1
0

runtime.spec.ts 294 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953195419551956195719581959196019611962196319641965196619671968196919701971197219731974197519761977197819791980198119821983198419851986198719881989199019911992199319941995199619971998199920002001200220032004200520062007200820092010201120122013201420152016201720182019202020212022202320242025202620272028202920302031203220332034203520362037203820392040204120422043204420452046204720482049205020512052205320542055205620572058205920602061206220632064206520662067206820692070207120722073207420752076207720782079208020812082208320842085208620872088208920902091209220932094209520962097209820992100210121022103210421052106210721082109211021112112211321142115211621172118211921202121212221232124212521262127212821292130213121322133213421352136213721382139214021412142214321442145214621472148214921502151215221532154215521562157215821592160216121622163216421652166216721682169217021712172217321742175217621772178217921802181218221832184218521862187218821892190219121922193219421952196219721982199220022012202220322042205220622072208220922102211221222132214221522162217221822192220222122222223222422252226222722282229223022312232223322342235223622372238223922402241224222432244224522462247224822492250225122522253225422552256225722582259226022612262226322642265226622672268226922702271227222732274227522762277227822792280228122822283228422852286228722882289229022912292229322942295229622972298229923002301230223032304230523062307230823092310231123122313231423152316231723182319232023212322232323242325232623272328232923302331233223332334233523362337233823392340234123422343234423452346234723482349235023512352235323542355235623572358235923602361236223632364236523662367236823692370237123722373237423752376237723782379238023812382238323842385238623872388238923902391239223932394239523962397239823992400240124022403240424052406240724082409241024112412241324142415241624172418241924202421242224232424242524262427242824292430243124322433243424352436243724382439244024412442244324442445244624472448244924502451245224532454245524562457245824592460246124622463246424652466246724682469247024712472247324742475247624772478247924802481248224832484248524862487248824892490249124922493249424952496249724982499250025012502250325042505250625072508250925102511251225132514251525162517251825192520252125222523252425252526252725282529253025312532253325342535253625372538253925402541254225432544254525462547254825492550255125522553255425552556255725582559256025612562256325642565256625672568256925702571257225732574257525762577257825792580258125822583258425852586258725882589259025912592259325942595259625972598259926002601260226032604260526062607260826092610261126122613261426152616261726182619262026212622262326242625262626272628262926302631263226332634263526362637263826392640264126422643264426452646264726482649265026512652265326542655265626572658265926602661266226632664266526662667266826692670267126722673267426752676267726782679268026812682268326842685268626872688268926902691269226932694269526962697269826992700270127022703270427052706270727082709271027112712271327142715271627172718271927202721272227232724272527262727272827292730273127322733273427352736273727382739274027412742274327442745274627472748274927502751275227532754275527562757275827592760276127622763276427652766276727682769277027712772277327742775277627772778277927802781278227832784278527862787278827892790279127922793279427952796279727982799280028012802280328042805280628072808280928102811281228132814281528162817281828192820282128222823282428252826282728282829283028312832283328342835283628372838283928402841284228432844284528462847284828492850285128522853285428552856285728582859286028612862286328642865286628672868286928702871287228732874287528762877287828792880288128822883288428852886288728882889289028912892289328942895289628972898289929002901290229032904290529062907290829092910291129122913291429152916291729182919292029212922292329242925292629272928292929302931293229332934293529362937293829392940294129422943294429452946294729482949295029512952295329542955295629572958295929602961296229632964296529662967296829692970297129722973297429752976297729782979298029812982298329842985298629872988298929902991299229932994299529962997299829993000300130023003300430053006300730083009301030113012301330143015301630173018301930203021302230233024302530263027302830293030303130323033303430353036303730383039304030413042304330443045304630473048304930503051305230533054305530563057305830593060306130623063306430653066306730683069307030713072307330743075307630773078307930803081308230833084308530863087308830893090309130923093309430953096309730983099310031013102310331043105310631073108310931103111311231133114311531163117311831193120312131223123312431253126312731283129313031313132313331343135313631373138313931403141314231433144314531463147314831493150315131523153315431553156315731583159316031613162316331643165316631673168316931703171317231733174317531763177317831793180318131823183318431853186318731883189319031913192319331943195319631973198319932003201320232033204320532063207320832093210321132123213321432153216321732183219322032213222322332243225322632273228322932303231323232333234323532363237323832393240324132423243324432453246324732483249325032513252325332543255325632573258325932603261326232633264326532663267326832693270327132723273327432753276327732783279328032813282328332843285328632873288328932903291329232933294329532963297329832993300330133023303330433053306330733083309331033113312331333143315331633173318331933203321332233233324332533263327332833293330333133323333333433353336333733383339334033413342334333443345334633473348334933503351335233533354335533563357335833593360336133623363336433653366336733683369337033713372337333743375337633773378337933803381338233833384338533863387338833893390339133923393339433953396339733983399340034013402340334043405340634073408340934103411341234133414341534163417341834193420342134223423342434253426342734283429343034313432343334343435343634373438343934403441344234433444344534463447344834493450345134523453345434553456345734583459346034613462346334643465346634673468346934703471347234733474347534763477347834793480348134823483348434853486348734883489349034913492349334943495349634973498349935003501350235033504350535063507350835093510351135123513351435153516351735183519352035213522352335243525352635273528352935303531353235333534353535363537353835393540354135423543354435453546354735483549355035513552355335543555355635573558355935603561356235633564356535663567356835693570357135723573357435753576357735783579358035813582358335843585358635873588358935903591359235933594359535963597359835993600360136023603360436053606360736083609361036113612361336143615361636173618361936203621362236233624362536263627362836293630363136323633363436353636363736383639364036413642364336443645364636473648364936503651365236533654365536563657365836593660366136623663366436653666366736683669367036713672367336743675367636773678367936803681368236833684368536863687368836893690369136923693369436953696369736983699370037013702370337043705370637073708370937103711371237133714371537163717371837193720372137223723372437253726372737283729373037313732373337343735373637373738373937403741374237433744374537463747374837493750375137523753375437553756375737583759376037613762376337643765376637673768376937703771377237733774377537763777377837793780378137823783378437853786378737883789379037913792379337943795379637973798379938003801380238033804380538063807380838093810381138123813381438153816381738183819382038213822382338243825382638273828382938303831383238333834383538363837383838393840384138423843384438453846384738483849385038513852385338543855385638573858385938603861386238633864386538663867386838693870387138723873387438753876387738783879388038813882388338843885388638873888388938903891389238933894389538963897389838993900390139023903390439053906390739083909391039113912391339143915391639173918391939203921392239233924392539263927392839293930393139323933393439353936393739383939394039413942394339443945394639473948394939503951395239533954395539563957395839593960396139623963396439653966396739683969397039713972397339743975397639773978397939803981398239833984398539863987398839893990399139923993399439953996399739983999400040014002400340044005400640074008400940104011401240134014401540164017401840194020402140224023402440254026402740284029403040314032403340344035403640374038403940404041404240434044404540464047404840494050405140524053405440554056405740584059406040614062406340644065406640674068406940704071407240734074407540764077407840794080408140824083408440854086408740884089409040914092409340944095409640974098409941004101410241034104410541064107410841094110411141124113411441154116411741184119412041214122412341244125412641274128412941304131413241334134413541364137413841394140414141424143414441454146414741484149415041514152415341544155415641574158415941604161416241634164416541664167416841694170417141724173417441754176417741784179418041814182418341844185418641874188418941904191419241934194419541964197419841994200420142024203420442054206420742084209421042114212421342144215421642174218421942204221422242234224422542264227422842294230423142324233423442354236423742384239424042414242424342444245424642474248424942504251425242534254425542564257425842594260426142624263426442654266426742684269427042714272427342744275427642774278427942804281428242834284428542864287428842894290429142924293429442954296429742984299430043014302430343044305430643074308430943104311431243134314431543164317431843194320432143224323432443254326432743284329433043314332433343344335433643374338433943404341434243434344434543464347434843494350435143524353435443554356435743584359436043614362436343644365436643674368436943704371437243734374437543764377437843794380438143824383438443854386438743884389439043914392439343944395439643974398439944004401440244034404440544064407440844094410441144124413441444154416441744184419442044214422442344244425442644274428442944304431443244334434443544364437443844394440444144424443444444454446444744484449445044514452445344544455445644574458445944604461446244634464446544664467446844694470447144724473447444754476447744784479448044814482448344844485448644874488448944904491449244934494449544964497449844994500450145024503450445054506450745084509451045114512451345144515451645174518451945204521452245234524452545264527452845294530453145324533453445354536453745384539454045414542454345444545454645474548454945504551455245534554455545564557455845594560456145624563456445654566456745684569457045714572457345744575457645774578457945804581458245834584458545864587458845894590459145924593459445954596459745984599460046014602460346044605460646074608460946104611461246134614461546164617461846194620462146224623462446254626462746284629463046314632463346344635463646374638463946404641464246434644464546464647464846494650465146524653465446554656465746584659466046614662466346644665466646674668466946704671467246734674467546764677467846794680468146824683468446854686468746884689469046914692469346944695469646974698469947004701470247034704470547064707470847094710471147124713471447154716471747184719472047214722472347244725472647274728472947304731473247334734473547364737473847394740474147424743474447454746474747484749475047514752475347544755475647574758475947604761476247634764476547664767476847694770477147724773477447754776477747784779478047814782478347844785478647874788478947904791479247934794479547964797479847994800480148024803480448054806480748084809481048114812481348144815481648174818481948204821482248234824482548264827482848294830483148324833483448354836483748384839484048414842484348444845484648474848484948504851485248534854485548564857485848594860486148624863486448654866486748684869487048714872487348744875487648774878487948804881488248834884488548864887488848894890489148924893489448954896489748984899490049014902490349044905490649074908490949104911491249134914491549164917491849194920492149224923492449254926492749284929493049314932493349344935493649374938493949404941494249434944494549464947494849494950495149524953495449554956495749584959496049614962496349644965496649674968496949704971497249734974497549764977497849794980498149824983498449854986498749884989499049914992499349944995499649974998499950005001500250035004500550065007500850095010501150125013501450155016501750185019502050215022502350245025502650275028502950305031503250335034503550365037503850395040504150425043504450455046504750485049505050515052505350545055505650575058505950605061506250635064506550665067506850695070507150725073507450755076507750785079508050815082508350845085508650875088508950905091509250935094509550965097509850995100510151025103510451055106510751085109511051115112511351145115511651175118511951205121512251235124512551265127512851295130513151325133513451355136513751385139514051415142514351445145514651475148514951505151515251535154515551565157515851595160516151625163516451655166516751685169517051715172517351745175517651775178517951805181518251835184518551865187518851895190519151925193519451955196519751985199520052015202520352045205520652075208520952105211521252135214521552165217521852195220522152225223522452255226522752285229523052315232523352345235523652375238523952405241524252435244524552465247524852495250525152525253525452555256525752585259526052615262526352645265526652675268526952705271527252735274527552765277527852795280528152825283528452855286528752885289529052915292529352945295529652975298529953005301530253035304530553065307530853095310531153125313531453155316531753185319532053215322532353245325532653275328532953305331533253335334533553365337533853395340534153425343534453455346534753485349535053515352535353545355535653575358535953605361536253635364536553665367536853695370537153725373537453755376537753785379538053815382538353845385538653875388538953905391539253935394539553965397539853995400540154025403540454055406540754085409541054115412541354145415541654175418541954205421542254235424542554265427542854295430543154325433543454355436543754385439544054415442544354445445544654475448544954505451545254535454545554565457545854595460546154625463546454655466546754685469547054715472547354745475547654775478547954805481548254835484548554865487548854895490549154925493549454955496549754985499550055015502550355045505550655075508550955105511551255135514551555165517551855195520552155225523552455255526552755285529553055315532553355345535553655375538553955405541554255435544554555465547554855495550555155525553555455555556555755585559556055615562556355645565556655675568556955705571557255735574557555765577557855795580558155825583558455855586558755885589559055915592559355945595559655975598559956005601560256035604560556065607560856095610561156125613561456155616561756185619562056215622562356245625562656275628562956305631563256335634563556365637563856395640564156425643564456455646564756485649565056515652565356545655565656575658565956605661566256635664566556665667566856695670567156725673567456755676567756785679568056815682568356845685568656875688568956905691569256935694569556965697569856995700570157025703570457055706570757085709571057115712571357145715571657175718571957205721572257235724572557265727572857295730573157325733573457355736573757385739574057415742574357445745574657475748574957505751575257535754575557565757575857595760576157625763576457655766576757685769577057715772577357745775577657775778577957805781578257835784578557865787578857895790579157925793579457955796579757985799580058015802580358045805580658075808580958105811581258135814581558165817581858195820582158225823582458255826582758285829583058315832583358345835583658375838583958405841584258435844584558465847584858495850585158525853585458555856585758585859586058615862586358645865586658675868586958705871587258735874587558765877587858795880588158825883588458855886588758885889589058915892589358945895589658975898589959005901590259035904590559065907590859095910591159125913591459155916591759185919592059215922592359245925592659275928592959305931593259335934593559365937593859395940594159425943594459455946594759485949595059515952595359545955595659575958595959605961596259635964596559665967596859695970597159725973597459755976597759785979598059815982598359845985598659875988598959905991599259935994599559965997599859996000600160026003600460056006600760086009601060116012601360146015601660176018601960206021602260236024602560266027602860296030603160326033603460356036603760386039604060416042604360446045604660476048604960506051605260536054605560566057605860596060606160626063606460656066606760686069607060716072607360746075607660776078607960806081608260836084608560866087608860896090609160926093609460956096609760986099610061016102610361046105610661076108610961106111611261136114611561166117611861196120
  1. import { execFileSync } from 'node:child_process'
  2. import { existsSync, mkdtempSync, realpathSync, rmSync, statSync, writeFileSync } from 'node:fs'
  3. import { mkdtemp, writeFile } from 'node:fs/promises'
  4. import { tmpdir } from 'node:os'
  5. import { basename, dirname, join, relative, resolve } from 'node:path'
  6. import { afterEach, describe, expect, it, vi } from 'vitest'
  7. import { Context } from '@deepseek-ai/cordis'
  8. import { PythonCodeRuntime, hostFrameParseCeiling, readProcessStart, resolvePythonBin } from '../src/index.ts'
  9. import { logTruncationMarker } from '../src/protocol.ts'
  10. import type { Config } from '../src/index.ts'
  11. // Absolute supported interpreter path for shell wrappers. The runtime gives a
  12. // child only TMPDIR, so a bare `python3` inside a wrapper would resolve against
  13. // /bin/sh's default PATH rather than the caller's selected interpreter.
  14. const PYABS = resolvePythonBin('python3') ?? 'python3'
  15. import type { CodeBindingFunction, CodeJsonValue, CodeRunResult } from '@deepseek-ai/dsh-code-runtime'
  16. /**
  17. * Names one `py/` script whose `copyFileSync` must fail, for the partial-staging
  18. * case. A real disk-full or missing-asset failure mid-copy cannot be produced
  19. * from a test, and the leak only shows when `mkdtempSync` has already succeeded.
  20. *
  21. * `stagedDirs` records every staging directory THIS test file creates, so the
  22. * leak assertions check the exact paths instead of a global tmpdir diff: a
  23. * parallel vitest worker running the same prefix could create or remove
  24. * `dsh-code-runtime-python-*` directories inside the sampling window, which a
  25. * readdir diff would misattribute to this test. `boot-write-failure.spec.ts`
  26. * records the same race and solves it with argv-based identity; recording the
  27. * mkdtempSync results is the fs-mock equivalent.
  28. */
  29. const { failNextCopyOf, stagedDirs, tempDirs, tempFiles } = vi.hoisted(() => ({
  30. failNextCopyOf: { value: undefined as string | undefined },
  31. stagedDirs: [] as string[],
  32. // Test-created temp dirs/files, registered by the helpers below and removed
  33. // after each test: a suite run over real python3 subprocesses must not
  34. // permanently accumulate `dsh-*` fixtures in the shared tmpdir (the runtime
  35. // cleans its own per-run staging dir; these are the stubs and wrappers the
  36. // tests themselves build).
  37. tempDirs: [] as string[],
  38. tempFiles: [] as string[],
  39. }))
  40. vi.mock('node:fs', async (importOriginal) => {
  41. const actual = await importOriginal<typeof import('node:fs')>()
  42. return {
  43. ...actual,
  44. copyFileSync(source: string, destination: string): void {
  45. if (failNextCopyOf.value !== undefined && basename(source) === failNextCopyOf.value) {
  46. failNextCopyOf.value = undefined
  47. throw Object.assign(new Error('simulated ENOSPC on copy'), { code: 'ENOSPC' })
  48. }
  49. actual.copyFileSync(source, destination)
  50. },
  51. mkdtempSync(prefix: string): string {
  52. const dir = actual.mkdtempSync(prefix)
  53. if (basename(prefix).startsWith('dsh-code-runtime-python-')) stagedDirs.push(dir)
  54. return dir
  55. },
  56. }
  57. })
  58. /**
  59. * Integration suite over REAL python3 subprocesses (no subprocess mocks — it is
  60. * cheap and local, per docs/testing.md's real-over-mock policy; the only mock is
  61. * `node:fs.copyFileSync` for the staging-failure cases). Each test builds a fresh
  62. * runtime so budgets can be tuned per case.
  63. */
  64. async function setup(config: Config = {}) {
  65. const ctx = new Context()
  66. const fiber = await ctx.plugin(PythonCodeRuntime, config)
  67. const runtime = ctx.codeRuntime as PythonCodeRuntime
  68. return { ctx, fiber, runtime }
  69. }
  70. /** Convenience: one namespace `tools` with the given functions. */
  71. function tools(functions: Record<string, CodeBindingFunction>) {
  72. return [{ global: 'tools', functions }]
  73. }
  74. /** Create a test temp dir registered for afterEach removal. */
  75. async function makeTempDir(prefix: string): Promise<string> {
  76. const dir = await mkdtemp(join(tmpdir(), prefix))
  77. tempDirs.push(dir)
  78. return dir
  79. }
  80. /** Synchronous variant of {@link makeTempDir} for the PATH-stub fixtures. */
  81. function makeTempDirSync(prefix: string): string {
  82. const dir = mkdtempSync(join(tmpdir(), prefix))
  83. tempDirs.push(dir)
  84. return dir
  85. }
  86. // Remove every fixture this file created, so repeated runs do not accumulate
  87. // `dsh-*` directories and wrappers in the shared tmpdir.
  88. afterEach(() => {
  89. for (const dir of tempDirs.splice(0)) rmSync(dir, { recursive: true, force: true })
  90. for (const file of tempFiles.splice(0)) rmSync(file, { force: true })
  91. })
  92. describe('PythonCodeRuntime — seam descriptors and misuse', () => {
  93. it('registers the seam descriptors', async () => {
  94. const { runtime } = await setup()
  95. expect(runtime.language).toBe('python')
  96. expect(runtime.isolation).toBe('process')
  97. })
  98. it('rejects non-positive config as seam misuse', async () => {
  99. const ctx = new Context()
  100. await expect(ctx.plugin(PythonCodeRuntime, { cpuSeconds: 0 }))
  101. .rejects.toThrow(/cpuSeconds must be a positive number/)
  102. await expect(ctx.plugin(PythonCodeRuntime, { maxWallMs: -1 }))
  103. .rejects.toThrow(/maxWallMs must be a positive number/)
  104. })
  105. it('rejects a non-integer cpuSeconds at load (setrlimit needs an int)', async () => {
  106. const ctx = new Context()
  107. await expect(ctx.plugin(PythonCodeRuntime, { cpuSeconds: 1.5 }))
  108. .rejects.toThrow(/cpuSeconds must be a positive integer, got 1.5/)
  109. })
  110. it('rejects a non-integer byte budget at load (the child int()-truncates it)', async () => {
  111. // maxLogBytes/maxValueBytes cross to the child, which reads them through
  112. // int(...): a float would floor there while the host meters the fraction, so
  113. // the two sides would enforce different public config. Reject at load.
  114. const ctxLog = new Context()
  115. await expect(ctxLog.plugin(PythonCodeRuntime, { maxLogBytes: 3.5 }))
  116. .rejects.toThrow(/maxLogBytes must be a positive integer/)
  117. const ctxValue = new Context()
  118. await expect(ctxValue.plugin(PythonCodeRuntime, { maxValueBytes: 1024.5 }))
  119. .rejects.toThrow(/maxValueBytes must be a positive integer/)
  120. })
  121. it('rejects finite numeric config that cannot cross as an exact rlimit integer', async () => {
  122. // `Number.isFinite` and `Number.isInteger` both admit values that cannot
  123. // round-trip. `addressSpaceMb: 1e308` overflows to `Infinity` once multiplied
  124. // by 1 MiB, and `encodeJsonPlain` renders that as `null`, so the child gets no
  125. // limit at all; `cpuSeconds: 1e100` clears `Number.isInteger` while sitting
  126. // far past the safe range, so `setrlimit` receives a different number than was
  127. // configured. Both used to end every run in a bootstrap exception instead of
  128. // failing at load, where a self-contained configuration error belongs.
  129. const ctx = new Context()
  130. await expect(ctx.plugin(PythonCodeRuntime, { addressSpaceMb: 1e308 }))
  131. .rejects.toThrow(/addressSpaceMb must be at most \d+ .*exact integer/)
  132. await expect(ctx.plugin(PythonCodeRuntime, { cpuSeconds: 1e100 }))
  133. .rejects.toThrow(/cpuSeconds must be at most \d+ .*exact integers/)
  134. // The boundary values still load: the bound rejects what cannot be encoded,
  135. // not everything large.
  136. const okMb = await ctx.plugin(PythonCodeRuntime, { addressSpaceMb: Math.floor(Number.MAX_SAFE_INTEGER / (1024 * 1024)) })
  137. await okMb.dispose()
  138. const okCpu = await ctx.plugin(PythonCodeRuntime, { cpuSeconds: Number.MAX_SAFE_INTEGER - 1 })
  139. await okCpu.dispose()
  140. })
  141. it('rejects an output cap whose payload could not cross the frame ceiling', async () => {
  142. // The caps budget a payload that must arrive inside ONE fd-3 frame, and the
  143. // 64 MiB frame parse cap is fixed. A larger cap is unsatisfiable rather
  144. // than generous: a completion the cap admits arrives as an over-ceiling
  145. // frame and fails the run as `worker-exit`, inverting the `output-limit`
  146. // the cap describes. Both budgets are metered in already-escaped serialized
  147. // bytes, so a payload occupies at most `cap + envelope` on the wire; the
  148. // bound is `parse-cap - envelope`, not `(ceiling - envelope) / 6` (that
  149. // divided in escape expansion the charge already counts). The receive path
  150. // rejects raw frames past the 64 MiB parse cap (the run settles as a
  151. // worker-exit), so a budget above it would admit a config whose honest
  152. // child frames the host then rejects.
  153. const admissible = 64 * 1024 * 1024 - 64
  154. const ctx = new Context()
  155. await expect(ctx.plugin(PythonCodeRuntime, { maxLogBytes: admissible + 1 }))
  156. .rejects.toThrow(/maxLogBytes must not exceed 67108800/)
  157. await expect(ctx.plugin(PythonCodeRuntime, { maxValueBytes: admissible + 1 }))
  158. .rejects.toThrow(/maxValueBytes must not exceed 67108800/)
  159. // The boundary value itself loads: the bound is the largest cap a frame can
  160. // still carry, not one below it. It needs an address space large enough to
  161. // clear the separate maxValueBytes/addressSpaceMb worst-case gate (the cap
  162. // times the 12x Unicode expansion must fit), so this pairs it with a 4 GiB
  163. // addressSpaceMb — the two load-time bounds are independent.
  164. const boundary = await ctx.plugin(PythonCodeRuntime, { maxValueBytes: admissible, addressSpaceMb: 4096 })
  165. await boundary.dispose()
  166. })
  167. it('rejects a completion budget whose frame a constrained host heap cannot safely parse', async () => {
  168. // The load gate bounds the CHILD's build-and-encode under RLIMIT_AS; it
  169. // does not bound the HOST's JSON.parse, which materializes several times a
  170. // wide frame's raw bytes in property storage. In a child node with a
  171. // 128 MiB old space the heap-derived frame cap is ~7 MiB, so a 50 MiB
  172. // budget is rejected at load even though the address-space gate alone
  173. // would admit it (50 MiB * 12 = 600 MiB < 1 GiB - 64 MiB).
  174. const script = [
  175. "import { Context } from '@deepseek-ai/cordis'",
  176. "import { PythonCodeRuntime } from './packages/experimental/code-runtime-python/src/index.ts'",
  177. 'const ctx = new Context()',
  178. 'try {',
  179. ' await ctx.plugin(PythonCodeRuntime, { maxValueBytes: 50 * 1024 * 1024, addressSpaceMb: 1024 })',
  180. " console.log('LOADED')",
  181. ' process.exit(1)',
  182. '} catch (error) {',
  183. " console.log('REJECTED:' + (error instanceof Error ? error.message : String(error)))",
  184. ' process.exit(0)',
  185. '}',
  186. ].join('\n')
  187. const out = execFileSync(process.execPath, ['--max-old-space-size=128', '--import', 'tsx', '-e', script], {
  188. cwd: resolve(import.meta.dirname, '../../../..'),
  189. encoding: 'utf8',
  190. timeout: 60_000,
  191. env: { ...process.env, TSX_TSCONFIG_PATH: resolve(import.meta.dirname, '../../../../tsconfig.json') },
  192. })
  193. expect(out).toContain('REJECTED:')
  194. expect(out).toContain('must not exceed')
  195. }, 60_000)
  196. it('parses a worst-shape frame at the derived cap on a constrained heap', async () => {
  197. // The host-heap frame cap must be measured against the WORST parse shape —
  198. // a dict of many short unique keys, which forces dictionary-mode property
  199. // storage plus interned keys (~6.4x at 3M keys, trending up), not the ~3x
  200. // of a repeated-key dict. A child node with a 128 MiB old space (~176 MiB
  201. // heap limit) derives a cap of floor((176 - 64) / 16) = 7 MiB; the
  202. // subprocess builds a unique-key dict whose frame is AT that cap and
  203. // parses it, which must survive. Verified fail-before: with the multiple
  204. // at 8 the derived cap doubles to 14 MiB and the same subprocess OOMs
  205. // during the parse (plain JS, no tsx — the frame and parse are builtins).
  206. const cap = hostFrameParseCeiling(176 * 1024 * 1024)
  207. const script = [
  208. `const cap = ${cap}`,
  209. // Each entry "k<base36>:1," is ~9-12 raw bytes; a few hundred thousand
  210. // unique keys put the frame just at the cap.
  211. 'const count = Math.floor(cap / 12)',
  212. 'const obj = {}',
  213. 'for (let i = 0; i < count; i++) obj[`k${i.toString(36)}`] = 1',
  214. 'const frame = JSON.stringify(obj)',
  215. "if (Buffer.byteLength(frame, 'utf8') > cap) throw new Error('frame over cap: ' + frame.length)",
  216. 'JSON.parse(frame)',
  217. "console.log('SURVIVED:' + Buffer.byteLength(frame, 'utf8'))",
  218. ].join('\n')
  219. const out = execFileSync(process.execPath, ['--max-old-space-size=128', '-e', script], {
  220. encoding: 'utf8',
  221. timeout: 60_000,
  222. })
  223. expect(out).toContain('SURVIVED:')
  224. }, 60_000)
  225. it('rejects a pythonBin that spawn() would throw on, at load', async () => {
  226. // Both values pass the string schema and both make `spawn` throw
  227. // SYNCHRONOUSLY from inside run() — ERR_INVALID_ARG_VALUE for the empty
  228. // path, ERR_INVALID_ARG_TYPE for the NUL — so run() would REJECT instead of
  229. // resolving the worker-exit the seam promises for a child that cannot
  230. // start. Both are self-contained configuration errors, so they fail here.
  231. const ctx = new Context()
  232. await expect(ctx.plugin(PythonCodeRuntime, { pythonBin: '' }))
  233. .rejects.toThrow(/pythonBin must be a non-empty path without NUL bytes/)
  234. await expect(ctx.plugin(PythonCodeRuntime, { pythonBin: 'py\u0000thon3' }))
  235. .rejects.toThrow(/pythonBin must be a non-empty path without NUL bytes/)
  236. })
  237. it('rejects an explicit pythonBin that is not an executable regular file, at load', async () => {
  238. // An explicit path (absolute, or containing a slash) bypasses PATH lookup,
  239. // so it must be validated directly: missing, non-executable, or directory
  240. // paths are self-contained configuration errors that used to slip through
  241. // load and surface only at the first run() as a misleading worker-exit.
  242. // The message distinguishes the explicit-path failure from a basename that
  243. // simply does not resolve on PATH.
  244. const nodePath = await import('node:path')
  245. const { writeFileSync, mkdirSync } = await import('node:fs')
  246. const dir = makeTempDirSync('dsh-bad-bin-')
  247. const notExecutable = nodePath.join(dir, 'not-executable')
  248. writeFileSync(notExecutable, '#!/bin/sh\nexit 0\n') // Regular file, but no X bit.
  249. const directory = nodePath.join(dir, 'is-a-directory')
  250. mkdirSync(directory)
  251. try {
  252. const missing = new Context()
  253. await expect(missing.plugin(PythonCodeRuntime, { pythonBin: nodePath.join(dir, 'missing') }))
  254. .rejects.toThrow(/is not an executable regular file/)
  255. const noX = new Context()
  256. await expect(noX.plugin(PythonCodeRuntime, { pythonBin: notExecutable }))
  257. .rejects.toThrow(/is not an executable regular file/)
  258. const isDir = new Context()
  259. await expect(isDir.plugin(PythonCodeRuntime, { pythonBin: directory }))
  260. .rejects.toThrow(/is not an executable regular file/)
  261. // A relative explicit path fails the same way, resolved against the host
  262. // CWD: `dir` is absolute, so a slash-containing relative form of it is
  263. // the dirname prefix plus the file, which does not exist as such.
  264. const rel = new Context()
  265. await expect(rel.plugin(PythonCodeRuntime, { pythonBin: './definitely-not-there-python' }))
  266. .rejects.toThrow(/is not an executable regular file/)
  267. } finally {
  268. const { rmSync } = await import('node:fs')
  269. rmSync(dir, { recursive: true, force: true })
  270. }
  271. })
  272. it('rejects a non-CPython, outdated, or probe-failing interpreter at load', async () => {
  273. const nonPython = new Context()
  274. await expect(nonPython.plugin(PythonCodeRuntime, { pythonBin: '/bin/echo' }))
  275. .rejects.toThrow(/did not report a CPython version/)
  276. const dir = await mkdtemp(join(tmpdir(), 'dsh-python-probe-'))
  277. const oldMajor = join(dir, 'python-old-major')
  278. const old = join(dir, 'python-old')
  279. const future = join(dir, 'python-future')
  280. const pypy = join(dir, 'pypy')
  281. const failed = join(dir, 'python-failed')
  282. await writeFile(oldMajor, '#!/bin/sh\nprintf \'cpython 2 99 0\\n\'\n', { mode: 0o755 })
  283. await writeFile(old, '#!/bin/sh\nprintf \'cpython 3 9 6\\n\'\n', { mode: 0o755 })
  284. await writeFile(future, '#!/bin/sh\nprintf \'cpython 4 0 0\\n\'\n', { mode: 0o755 })
  285. await writeFile(pypy, '#!/bin/sh\nprintf \'pypy 3 10 0\\n\'\n', { mode: 0o755 })
  286. await writeFile(failed, '#!/bin/sh\nexit 7\n', { mode: 0o755 })
  287. try {
  288. expect(resolvePythonBin(relative(process.cwd(), old))).toBe(old)
  289. const obsolete = new Context()
  290. await expect(obsolete.plugin(PythonCodeRuntime, { pythonBin: oldMajor }))
  291. .rejects.toThrow(/must be CPython 3\.10 or newer, got cpython 2\.99\.0/)
  292. const outdated = new Context()
  293. await expect(outdated.plugin(PythonCodeRuntime, { pythonBin: old }))
  294. .rejects.toThrow(/must be CPython 3\.10 or newer, got cpython 3\.9\.6/)
  295. const forwardCompatible = new Context()
  296. const fiber = await forwardCompatible.plugin(PythonCodeRuntime, { pythonBin: future })
  297. await fiber.dispose()
  298. const alternative = new Context()
  299. await expect(alternative.plugin(PythonCodeRuntime, { pythonBin: pypy }))
  300. .rejects.toThrow(/must be CPython, got pypy/)
  301. const probeFailure = new Context()
  302. await expect(probeFailure.plugin(PythonCodeRuntime, { pythonBin: failed }))
  303. .rejects.toThrow(/failed the CPython version probe/)
  304. } finally {
  305. rmSync(dir, { recursive: true, force: true })
  306. }
  307. })
  308. it('keeps an explicit executable pythonBin working through load and run', async () => {
  309. // The same validation that rejects bad explicit paths must admit a good
  310. // one: an absolute path to the real interpreter (or a wrapper around it)
  311. // is the deployment form the validation exists to serve.
  312. const pyAbs = resolvePythonBin('python3') ?? 'python3'
  313. const { runtime, fiber } = await setup({ pythonBin: pyAbs, maxWallMs: 30_000 })
  314. const result = await runtime.run({ program: 'return 1', bindings: [] })
  315. expect(result.error).toBeUndefined()
  316. expect(result.value).toBe(1)
  317. await fiber.dispose()
  318. })
  319. it('rejects a binding member accessor that throws, as seam misuse', async () => {
  320. // `namespace.functions` is caller-supplied, so its members may come from a
  321. // getter or Proxy. Reading one of them inside the fd-3 `data` callback used
  322. // to throw OUTSIDE the dispatcher's try and terminate the host; the
  323. // validation now snapshots the callables synchronously, so the throw
  324. // surfaces as the seam-misuse rejection run() reserves for malformed
  325. // bindings — the child is never spawned.
  326. const { runtime } = await setup()
  327. const exploding = {
  328. get explode(): CodeBindingFunction {
  329. throw new Error('getter blew up')
  330. },
  331. }
  332. await expect(runtime.run({
  333. program: 'return 1',
  334. bindings: [{ global: 'tools', functions: exploding }],
  335. })).rejects.toThrow(/getter blew up/)
  336. })
  337. it('snapshots binding callables once, so a getter is read exactly once', async () => {
  338. // The snapshot also fixes the key set the boot frame advertises: the child
  339. // learns the namespace names from the SAME record dispatch reads, so a
  340. // getter whose keys differ between reads cannot desynchronize the two.
  341. let reads = 0
  342. const countReads = {
  343. get first(): CodeBindingFunction {
  344. reads += 1
  345. return async () => 1
  346. },
  347. }
  348. const { runtime, fiber } = await setup()
  349. const result = await runtime.run({
  350. program: 'return 1',
  351. bindings: [{ global: 'tools', functions: countReads }],
  352. })
  353. expect(result.error).toBeUndefined()
  354. // One read for the validation snapshot; the boot frame and every dispatch
  355. // read the snapshot, not the getter.
  356. expect(reads).toBe(1)
  357. await fiber.dispose()
  358. })
  359. it('keeps a __proto__ binding member dispatchable', async () => {
  360. // The seam contract treats member names like `__proto__` or `constructor`
  361. // as ordinary own properties (null-prototype construction). The binding
  362. // snapshot must preserve that: a plain `{}` record would hit the prototype
  363. // setter on assignment and drop the member, so the child would never learn
  364. // the name and a call to it would fail with KeyError.
  365. const { runtime, fiber } = await setup()
  366. const result = await runtime.run({
  367. program: 'return await tools["__proto__"]({})',
  368. bindings: [{
  369. global: 'tools',
  370. functions: { ['__proto__']: async () => 'proto-callable' },
  371. }],
  372. })
  373. expect(result.error).toBeUndefined()
  374. expect(result.value).toBe('proto-callable')
  375. await fiber.dispose()
  376. })
  377. it('resolves pythonBin once so a later PATH change cannot switch interpreters', async () => {
  378. const firstDir = await mkdtemp(join(tmpdir(), 'dsh-python-first-'))
  379. const secondDir = await mkdtemp(join(tmpdir(), 'dsh-python-second-'))
  380. const wrapper = (marker: string): string => `#!/bin/sh\nDSH_TEST_PYTHON=${marker}\nexport DSH_TEST_PYTHON\nexec "${PYABS}" "$@"\n`
  381. await writeFile(join(firstDir, 'python3'), wrapper('first'), { mode: 0o755 })
  382. await writeFile(join(secondDir, 'python3'), wrapper('second'), { mode: 0o755 })
  383. vi.stubEnv('PATH', firstDir)
  384. let fiber: Awaited<ReturnType<typeof setup>>['fiber'] | undefined
  385. try {
  386. const mounted = await setup({ pythonBin: 'python3' })
  387. fiber = mounted.fiber
  388. vi.stubEnv('PATH', secondDir)
  389. const result = await mounted.runtime.run({
  390. program: 'import os\nreturn os.environ.get("DSH_TEST_PYTHON")',
  391. bindings: [],
  392. })
  393. expect(result.error).toBeUndefined()
  394. expect(result.value).toBe('first')
  395. } finally {
  396. await fiber?.dispose()
  397. vi.unstubAllEnvs()
  398. rmSync(firstDir, { recursive: true, force: true })
  399. rmSync(secondDir, { recursive: true, force: true })
  400. }
  401. })
  402. it('skips relative PATH entries when resolving a basename pythonBin', async () => {
  403. // resolvePythonBin must return an absolute path: a RELATIVE PATH entry
  404. // ('.' here) would otherwise resolve the basename against the host CWD.
  405. // This run's CWD holds no executable named python3, so both the relative
  406. // skip and the accessSync-miss fall through to the absolute entry — the
  407. // case pins the contract (absolute candidate wins over a relative PATH
  408. // prefix), not a worker-exit distinction, which would need an executable
  409. // named python3 in the test CWD.
  410. const cp = await import('node:child_process')
  411. const nodePath = await import('node:path')
  412. const pythonDir = nodePath.dirname(cp.execFileSync('which', ['python3'], { encoding: 'utf8' }).trim())
  413. vi.stubEnv('PATH', `.:${pythonDir}`)
  414. try {
  415. const { runtime, fiber } = await setup({ pythonBin: 'python3', maxWallMs: 30_000 })
  416. const result = await runtime.run({ program: 'return 1', bindings: [] })
  417. expect(result.error).toBeUndefined()
  418. expect(result.value).toBe(1)
  419. await fiber.dispose()
  420. } finally {
  421. vi.unstubAllEnvs()
  422. }
  423. }, 45_000)
  424. it('ignores a forged second boot-ack without re-sending the run frame', async () => {
  425. // The run frame is sent once, from the first boot-ack; a program that
  426. // forges an extra boot-ack frame on fd 3 must not re-enter the gate (a
  427. // second run frame would confuse the child's frame reader). The honest
  428. // child sends exactly one ack; the forged one exercises the re-entry
  429. // guard.
  430. const { runtime } = await setup()
  431. const result = await runtime.run({
  432. program: [
  433. 'import os',
  434. // One forged boot-ack after the program starts; the run already went
  435. // out on the real ack.
  436. "os.write(3, b'{\"type\":\"boot-ack\"}\\n')",
  437. 'return "done"',
  438. ].join('\n'),
  439. bindings: [],
  440. })
  441. expect(result.error).toBeUndefined()
  442. expect(result.value).toBe('done')
  443. }, 15_000)
  444. it('skips a PATH entry that is an executable DIRECTORY named like the interpreter', async () => {
  445. // accessSync(X_OK) succeeds on directories, so without the isFile guard a
  446. // PATH entry like a `python3` directory would be chosen over a later real
  447. // interpreter. The stub PATH puts such a directory first and asserts the
  448. // real interpreter is used.
  449. const cp = await import('node:child_process')
  450. const nodePath = await import('node:path')
  451. const { mkdirSync } = await import('node:fs')
  452. const realPythonDir = nodePath.dirname(cp.execFileSync('which', ['python3'], { encoding: 'utf8' }).trim())
  453. const fakeDir = makeTempDirSync('dsh-fake-bin-')
  454. mkdirSync(nodePath.join(fakeDir, 'python3')) // A directory named python3, executable by default.
  455. vi.stubEnv('PATH', `${fakeDir}:${realPythonDir}`)
  456. try {
  457. const { runtime, fiber } = await setup({ pythonBin: 'python3', maxWallMs: 30_000 })
  458. const result = await runtime.run({ program: 'return 1', bindings: [] })
  459. expect(result.error).toBeUndefined()
  460. expect(result.value).toBe(1)
  461. await fiber.dispose()
  462. } finally {
  463. vi.unstubAllEnvs()
  464. }
  465. }, 45_000)
  466. it('rejects a timer budget setTimeout would silently clamp to 1 ms', async () => {
  467. // Node stores a setTimeout delay as a signed 32-bit value and substitutes
  468. // 1 ms for anything larger, inverting the knob's meaning: a huge maxWallMs
  469. // would time every run out at once, and a huge graceMs would SIGKILL one
  470. // millisecond after SIGTERM. Both must fail at load instead.
  471. const ctx = new Context()
  472. await expect(ctx.plugin(PythonCodeRuntime, { maxWallMs: 2_147_483_648 }))
  473. .rejects.toThrow(/maxWallMs must not exceed 2147483647/)
  474. // graceMs is bounded by the close deadline's added margin, not by the raw
  475. // timer maximum, because that sum is what gets armed.
  476. await expect(ctx.plugin(PythonCodeRuntime, { graceMs: 2_147_481_648 }))
  477. .rejects.toThrow(/graceMs must not exceed 2147481647/)
  478. // The exact maxima still load.
  479. await expect(ctx.plugin(PythonCodeRuntime, { maxWallMs: 2_147_483_647, graceMs: 2_147_481_647 }))
  480. .resolves.toBeDefined()
  481. })
  482. it('rejects loading this Unix-only backend on Windows', async () => {
  483. // The bootstrap needs the POSIX `resource` module, a positional fd 3, and
  484. // negative-PID process-group signals — none on Windows. The constructor
  485. // must throw at load rather than register ctx.codeRuntime and defer the
  486. // failure to the first run.
  487. const original = process.platform
  488. Object.defineProperty(process, 'platform', { value: 'win32', configurable: true })
  489. try {
  490. const ctx = new Context()
  491. await expect(ctx.plugin(PythonCodeRuntime, {})).rejects.toThrow(/requires a Unix platform/)
  492. } finally {
  493. Object.defineProperty(process, 'platform', { value: original, configurable: true })
  494. }
  495. })
  496. it('rejects a binding global that is not a Python identifier or is reserved', async () => {
  497. const { runtime } = await setup()
  498. await expect(runtime.run({
  499. program: 'return 1',
  500. bindings: [{ global: '1bad', functions: {} }],
  501. })).rejects.toThrow(/is not a usable Python identifier/)
  502. await expect(runtime.run({
  503. program: 'return 1',
  504. bindings: [{ global: 'class', functions: {} }],
  505. })).rejects.toThrow(/is not a usable Python identifier/)
  506. })
  507. it('rejects duplicate binding namespaces', async () => {
  508. const { runtime } = await setup()
  509. await expect(runtime.run({
  510. program: 'return 1',
  511. bindings: [
  512. { global: 'tools', functions: {} },
  513. { global: 'tools', functions: {} },
  514. ],
  515. })).rejects.toThrow(/duplicate binding global/)
  516. })
  517. it('rejects run() after disposal, and unregisters ctx.codeRuntime', async () => {
  518. const { ctx, fiber, runtime } = await setup()
  519. await fiber.dispose()
  520. await expect(runtime.run({ program: 'return 1', bindings: [] }))
  521. .rejects.toThrow(/after disposal/)
  522. expect(ctx.get('codeRuntime')).toBeUndefined()
  523. })
  524. it('short-circuits when the request signal is already aborted', async () => {
  525. const { runtime } = await setup()
  526. const signal = AbortSignal.abort('already-cancelled')
  527. const result = await runtime.run({ program: 'return 1', bindings: [], signal })
  528. expect(result.error?.kind).toBe('abort')
  529. expect(result.error?.message).toContain('already-cancelled')
  530. expect(result.logs).toEqual([])
  531. })
  532. it('short-circuits on an already-aborted signal whose reason cannot be converted', async () => {
  533. // The pre-flight arm converted the reason with a bare `String()`, so a
  534. // hostile reason threw out of `run()` — the seam promises to reject only for
  535. // misuse, and a caller's cancellation token is not misuse.
  536. const { runtime } = await setup()
  537. const signal = AbortSignal.abort({
  538. [Symbol.toPrimitive]() { throw new Error('reason blew up') },
  539. })
  540. const result = await runtime.run({ program: 'return 1', bindings: [], signal })
  541. expect(result.error?.kind).toBe('abort')
  542. expect(result.error?.message).toBe('<unrenderable rejection value>')
  543. expect(result.logs).toEqual([])
  544. })
  545. it('runs the interpreter from materialized scripts outside the package, and removes them per run', async () => {
  546. // The interpreter is an EXTERNAL process, so it can only open paths the OS
  547. // resolves. Inside the single-file Python-SDK executable the packaged `py/`
  548. // directory lives in pkg's virtual filesystem, which Node reads through its
  549. // patched `fs` but `python3` cannot see, so spawning from that path fails
  550. // with ENOENT. The scripts are therefore copied to a real directory first.
  551. //
  552. // The path is read from the child's own `__main__` module, so it proves
  553. // where the interpreter actually loaded the entry script — asserting on a
  554. // host-side constant would only restate the source. The program namespace
  555. // seeds `__name__` but no `__file__`, hence the module lookup.
  556. // `protocol.py` must land in the SAME directory, since `bootstrap.py` puts
  557. // its own directory on `sys.path` to import it; the run completing at all
  558. // already exercises that import.
  559. const { runtime } = await setup()
  560. const entryOf = async (): Promise<string> => {
  561. const result = await runtime.run({ program: 'import sys\nreturn sys.modules["__main__"].__file__', bindings: [] })
  562. expect(result.error).toBeUndefined()
  563. return result.value as string
  564. }
  565. const entry = await entryOf()
  566. expect(entry.endsWith('/bootstrap.py')).toBe(true)
  567. const dir = dirname(entry)
  568. expect(realpathSync(dirname(dir))).toBe(realpathSync(tmpdir()))
  569. expect(basename(dir)).toMatch(/^dsh-code-runtime-python-/)
  570. expect(dir).not.toContain('/packages/')
  571. // Staging is per RUN and removed at settlement, so by the time `run()`
  572. // resolved the directory is already gone — nothing survives to be rewritten
  573. // by a later run. `protocol.py` had to be beside the entry script for the run
  574. // to complete at all, since `bootstrap.py` imports it off `sys.path`.
  575. expect(existsSync(dir)).toBe(false)
  576. // A second run stages its own copy rather than reusing the first.
  577. expect(dirname(await entryOf())).not.toBe(dir)
  578. })
  579. it('contains a program that rewrites its own bootstrap to the run that did it', async () => {
  580. // The child runs as the same UID as the host, so `0o700` does not stop model
  581. // code from rewriting the scripts it was started from —
  582. // `sys.modules['__main__'].__file__` names them. While all runs shared one
  583. // staged copy, a program that overwrote `bootstrap.py` broke the NEXT run
  584. // (measured: it settled as `worker-exit`), and substituted code would have
  585. // run before the resource limits were applied.
  586. const { runtime } = await setup({ maxWallMs: 10_000 })
  587. const sabotage = await runtime.run({
  588. program: [
  589. 'import sys',
  590. 'path = sys.modules["__main__"].__file__',
  591. 'open(path, "w").write("raise SystemExit(1)\\n")',
  592. 'return path',
  593. ].join('\n'),
  594. bindings: [],
  595. })
  596. expect(sabotage.error).toBeUndefined()
  597. // The damage stayed inside the run that caused it.
  598. const after = await runtime.run({ program: 'return 1 + 1', bindings: [] })
  599. expect(after.error).toBeUndefined()
  600. expect(after.value).toBe(2)
  601. }, 20_000)
  602. it('leaves no subprocess or scripts behind when disposal races the first run', async () => {
  603. // Staging runs SYNCHRONOUSLY so no async boundary opens between `run()` and
  604. // the point where `execute` registers the run in `live` and installs the
  605. // abort listener. With an `await` there, a disposal landing in that window
  606. // saw an empty `live`, returned, removed the script directory, and let the
  607. // continuation spawn a subprocess after the fiber was gone.
  608. //
  609. // `dispose()` is called in the same synchronous turn as `run()`, with no
  610. // `await` between them, so it lands exactly in that window.
  611. //
  612. // The leak assertion checks the EXACT paths this test file staged (recorded
  613. // by the mocked mkdtempSync) rather than diffing a global tmpdir: a
  614. // parallel vitest worker can create or remove same-prefix directories
  615. // inside the sampling window, which a readdir diff would misattribute to
  616. // this test (boot-write-failure.spec.ts records the same race).
  617. const stagedBefore = stagedDirs.length
  618. const { fiber, runtime } = await setup({ maxWallMs: 8_000 })
  619. const pending = runtime.run({ program: 'import time\nwhile True: time.sleep(0.1)', bindings: [] })
  620. const disposed = fiber.dispose()
  621. const result = await pending
  622. await disposed
  623. // Whatever the run reports, it must be terminal and must not be a success.
  624. expect(result.value).toBeUndefined()
  625. expect(['abort', 'worker-exit', 'timeout']).toContain(result.error?.kind)
  626. // Disposal is to quiescence, so every directory this run staged is gone.
  627. const created = stagedDirs.slice(stagedBefore)
  628. for (const dir of created) expect(existsSync(dir)).toBe(false)
  629. }, 15_000)
  630. it('settles as abort when the signal fires in the same turn as the first run', async () => {
  631. // Same window, the other listener. `addEventListener('abort')` does not
  632. // replay an event that already fired, so an abort landing before the
  633. // listener was installed used to be missed entirely and the program ran to
  634. // success or the wall ceiling instead of resolving as `abort`. Synchronous
  635. // staging keeps the pre-flight check and the listener in one turn, leaving
  636. // no gap for the signal to slip through.
  637. const { runtime } = await setup({ maxWallMs: 4_000, graceMs: 200 })
  638. const controller = new AbortController()
  639. const pending = runtime.run({
  640. program: 'import time\nwhile True: time.sleep(0.1)',
  641. bindings: [],
  642. signal: controller.signal,
  643. })
  644. controller.abort('same-turn-abort')
  645. const result = await pending
  646. expect(result.error?.kind).toBe('abort')
  647. expect(result.error?.message).toContain('same-turn-abort')
  648. }, 15_000)
  649. it('reports a staging failure as worker-exit instead of rejecting run()', async () => {
  650. // Staging touches the filesystem, so it can fail for reasons that are not
  651. // the caller's doing: a full or read-only temp filesystem, or a deployment
  652. // that failed to ship the packaged scripts. Those are SUBSTRATE failures,
  653. // the same class as a child that cannot start, and the seam reserves
  654. // rejection for misuse — so `run()` must resolve, not throw.
  655. //
  656. // `TMPDIR` is the honest lever: `mkdtempSync` builds its path from
  657. // `os.tmpdir()`, so pointing it at a path that is not a directory makes the
  658. // real call fail without stubbing the module under test.
  659. const previous = process.env.TMPDIR
  660. const notADirectory = join(await makeTempDir('dsh-staging-'), 'file')
  661. await writeFile(notADirectory, '')
  662. process.env.TMPDIR = notADirectory
  663. try {
  664. const { runtime } = await setup()
  665. const result = await runtime.run({ program: 'return 1', bindings: [] })
  666. expect(result.error?.kind).toBe('worker-exit')
  667. expect(result.error?.message).toContain('failed to stage the python bootstrap')
  668. expect(result.logs).toEqual([])
  669. } finally {
  670. if (previous === undefined) delete process.env.TMPDIR
  671. else process.env.TMPDIR = previous
  672. }
  673. })
  674. it('leaves no staging directory behind when a script copy fails', async () => {
  675. // `mkdtempSync` succeeding and a later `copyFileSync` failing is its own
  676. // case: the directory exists but is only partially populated. Recording it
  677. // before the copies would leak it, because `run` retries staging on the next
  678. // call and overwrites the single recorded path — teardown could then remove
  679. // only the newest attempt. Staging must clean up its own partial directory.
  680. //
  681. // Only `copyFileSync` is stubbed, and only for the second script, so
  682. // `mkdtempSync` really runs and the directory under assertion is real.
  683. // The assertion checks the exact paths this test staged (see the sibling
  684. // disposal-race test for why a global tmpdir diff races parallel workers).
  685. const stagedBefore = stagedDirs.length
  686. failNextCopyOf.value = 'protocol.py'
  687. try {
  688. const { runtime } = await setup()
  689. const result = await runtime.run({ program: 'return 1', bindings: [] })
  690. expect(result.error?.kind).toBe('worker-exit')
  691. expect(result.error?.message).toContain('failed to stage the python bootstrap')
  692. // The partial directory is gone, so nothing accumulates across retries.
  693. for (const dir of stagedDirs.slice(stagedBefore)) expect(existsSync(dir)).toBe(false)
  694. } finally {
  695. failNextCopyOf.value = undefined
  696. }
  697. }, 15_000)
  698. })
  699. describe('PythonCodeRuntime — process identity', () => {
  700. it('reads a live process start time and distinguishes it from an absent pid', () => {
  701. // The teardown guard signals `-child.pid` with a RAW `process.kill`, which
  702. // (unlike `child.kill()`) has no handle check, so it would reach a recycled
  703. // pgid during the window between the leader being reaped and `close` firing.
  704. // A pid alone cannot separate the original from its replacement -- both
  705. // answer `kill(pid, 0)` -- so the guard compares START TIME, and this pins
  706. // that the reading is stable for one process and absent for a pid that
  707. // cannot be read.
  708. const own = readProcessStart(process.pid)
  709. if (process.platform === 'linux') {
  710. // Same process, two reads: the identity must be stable, or the guard would
  711. // refuse to signal its own live group.
  712. expect(own).toBeDefined()
  713. expect(readProcessStart(process.pid)).toBe(own)
  714. // Pid 0 is never a readable /proc entry, so the guard degrades to
  715. // undefined rather than throwing on a teardown path. This is also the
  716. // reading a REAPED leader produces -- its /proc entry is gone while the
  717. // group it led can still hold survivors -- so `undefined` must NOT be
  718. // treated as an identity mismatch. Reading it as one refused the SIGKILL
  719. // that the same-group survivor tests depend on, which is why they went red
  720. // on Linux while passing on Darwin (where the reader always returns
  721. // undefined and the guard is inert).
  722. expect(readProcessStart(0)).toBeUndefined()
  723. } else {
  724. // Darwin has no /proc: the reader reports undefined, and `killGroup`
  725. // signals the pgid without the identity re-check instead of paying a `ps`
  726. // fork per signal.
  727. expect(own).toBeUndefined()
  728. }
  729. })
  730. })
  731. describe('PythonCodeRuntime — inherited resource limits', () => {
  732. // Darwin deliberately does not apply RLIMIT_AS, and its shell rejects `ulimit -v`.
  733. it.skipIf(process.platform === 'darwin')('runs under an inherited hard limit tighter than addressSpaceMb', async () => {
  734. // An unprivileged process may lower a hard rlimit but never raise it. Under
  735. // a harness started with `ulimit -v` below `addressSpaceBytes`, requesting
  736. // the configured cap made `setrlimit` raise `ValueError` and every run
  737. // returned a bootstrap exception — even though the inherited limit is
  738. // STRONGER than the one asked for. The bootstrap clamps to the inherited
  739. // hard limit instead, so the run proceeds under the stricter bound.
  740. //
  741. // `pythonBin` is the honest lever: a wrapper that lowers RLIMIT_AS and then
  742. // execs the real interpreter reproduces the inherited-limit condition
  743. // without touching this test process's own limits.
  744. const dir = await makeTempDir('dsh-rlimit-')
  745. const wrapper = join(dir, 'python3-capped')
  746. // 256 MiB, half the 512 MiB addressSpaceMb default, so the requested cap is
  747. // unambiguously above the inherited ceiling.
  748. await writeFile(wrapper, `#!/bin/sh\nulimit -v 262144\nexec "${PYABS}" "$@"\n`, { mode: 0o755 })
  749. const { runtime } = await setup({ pythonBin: wrapper })
  750. const result = await runtime.run({
  751. program: 'import resource\nreturn resource.getrlimit(resource.RLIMIT_AS)[1]',
  752. bindings: [],
  753. })
  754. expect(result.error).toBeUndefined()
  755. // The applied hard limit is the inherited one, not the configured 512 MiB.
  756. expect(result.value).toBe(256 * 1024 * 1024)
  757. }, 15_000)
  758. it('rejects at boot when an inherited RLIMIT_AS is too tight for the output budgets', async () => {
  759. // The host gate validates the output budgets against the CONFIGURED
  760. // addressSpaceMb, but a launch environment can inherit a STRICTER RLIMIT_AS
  761. // (a `ulimit -v` wrapper below addressSpaceMb), which the bootstrap clamps the
  762. // effective limit down to — leaving the budgets sized for a ceiling the child
  763. // never gets, so a near-budget output would OOM mid-run as an opaque
  764. // worker-exit. The bootstrap re-checks both budgets against the EFFECTIVE
  765. // clamped limit and fails loud at boot instead. A 128 MiB inherited limit
  766. // leaves 64 MiB budgetable (~5 MiB admissible under the 12x multiple), under
  767. // which a 32 MiB maxLogBytes — admitted by the 512 MiB configured default — is
  768. // rejected. The rejection surfaces as an 'exception' (bootstrap's
  769. // setrlimit-phase failure class), not a mid-run OOM. The repro is Linux-only
  770. // (macOS ignores `ulimit -v`); there the run proceeds.
  771. const dir = await makeTempDir('dsh-rlimit-')
  772. const wrapper = join(dir, 'python3-tight')
  773. await writeFile(wrapper, `#!/bin/sh\nulimit -v 131072\nexec "${PYABS}" "$@"\n`, { mode: 0o755 })
  774. const { runtime } = await setup({ pythonBin: wrapper, maxLogBytes: 32 * 1024 * 1024, addressSpaceMb: 512 })
  775. const result = await runtime.run({ program: 'return 1', bindings: [] })
  776. if (process.platform === 'darwin') {
  777. expect(result.error).toBeUndefined()
  778. } else {
  779. // The re-check raises inside bootstrap's resource-limit block, which
  780. // reports every setrlimit-phase failure as kind 'exception'; the message
  781. // discriminates this config rejection from a generic setrlimit error.
  782. expect(result.error?.kind).toBe('exception')
  783. expect(result.error?.message).toContain('too large for the inherited RLIMIT_AS')
  784. }
  785. }, 15_000)
  786. // The expected tuple includes RLIMIT_AS, which the backend deliberately skips on Darwin.
  787. it.skipIf(process.platform === 'darwin')('applies the configured limits when nothing tighter is inherited', async () => {
  788. // The clamp must not weaken the normal path: with an infinite inherited hard
  789. // limit there is nothing to clamp against, and RLIM_INFINITY compares as -1,
  790. // so treating it as a numeric bound would collapse every limit to -1.
  791. const { runtime } = await setup({ cpuSeconds: 42, addressSpaceMb: 400 })
  792. const result = await runtime.run({
  793. // `getrlimit` returns a tuple, which the lossless-JSON completion check
  794. // rejects; the pair is listed explicitly rather than converted.
  795. program: [
  796. 'import resource, sys',
  797. 'cpu = resource.getrlimit(resource.RLIMIT_CPU)',
  798. 'address_space = None if sys.platform == "darwin" else resource.getrlimit(resource.RLIMIT_AS)[1]',
  799. 'return {"cpu": [cpu[0], cpu[1]], "addressSpace": address_space}',
  800. ].join('\n'),
  801. bindings: [],
  802. })
  803. expect(result.error).toBeUndefined()
  804. // Darwin deliberately skips RLIMIT_AS; every other Unix host applies the
  805. // configured bytes alongside the CPU soft/hard pair.
  806. expect(result.value).toEqual({
  807. cpu: [42, 43],
  808. addressSpace: process.platform === 'darwin' ? null : 400 * 1024 * 1024,
  809. })
  810. }, 15_000)
  811. it('preserves an inherited soft limit stricter than the configured cap', async () => {
  812. // Clamping reads BOTH inherited bounds, not just the hard one. A deployment
  813. // that inherited a soft rlimit below the configured cap must keep that
  814. // stricter soft: returning the configured value would RAISE the effective
  815. // soft limit, loosening containment. The wrapper lowers only the SOFT CPU
  816. // limit (`ulimit -S -t`) and leaves the hard limit unlimited, so the
  817. // requested soft (`cpuSeconds`) sits above the inherited soft — the case that
  818. // exposed the bug. RLIMIT_CPU is used because macOS ignores `ulimit -v`
  819. // (RLIMIT_AS), which is exactly why the backend skips address space there.
  820. const dir = await makeTempDir('dsh-rlimit-soft-')
  821. const wrapper = join(dir, 'python3-soft-capped')
  822. // Soft CPU 5 s, well below the configured 30 s, hard left unlimited.
  823. await writeFile(wrapper, `#!/bin/sh\nulimit -S -t 5\nexec "${PYABS}" "$@"\n`, { mode: 0o755 })
  824. const { runtime } = await setup({ pythonBin: wrapper, cpuSeconds: 30 })
  825. const result = await runtime.run({
  826. program: 'import resource\nreturn resource.getrlimit(resource.RLIMIT_CPU)[0]',
  827. bindings: [],
  828. })
  829. expect(result.error).toBeUndefined()
  830. // The applied SOFT limit is the inherited 5 s, not the configured 30 s.
  831. expect(result.value).toBe(5)
  832. }, 15_000)
  833. it('reports a CPU overrun under a dual-limit ulimit as a timeout, not a worker-exit', async () => {
  834. // `ulimit -t N` sets BOTH the soft and hard CPU limit to N. The kernel
  835. // checks the hard limit and SIGKILLs a busy loop directly, so with
  836. // soft == hard the SIGXCPU signal is never delivered — and the host
  837. // classifies a CPU overrun ONLY on `signal === 'SIGXCPU'`, so the overrun
  838. // would be misreported as a `worker-exit` instead of a timeout. `_clamped`
  839. // now lowers a clamped soft==hard result by one unit (when hard >= 2), so
  840. // the SIGXCPU signal fires at the softer limit and the run reports a
  841. // timeout. This uses `ulimit -t 2` (hard == 2, so the soft is lowered to 1)
  842. // and leaves SIGXCPU unhandled, so the kernel terminates the busy loop at
  843. // 1 s with SIGXCPU and the host classifies it as a timeout.
  844. const dir = await makeTempDir('dsh-rlimit-dual-')
  845. const wrapper = join(dir, 'python3-dual-capped')
  846. // Both soft and hard CPU 2 s; configured cpuSeconds 30 s.
  847. await writeFile(wrapper, `#!/bin/sh\nulimit -t 2\nexec "${PYABS}" "$@"\n`, { mode: 0o755 })
  848. const { runtime } = await setup({ pythonBin: wrapper, cpuSeconds: 30, maxWallMs: 12_000 })
  849. const result = await runtime.run({
  850. program: [
  851. 'while True:',
  852. ' pass',
  853. 'return "unreachable"',
  854. ].join('\n'),
  855. bindings: [],
  856. })
  857. expect(result.error?.kind).toBe('timeout')
  858. expect(result.error?.message).toContain('CPU time exhausted')
  859. }, 15_000)
  860. it('reports a timeout when a program masks SIGXCPU and returns past the soft limit', async () => {
  861. // A program can mask SIGXCPU (pthread_sigmask SIG_BLOCK), burn past the
  862. // soft CPU limit, and return during the soft-to-hard gap. The settlement
  863. // recheck (`die_if_cpu_exhausted`) must UNBLOCK the signal before
  864. // re-delivering it, or the SIGXCPU stays pending and the child exits
  865. // normally with a success result. With the unblock, the re-delivered
  866. // SIGXCPU (default disposition) terminates the child and the host
  867. // classifies the run as a timeout. Fail-before: without the unblock the
  868. // run reports `value: "escaped"` and no error. The masking is guarded by
  869. // hasattr so the case is a no-op on platforms without pthread_sigmask.
  870. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 12_000 })
  871. const result = await runtime.run({
  872. program: [
  873. 'import signal, time',
  874. 'if hasattr(signal, "pthread_sigmask"):',
  875. ' signal.pthread_sigmask(signal.SIG_BLOCK, {signal.SIGXCPU})',
  876. 'end = time.process_time() + 1.05',
  877. 'while time.process_time() < end:',
  878. ' pass',
  879. 'return "escaped"',
  880. ].join('\n'),
  881. bindings: [],
  882. })
  883. expect(result.error?.kind).toBe('timeout')
  884. expect(result.value).toBeUndefined()
  885. }, 20_000)
  886. it('reports a timeout when the interpreter was started with SIGXCPU ignored (inherited state)', async () => {
  887. // The child inherits the host's SIGXCPU disposition: a wrapper that
  888. // ignores SIGXCPU before exec'ing python3 hands the child a soft
  889. // RLIMIT_CPU that cannot stop it. The bootstrap resets SIGXCPU to SIG_DFL
  890. // before model code runs, so a busy loop still ends as a timeout rather
  891. // than running to the hard limit and being misclassified as worker-exit.
  892. const wrapper = join(tmpdir(), `dsh-xcpu-ignore-${process.pid}.sh`)
  893. tempFiles.push(wrapper)
  894. writeFileSync(wrapper, `#!/bin/sh\ntrap "" XCPU\nexec "${PYABS}" "$@"\n`, { mode: 0o755 })
  895. try {
  896. const { runtime } = await setup({ maxWallMs: 30_000, cpuSeconds: 1, pythonBin: wrapper })
  897. const result = await runtime.run({
  898. program: ['while True: pass'].join('\n'),
  899. bindings: [],
  900. })
  901. expect(result.error?.kind).toBe('timeout')
  902. } finally {
  903. rmSync(wrapper, { force: true })
  904. }
  905. }, 20_000)
  906. it('reports a timeout when a program traps AND masks SIGXCPU and returns past the soft limit', async () => {
  907. // The mask-only case exercises the unblock; the trap+mask combination is
  908. // the harder one: a program that installed a custom handler AND masked the
  909. // signal has that PENDING handler run the moment the signal is unblocked
  910. // (CPython delivers it at the next eval-breaker checkpoint in model code),
  911. // and the handler re-masks — so the settlement recheck must restore the
  912. // default disposition BEFORE unblocking. With SIG_DFL restored first, the
  913. // pending signal kills the process inside the kernel with no bytecode
  914. // window; without it, the handler re-blocks and the child exits normally
  915. // with a success value. Fail-before: the run reports `value: "escaped"`.
  916. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 12_000 })
  917. const result = await runtime.run({
  918. program: [
  919. 'import signal, time',
  920. 'if hasattr(signal, "pthread_sigmask"):',
  921. ' def h(signum, frame):',
  922. ' signal.pthread_sigmask(signal.SIG_BLOCK, {signal.SIGXCPU})',
  923. ' signal.signal(signal.SIGXCPU, h)',
  924. ' signal.pthread_sigmask(signal.SIG_BLOCK, {signal.SIGXCPU})',
  925. ' end = time.process_time() + 1.05',
  926. ' while time.process_time() < end:',
  927. ' pass',
  928. 'return "escaped"',
  929. ].join('\n'),
  930. bindings: [],
  931. })
  932. expect(result.error?.kind).toBe('timeout')
  933. expect(result.value).toBeUndefined()
  934. }, 20_000)
  935. it('rechecks CPU at settlement against the effective inherited soft limit', async () => {
  936. // The settlement-time CPU recheck must compare against the EFFECTIVE soft
  937. // limit (`_clamped` may have lowered it to a stricter inherited value), not
  938. // the configured `cpuSeconds`. A program that traps SIGXCPU, burns past the
  939. // inherited soft, and returns inside the soft-to-hard gap would otherwise be
  940. // compared to the configured value and falsely reported successful, bypassing
  941. // the inherited limit. The wrapper sets a 1 s soft CPU limit; the program
  942. // traps SIGXCPU and busy-loops past it, then returns — the recheck must
  943. // re-deliver SIGXCPU so the host classifies the run as a timeout.
  944. const dir = await makeTempDir('dsh-cpu-recheck-')
  945. const wrapper = join(dir, 'python3-cpu-capped')
  946. await writeFile(wrapper, `#!/bin/sh\nulimit -S -t 1\nexec "${PYABS}" "$@"\n`, { mode: 0o755 })
  947. const { runtime } = await setup({ pythonBin: wrapper, cpuSeconds: 30, maxWallMs: 12_000 })
  948. const result = await runtime.run({
  949. program: [
  950. 'import signal, time',
  951. // Trap SIGXCPU so the soft limit does not terminate the program; burn
  952. // CPU well past the inherited 1 s soft, then return normally.
  953. 'signal.signal(signal.SIGXCPU, lambda *a: None)',
  954. 'end = time.process_time() + 2.5',
  955. 'while time.process_time() < end:',
  956. ' pass',
  957. 'return "returned"',
  958. ].join('\n'),
  959. bindings: [],
  960. })
  961. // The recheck compares spent CPU against the effective 1 s soft, not 30 s, so
  962. // the run is a timeout rather than a false success.
  963. expect(result.error?.kind).toBe('timeout')
  964. }, 20_000)
  965. })
  966. describe('PythonCodeRuntime — programs and bindings', () => {
  967. it('runs a top-level script, captures print output, and returns `result`', async () => {
  968. const { runtime } = await setup()
  969. const result = await runtime.run({
  970. program: [
  971. 'x = 40',
  972. 'y = 2',
  973. 'print("hello", x + y)',
  974. 'return {"answer": x + y}',
  975. ].join('\n'),
  976. bindings: [],
  977. })
  978. expect(result.error).toBeUndefined()
  979. expect(result.value).toEqual({ answer: 42 })
  980. // `print` in Python emits: text, ' ', text, '\n'. Concat the captured
  981. // fragments and assert the model-visible message survives.
  982. expect(result.logs.join('')).toContain('hello 42')
  983. // 15s: this is usually the suite's first real subprocess — a cold python3
  984. // start (interpreter + asyncio import) on a loaded CI runner can exceed
  985. // the 5s default alone; later tests reuse the warm page cache.
  986. }, 15_000)
  987. it('exposes only the platform temp directory from the host environment', async () => {
  988. const { runtime } = await setup()
  989. const result = await runtime.run({
  990. program: [
  991. 'import os',
  992. 'return {',
  993. ' "tmpdir": os.environ.get("TMPDIR"),',
  994. ' "path": os.environ.get("PATH"),',
  995. ' "home": os.environ.get("HOME"),',
  996. ' "token": os.environ.get("DEEPSEEK_API_KEY"),',
  997. '}',
  998. ].join('\n'),
  999. bindings: [],
  1000. })
  1001. expect(result.error).toBeUndefined()
  1002. expect(result.value).toEqual({ tmpdir: tmpdir(), path: null, home: null, token: null })
  1003. expect(result.logs).toEqual([])
  1004. })
  1005. it('bridges binding calls both ways and rejects the program-side call on a host rejection', async () => {
  1006. const { runtime } = await setup()
  1007. const calls: unknown[] = []
  1008. const result = await runtime.run({
  1009. program: [
  1010. 'first = await tools.echo({"n": 1})',
  1011. 'caught = ""',
  1012. 'try:',
  1013. ' await tools.fail({})',
  1014. 'except RuntimeError as e:',
  1015. ' caught = str(e)',
  1016. 'return {"first": first, "caught": caught}',
  1017. ].join('\n'),
  1018. bindings: tools({
  1019. echo: async (args) => { calls.push(args); return { echoed: args as CodeJsonValue } },
  1020. fail: async () => { throw new Error('nope') },
  1021. }),
  1022. })
  1023. expect(result.error).toBeUndefined()
  1024. expect(result.value).toEqual({ first: { echoed: { n: 1 } }, caught: 'nope' })
  1025. expect(calls).toEqual([{ n: 1 }])
  1026. })
  1027. it('keeps decoding binding replies when _decode_json_plain is rebound', async () => {
  1028. // read_frame_async resolves _decode_json_plain at call time; a program that
  1029. // rebinds __main__._decode_json_plain would otherwise kill the reply pump
  1030. // (a broken decode strands every pending Future to the wall clock). The
  1031. // decode primitives are def-time captures on the channel methods, so a
  1032. // rebind cannot break reply delivery.
  1033. const { runtime } = await setup()
  1034. const result = await runtime.run({
  1035. program: [
  1036. 'import __main__',
  1037. '__main__._decode_json_plain = None',
  1038. 'first = await tools.echo({"n": 1})',
  1039. 'return first',
  1040. ].join('\n'),
  1041. bindings: tools({
  1042. echo: async args => ({ echoed: args as CodeJsonValue }),
  1043. }),
  1044. })
  1045. expect(result.error).toBeUndefined()
  1046. expect(result.value).toEqual({ echoed: { n: 1 } })
  1047. }, 15_000)
  1048. it('keeps dispatch working when _lossless_json_violation, asyncio, and send_sync are rebound', async () => {
  1049. // dispatch binds _lossless_json_violation, asyncio.get_event_loop, and the
  1050. // channel's send method into _run locals before the program runs, so a
  1051. // rebind of __main__._lossless_json_violation/__main__.asyncio/
  1052. // __main__.ProtocolChannel.send_sync cannot turn a legitimate binding call
  1053. // into an exception or a wall-clock timeout.
  1054. const { runtime } = await setup()
  1055. const result = await runtime.run({
  1056. program: [
  1057. 'import __main__',
  1058. 'def boom(*a, **k):',
  1059. ' raise RuntimeError("hijacked")',
  1060. '__main__._lossless_json_violation = boom',
  1061. '__main__.asyncio = boom',
  1062. '__main__.ProtocolChannel.send_sync = boom',
  1063. '__main__._encode_json_plain = boom',
  1064. '__main__.ProtocolChannel.write_encoded = boom',
  1065. 'first = await tools.echo({"n": 1})',
  1066. 'return first',
  1067. ].join('\n'),
  1068. bindings: tools({
  1069. echo: async args => ({ echoed: args as CodeJsonValue }),
  1070. }),
  1071. })
  1072. expect(result.error).toBeUndefined()
  1073. expect(result.value).toEqual({ echoed: { n: 1 } })
  1074. }, 15_000)
  1075. it('keeps the reply pump reading when the read_frame_async class attribute is rebound', async () => {
  1076. // _pump_replies' frame reader is a bound method captured by _run before the
  1077. // program runs and passed in as an explicit argument, so a program rebinding
  1078. // `__main__.ProtocolChannel.read_frame_async` cannot redirect the pump (a
  1079. // body-local `channel.read_frame_async` lookup would resolve the rebound
  1080. // class attribute, since the pump starts after the program's top-level
  1081. // statements).
  1082. const { runtime } = await setup()
  1083. const result = await runtime.run({
  1084. program: [
  1085. 'import __main__',
  1086. 'async def boom(*a, **k):',
  1087. ' raise RuntimeError("hijacked reader")',
  1088. '__main__.ProtocolChannel.read_frame_async = boom',
  1089. 'first = await tools.echo({"n": 1})',
  1090. 'return first',
  1091. ].join('\n'),
  1092. bindings: tools({
  1093. echo: async args => ({ echoed: args as CodeJsonValue }),
  1094. }),
  1095. })
  1096. expect(result.error).toBeUndefined()
  1097. expect(result.value).toEqual({ echoed: { n: 1 } })
  1098. }, 15_000)
  1099. it('keeps the rejection contract when _BindingRejection is rebound', async () => {
  1100. // `dispatch`'s except clause resolves `_BindingRejection` at call time; a
  1101. // program that rebinds `__main__._BindingRejection = ValueError` would
  1102. // otherwise let the internal marker type leak into model code (the program
  1103. // would catch a `ValueError` for a host rejection). The class is now bound
  1104. // into `_run` locals before the program runs, so a host rejection still
  1105. // surfaces as the declared `RuntimeError`.
  1106. const { runtime } = await setup()
  1107. const result = await runtime.run({
  1108. program: [
  1109. 'import __main__',
  1110. '__main__._BindingRejection = ValueError',
  1111. 'caught = ""',
  1112. 'try:',
  1113. ' await tools.fail({})',
  1114. 'except RuntimeError as e:',
  1115. ' caught = e.args[0] if e.args else ""',
  1116. 'except Exception as e:',
  1117. ' caught = "WRONG TYPE: " + type(e).__name__',
  1118. 'return caught',
  1119. ].join('\n'),
  1120. bindings: tools({
  1121. fail: async () => { throw new Error('nope') },
  1122. }),
  1123. })
  1124. expect(result.error).toBeUndefined()
  1125. expect(result.value).toBe('nope')
  1126. }, 15_000)
  1127. it('still answers the call when the rejection value cannot be converted to a string', async () => {
  1128. // `messageOf` calls `String(error)`, which runs the value's own conversion,
  1129. // and this call site is a DETACHED async reply callback. A rejection whose
  1130. // `Symbol.toPrimitive` throws therefore escaped as an unhandled rejection:
  1131. // the reply frame was never written, the program stayed blocked on `await`,
  1132. // and the run degraded to a `maxWallMs` timeout (observed) — a host with no
  1133. // `unhandledRejection` listener would exit instead. The rejection must reach
  1134. // the program as an ordinary error carrying a fixed placeholder.
  1135. const { runtime } = await setup({ maxWallMs: 8_000 })
  1136. const result = await runtime.run({
  1137. program: [
  1138. 'try:',
  1139. ' await tools.hostile({})',
  1140. 'except RuntimeError as e:',
  1141. ' return "rejected: " + str(e)',
  1142. 'return "no rejection"',
  1143. ].join('\n'),
  1144. bindings: tools({
  1145. hostile: async () => {
  1146. throw { [Symbol.toPrimitive]() { throw new Error('toPrimitive blew up') } }
  1147. },
  1148. }),
  1149. })
  1150. expect(result.error).toBeUndefined()
  1151. expect(result.value).toBe('rejected: <unrenderable rejection value>')
  1152. }, 15_000)
  1153. it('still answers the call when an Error carries a cyclic value in place of its message', async () => {
  1154. // `Error.message` is typed `string` but is a plain writable property, so a
  1155. // rejection can carry any value there. Returning it verbatim handed a
  1156. // non-string to `sendReply`, breaching `encodeJsonPlain`'s JSON-plain
  1157. // precondition: a cyclic object grew the encoder stack until the host threw
  1158. // RangeError from the detached reply callback, so no reply frame was written
  1159. // and the run degraded to a `maxWallMs` timeout (observed). The conversion
  1160. // must contain it — `String()` on a cycle throws inside the guard and lands
  1161. // on the placeholder, so the program sees an ordinary error.
  1162. const { runtime } = await setup({ maxWallMs: 8_000 })
  1163. const result = await runtime.run({
  1164. program: [
  1165. 'try:',
  1166. ' await tools.hostile({})',
  1167. 'except RuntimeError as e:',
  1168. ' return "rejected: " + str(e)',
  1169. 'return "no rejection"',
  1170. ].join('\n'),
  1171. bindings: tools({
  1172. hostile: async () => {
  1173. const cyclic: { self?: unknown; [Symbol.toPrimitive]: () => string } = {
  1174. // A cycle alone is inert for `String()`; the throwing conversion is
  1175. // what proves the guard runs rather than the encoder.
  1176. [Symbol.toPrimitive]: () => { throw new Error('cyclic message') },
  1177. }
  1178. cyclic.self = cyclic
  1179. const error = new Error('placeholder')
  1180. // Writable per spec, so no cast is needed to install a non-string.
  1181. ;(error as unknown as { message: unknown }).message = cyclic
  1182. throw error
  1183. },
  1184. }),
  1185. })
  1186. expect(result.error).toBeUndefined()
  1187. expect(result.value).toBe('rejected: <unrenderable rejection value>')
  1188. }, 15_000)
  1189. it('renders an Error whose message is a value with no JSON form', async () => {
  1190. // The non-cyclic arm. A number would not discriminate: `scalarJson` renders
  1191. // it as digits and the child `str()`s the field back, so it survives the
  1192. // wire either way. `undefined` is the value that separates the two orders —
  1193. // `scalarJson` emits a bare `undefined` token, so the reply line is not JSON
  1194. // at all, the child's parse drops the frame, and the program stays blocked
  1195. // on `await` until the wall ceiling (observed). Converting first sends the
  1196. // string "undefined", which the program receives as an ordinary rejection.
  1197. const { runtime } = await setup({ maxWallMs: 8_000 })
  1198. const result = await runtime.run({
  1199. program: [
  1200. 'try:',
  1201. ' await tools.absent({})',
  1202. 'except RuntimeError as e:',
  1203. ' return "rejected: " + str(e)',
  1204. 'return "no rejection"',
  1205. ].join('\n'),
  1206. bindings: tools({
  1207. absent: async () => {
  1208. const error = new Error('placeholder')
  1209. ;(error as unknown as { message: unknown }).message = undefined
  1210. throw error
  1211. },
  1212. }),
  1213. })
  1214. expect(result.error).toBeUndefined()
  1215. expect(result.value).toBe('rejected: undefined')
  1216. }, 15_000)
  1217. it('runs a program with no await', async () => {
  1218. const { runtime } = await setup()
  1219. const result = await runtime.run({
  1220. program: 'return 2 + 2',
  1221. bindings: [],
  1222. })
  1223. expect(result.error).toBeUndefined()
  1224. expect(result.value).toBe(4)
  1225. })
  1226. it('returns JSON null whether the program returns None or falls off the end', async () => {
  1227. // Python has no `undefined`: an async body that returns None and one that
  1228. // never returns both yield None, so both complete as an exact JSON null.
  1229. // (The worker/TS backend can tell `return undefined` from `return null`;
  1230. // Python cannot, and reporting null for both is the honest rendering.)
  1231. const { runtime } = await setup()
  1232. const explicit = await runtime.run({ program: 'return None', bindings: [] })
  1233. expect(explicit.error).toBeUndefined()
  1234. expect(explicit.value).toBeNull()
  1235. const noReturn = await runtime.run({ program: 'x = 1', bindings: [] })
  1236. expect(noReturn.error).toBeUndefined()
  1237. expect(noReturn.value).toBeNull()
  1238. })
  1239. it('settles with no value on a forged valueless done frame', async () => {
  1240. // The child always sends a value now (return None → JSON null), so a done
  1241. // frame with no value key can only be forged; the host settles it as a
  1242. // value-less completion rather than crashing on the absent field.
  1243. const { runtime } = await setup()
  1244. const result = await runtime.run({
  1245. program: [
  1246. 'import os',
  1247. 'os.write(3, b\'{"type":"done"}\\n\')',
  1248. 'import time',
  1249. 'time.sleep(5)',
  1250. ].join('\n'),
  1251. bindings: [],
  1252. })
  1253. expect(result.error).toBeUndefined()
  1254. expect(result.value).toBeUndefined()
  1255. })
  1256. it('coalesces print arguments into one log line, not per-write fragments', async () => {
  1257. // print("a","b") calls write() per arg/sep/newline; the stream must emit
  1258. // one logical line "a b" so PTC mode's join(newline) does not insert
  1259. // spurious blank lines. Two prints → exactly two entries, no empties.
  1260. const { runtime } = await setup()
  1261. const result = await runtime.run({
  1262. program: ['print("a", "b")', 'print("c")', 'return None'].join('\n'),
  1263. bindings: [],
  1264. })
  1265. expect(result.error).toBeUndefined()
  1266. expect(result.logs).toEqual(['a b', 'c'])
  1267. })
  1268. it('flushes a print with no trailing newline', async () => {
  1269. const { runtime } = await setup()
  1270. const result = await runtime.run({
  1271. program: ['print("partial", end="")', 'return None'].join('\n'),
  1272. bindings: [],
  1273. })
  1274. expect(result.error).toBeUndefined()
  1275. expect(result.logs).toEqual(['partial'])
  1276. })
  1277. it('aggregates a large newline-free native write into one log entry, not one per pipe chunk', async () => {
  1278. // A single `os.write` larger than one pipe read arrives as several Node
  1279. // `data` chunks. `logs` entries are joined with `\n` downstream, so pushing
  1280. // one entry per transport chunk would insert model-visible newlines at
  1281. // arbitrary pipe boundaries inside one native write. Stray capture holds a
  1282. // per-stream residual and admits only on a real `\n`, so a 200 KiB blast
  1283. // with no newline reads back as exactly one entry with no interior breaks.
  1284. const { runtime } = await setup({ maxLogBytes: 300_000 })
  1285. const size = 200_000
  1286. const result = await runtime.run({
  1287. program: ['import os', `os.write(1, b"A" * ${size})`, 'return None'].join('\n'),
  1288. bindings: [],
  1289. })
  1290. expect(result.error).toBeUndefined()
  1291. expect(result.logs).toEqual(['A'.repeat(size)])
  1292. })
  1293. it('splits native output on its own newlines, one entry per line', async () => {
  1294. // The complement of the aggregation case: real newlines in a native write
  1295. // still delimit entries, matching the child's line-granular `log` frames.
  1296. const { runtime } = await setup()
  1297. const result = await runtime.run({
  1298. program: ['import os', 'os.write(1, b"one\\ntwo\\nthree")', 'return None'].join('\n'),
  1299. bindings: [],
  1300. })
  1301. expect(result.error).toBeUndefined()
  1302. expect(result.logs).toEqual(['one', 'two', 'three'])
  1303. })
  1304. it('preserves each native stream order while allowing backend-dependent interleaving', async () => {
  1305. const { runtime } = await setup()
  1306. const result = await runtime.run({
  1307. program: [
  1308. 'import os',
  1309. 'os.write(1, b"stdout-one\\n")',
  1310. 'os.write(2, b"stderr-one\\n")',
  1311. 'os.write(1, b"stdout-two\\n")',
  1312. 'os.write(2, b"stderr-two\\n")',
  1313. 'return None',
  1314. ].join('\n'),
  1315. bindings: [],
  1316. })
  1317. expect(result.error).toBeUndefined()
  1318. expect(result.logs.indexOf('stdout-one')).toBeLessThan(result.logs.indexOf('stdout-two'))
  1319. expect(result.logs.indexOf('stderr-one')).toBeLessThan(result.logs.indexOf('stderr-two'))
  1320. })
  1321. it('bounds a newline-free native flood by the ledger instead of buffering it whole', async () => {
  1322. // A newline-free write far larger than maxLogBytes must not accumulate in
  1323. // the host-side residual: when the pending residual would cross the budget
  1324. // it is admitted (and truncated) immediately, and once the ledger has
  1325. // truncated, later chunks stop buffering entirely. The run still completes
  1326. // and the captured output ends at the truncation marker rather than
  1327. // retaining the whole flood.
  1328. const { runtime } = await setup({ maxLogBytes: 4096 })
  1329. const result = await runtime.run({
  1330. program: ['import os', 'os.write(1, b"A" * 2_000_000)', 'return None'].join('\n'),
  1331. bindings: [],
  1332. })
  1333. expect(result.error).toBeUndefined()
  1334. expect(result.logs.at(-1)).toBe(logTruncationMarker(4096))
  1335. // The retained output is bounded by the budget, not the 2 MB flood.
  1336. expect(result.logs.join('').length).toBeLessThan(4096)
  1337. })
  1338. it('bounds a newline-free single-character Python write drip by the fragment cap, not OOM', async () => {
  1339. // The child-side `_LogStream` buffers one fragment per `write` (so
  1340. // `print("x", end="")` does not concatenate quadratically). A newline-free
  1341. // drip of one character per call past a large `maxLogBytes` would otherwise
  1342. // accumulate one list slot (and one str object) per call — 25 M calls =
  1343. // ~25 M slots, which OOMs the host on its own accounting before the byte
  1344. // budget is reached. The stream seals the fragment list past
  1345. // `_PENDING_MAX_CHUNKS` into one joined block (character count unchanged),
  1346. // bounding the live fragment count exactly as the host-side `captureStray`
  1347. // seal does. This drives well past the cap and asserts the run still
  1348. // completes with a truncation marker rather than a MemoryError.
  1349. const { runtime } = await setup({ maxLogBytes: 4096 })
  1350. const result = await runtime.run({
  1351. program: [
  1352. 'import sys',
  1353. 'for _ in range(200_000):',
  1354. ' sys.stdout.write("x")',
  1355. 'return None',
  1356. ].join('\n'),
  1357. bindings: [],
  1358. })
  1359. expect(result.error).toBeUndefined()
  1360. expect(result.logs.at(-1)).toBe(logTruncationMarker(4096))
  1361. })
  1362. it('bounds a control-char-dense native residual by serialized cost, not raw length', async () => {
  1363. // A newline-free NUL flood passes the cheap `length + 3` lower bound at a
  1364. // raw length well under the budget, but each NUL serializes to `\u0000` (6
  1365. // bytes), so the true JSON cost is ~6x. The ledger must charge that
  1366. // serialized cost — and `jsonStringCostUpTo` must measure it WITHOUT
  1367. // allocating the escaped copy, so a near-budget line under a large
  1368. // maxLogBytes cannot momentarily allocate a multi-gigabyte `JSON.stringify`
  1369. // result. Under a small budget the residual is truncated once the serialized
  1370. // cost crosses it.
  1371. const { runtime } = await setup({ maxLogBytes: 4096 })
  1372. const result = await runtime.run({
  1373. program: ['import os', 'os.write(1, b"\\x00" * 4000)', 'return None'].join('\n'),
  1374. bindings: [],
  1375. })
  1376. expect(result.error).toBeUndefined()
  1377. expect(result.logs.at(-1)).toBe(logTruncationMarker(4096))
  1378. })
  1379. it('rejects an output budget that could breach addressSpaceMb during encode at load', async () => {
  1380. // The child builds, charges, and encodes a `maxLogBytes` log entry or a
  1381. // `maxValueBytes` completion value under RLIMIT_AS, and both trigger on
  1382. // character count against a serialized-byte budget — an astral character is
  1383. // one character but ~4 bytes stored and ~4 encoded, and THREE such copies are
  1384. // live at the peak (the caller's write argument, the slice/join handed to
  1385. // push, and the encode copy), so a budget approaching the address space lets a
  1386. // legitimate near-budget output breach it and die as worker-exit. The
  1387. // incompatible pair is rejected at load: each budget times the worst-case
  1388. // multiple (12) must fit the address space LEFT after the fixed interpreter
  1389. // baseline. Against a 256 MiB address space that leaves 192 MiB budgetable
  1390. // (~16 MiB admissible), so a 50 MB cap is far over; the default caps against
  1391. // 512 MiB are not. Both budgets are gated symmetrically — the value case sets
  1392. // a default-fitting maxLogBytes so the maxValueBytes check is what fires.
  1393. const ctxLog = new Context()
  1394. await expect(ctxLog.plugin(PythonCodeRuntime, { maxLogBytes: 50_000_000, addressSpaceMb: 256 }))
  1395. .rejects.toThrow(/maxLogBytes times the 12x worst-case Unicode expansion must fit/)
  1396. const ctxValue = new Context()
  1397. await expect(ctxValue.plugin(PythonCodeRuntime, { maxValueBytes: 50_000_000, addressSpaceMb: 256 }))
  1398. .rejects.toThrow(/maxValueBytes times the 12x worst-case Unicode expansion must fit/)
  1399. // Discriminates 12 from 8: a 48 MiB maxLogBytes against a 512 MiB address
  1400. // space leaves 448 MiB budgetable. 48*8 = 384 MiB fits (the old 8x multiple
  1401. // wrongly ADMITTED this), but 48*12 = 576 MiB does not. The ~12x peak this
  1402. // guards is the NEWLINE path's single near-budget write — the caller's own
  1403. // string, the line slice, and the encode copy live at once. The settlement
  1404. // flush is no longer the binding case: `flush_line` drops the pending chunks
  1405. // before its push, so it holds two copies, not three.
  1406. const ctxTwelve = new Context()
  1407. await expect(ctxTwelve.plugin(PythonCodeRuntime, { maxLogBytes: 48 * 1024 * 1024, addressSpaceMb: 512 }))
  1408. .rejects.toThrow(/maxLogBytes times the 12x worst-case Unicode expansion must fit/)
  1409. // An addressSpaceMb at or below the interpreter baseline leaves nothing
  1410. // budgetable, so no budget value can pass. It is rejected on its own terms:
  1411. // the budget loop would otherwise report "a limit of -1" (or -2796203 at
  1412. // 32 MiB) while naming maxLogBytes, sending the operator to the wrong knob.
  1413. const ctxBaseline = new Context()
  1414. await expect(ctxBaseline.plugin(PythonCodeRuntime, { addressSpaceMb: 64 }))
  1415. .rejects.toThrow(/addressSpaceMb must exceed the 67108864-byte interpreter baseline/)
  1416. const ctxBelow = new Context()
  1417. await expect(ctxBelow.plugin(PythonCodeRuntime, { addressSpaceMb: 32 }))
  1418. .rejects.toThrow(/addressSpaceMb must exceed the 67108864-byte interpreter baseline/)
  1419. // The default caps against the default 512 MiB address space load.
  1420. const ok = new Context()
  1421. const fiber = await ok.plugin(PythonCodeRuntime, { maxLogBytes: 65536, maxValueBytes: 32768, addressSpaceMb: 512 })
  1422. await fiber.dispose()
  1423. })
  1424. it('bounds an illegal-UTF-8 native residual by its U+FFFD-decoded cost', async () => {
  1425. // Every 0xFF byte is illegal in any UTF-8 sequence, so `toString('utf8')`
  1426. // renders each as U+FFFD (3 serialized bytes). `accrueStrayCost` must charge
  1427. // that 3, not the raw 1: otherwise the newline-free residual grows to a full
  1428. // budget's worth of RAW bytes before flushing — a ~3x undercount that near a
  1429. // large maxLogBytes retains hundreds of MiB then expands toward a ~1 GiB peak
  1430. // in flushStray's concat + toString. Paced single-byte writes (each its own
  1431. // `data` chunk, like the sealing case) expose the sub-chunk accrual: charged
  1432. // at 3 the residual crosses a 3072-byte budget after ~1024 bytes and flushes;
  1433. // charged at 1 it would need ~3072 bytes, so the peak residual triples. The
  1434. // largest merged buffer is the discriminator.
  1435. const realConcat = Buffer.concat.bind(Buffer)
  1436. let maxConcat = 0
  1437. Buffer.concat = (list: readonly Uint8Array[], total?: number): Buffer<ArrayBuffer> => {
  1438. const merged = realConcat(list, total)
  1439. if (merged.length > maxConcat) maxConcat = merged.length
  1440. return merged
  1441. }
  1442. let result: CodeRunResult
  1443. try {
  1444. const { runtime } = await setup({ maxLogBytes: 3072, maxWallMs: 30_000 })
  1445. result = await runtime.run({
  1446. program: [
  1447. 'import os',
  1448. 'for _ in range(6000):',
  1449. ' os.write(1, b"\\xff")',
  1450. ' os.sched_yield()',
  1451. 'return None',
  1452. ].join('\n'),
  1453. bindings: [],
  1454. })
  1455. } finally {
  1456. Buffer.concat = realConcat
  1457. }
  1458. expect(result.error).toBeUndefined()
  1459. expect(result.logs.at(-1)).toBe(logTruncationMarker(3072))
  1460. // Charged at 3, the residual flushes around 1024 raw bytes; the largest
  1461. // merged buffer stays well under 2048. A raw-byte undercount would let it
  1462. // reach ~3072 before flushing, so 2048 discriminates.
  1463. expect(maxConcat).toBeLessThan(2048)
  1464. })
  1465. it('charges a structurally-valid but illegal UTF-8 sequence its U+FFFD-decoded cost', async () => {
  1466. // A CESU-8 lone surrogate `ED A0 80` is structurally well-formed (a 3-byte
  1467. // lead plus two 0x80–0xBF continuations) but ILLEGAL: `toString('utf8')`
  1468. // renders each of the three bytes as its own U+FFFD (serialized cost 9), not
  1469. // one width-3 character. The newline-free flush trigger weighs the residual
  1470. // through `accrueStrayCost`, which must validate each lead's
  1471. // first-continuation range (ED excludes A0–BF) and charge the true 9 — else a
  1472. // CESU flood undercounts 3x and the residual grows toward a full budget's raw
  1473. // bytes before flushing, the same peak-memory vector as the 0xFF case. The
  1474. // bytes are written one at a time (each its own `data` chunk, no pipe
  1475. // coalescing) and `Buffer.concat` is wrapped to measure the peak residual.
  1476. const realConcat = Buffer.concat.bind(Buffer)
  1477. let maxConcat = 0
  1478. Buffer.concat = (list: readonly Uint8Array[], total?: number): Buffer<ArrayBuffer> => {
  1479. const merged = realConcat(list, total)
  1480. if (merged.length > maxConcat) maxConcat = merged.length
  1481. return merged
  1482. }
  1483. let result: CodeRunResult
  1484. try {
  1485. const { runtime } = await setup({ maxLogBytes: 3072, maxWallMs: 30_000 })
  1486. result = await runtime.run({
  1487. program: [
  1488. 'import os',
  1489. 'seq = (0xed, 0xa0, 0x80)',
  1490. 'for _ in range(2000):',
  1491. ' for b in seq:',
  1492. ' os.write(1, bytes((b,)))',
  1493. ' os.sched_yield()',
  1494. 'return None',
  1495. ].join('\n'),
  1496. bindings: [],
  1497. })
  1498. } finally {
  1499. Buffer.concat = realConcat
  1500. }
  1501. expect(result.error).toBeUndefined()
  1502. expect(result.logs.at(-1)).toBe(logTruncationMarker(3072))
  1503. // Each 3-byte sequence costs 9 (three U+FFFD), so single-byte-paced the
  1504. // residual crosses the 3072 budget after ~342 raw bytes and flushes; the
  1505. // largest merged buffer stays well under 2048. Charging the structural width
  1506. // 3 would need ~1024 raw bytes, tripling the peak past 2048.
  1507. expect(maxConcat).toBeLessThan(2048)
  1508. })
  1509. it('charges a lone surrogate its full six escaped bytes, not three', async () => {
  1510. // A forged `log` frame carrying `\ud800` escapes materializes lone
  1511. // surrogates after JSON.parse. `Buffer.byteLength` of U+FFFD is 3, but
  1512. // ES2019 well-formed `JSON.stringify` emits `\ud800` at 6 bytes, so charging
  1513. // the raw width would admit ~2x the configured budget of serialized bytes
  1514. // (the same family as the NUL-flood undercount, at 2x rather than 6x). The
  1515. // cost walker charges surrogates the full 6, so a flood truncates at budget.
  1516. // Forged on fd 3 because Python stdout will not emit lone surrogates.
  1517. const { runtime } = await setup({ maxLogBytes: 4096 })
  1518. const result = await runtime.run({
  1519. program: [
  1520. 'import os',
  1521. // 1000 \ud800 escapes: charged at the buggy raw width 1000 * 3 = 3000
  1522. // bytes fits under 4096 (wrongly admitted), but the correct serialized
  1523. // width 1000 * 6 = 6000 bytes is over budget — so the ledger must
  1524. // truncate. The count sits in the 683..1365 window where the two
  1525. // chargings disagree, making the test discriminate.
  1526. String.raw`frame = b'{"type":"log","text":"' + b'\\ud800' * 1000 + b'"}\n'`,
  1527. 'os.write(3, frame)',
  1528. 'return None',
  1529. ].join('\n'),
  1530. bindings: [],
  1531. })
  1532. expect(result.error).toBeUndefined()
  1533. expect(result.logs.at(-1)).toBe(logTruncationMarker(4096))
  1534. })
  1535. it('drops a second stray line in the same chunk once the first truncated the ledger', async () => {
  1536. // One `os.write` carrying two newline-terminated lines where the first
  1537. // exhausts maxLogBytes: the first line's admit truncates and marks the
  1538. // ledger, and the second line's admit — reached in the same `data` callback
  1539. // — must be the post-truncation no-op. Proves that branch is exercised, so
  1540. // it carries no v8-ignore. Kept to 108 bytes (< the smallest PIPE_BUF, 512 on
  1541. // macOS) so the whole payload lands in ONE atomic write and one `data`
  1542. // callback — the two newlines cannot split across callbacks and leave the
  1543. // branch un-exercised, which would be a hard-to-attribute per-file coverage
  1544. // flake. The first line's 100 bytes already exceed the 64-byte budget, so it truncates.
  1545. const { runtime } = await setup({ maxLogBytes: 64 })
  1546. const result = await runtime.run({
  1547. program: ['import os', 'os.write(1, b"A" * 100 + b"\\nSECOND\\n")', 'return None'].join('\n'),
  1548. bindings: [],
  1549. })
  1550. expect(result.error).toBeUndefined()
  1551. expect(result.logs.at(-1)).toBe(logTruncationMarker(64))
  1552. expect(result.logs.join('\n')).not.toContain('SECOND')
  1553. })
  1554. it('charges a broken multibyte sequence its U+FFFD bytes, split across pipe chunks', async () => {
  1555. // A 3-byte lead (0xE4) whose continuation never arrives — the next byte is a
  1556. // fresh ASCII 'A' — must be costed as U+FFFD (3) for the orphaned lead, not
  1557. // folded into a phantom character. Driven byte-by-byte so the lead and the
  1558. // breaking byte land in separate `data` chunks, exercising accrueStrayCost's
  1559. // cross-chunk broken-sequence branch. The run completes and the bytes are
  1560. // captured (rendered U+FFFD by toString), proving the walk resynchronizes.
  1561. const { runtime } = await setup({ maxLogBytes: 1024 })
  1562. const result = await runtime.run({
  1563. program: [
  1564. 'import os',
  1565. 'os.write(1, b"\\xe4")',
  1566. 'os.sched_yield()',
  1567. 'os.write(1, b"A\\n")',
  1568. 'return None',
  1569. ].join('\n'),
  1570. bindings: [],
  1571. })
  1572. expect(result.error).toBeUndefined()
  1573. expect(result.logs.join('')).toContain('A')
  1574. expect(result.logs.join('')).toContain('�')
  1575. })
  1576. it('charges the exact serialized cost of short-escape and quote/backslash characters', async () => {
  1577. // Exercises every branch of jsonStringCostUpTo's per-character cost: a tab
  1578. // and other C0 controls with short JSON forms (\t etc., 2 bytes), a quote
  1579. // and backslash (2 bytes each), a `\uXXXX` control (6 bytes), a multibyte
  1580. // BMP character (raw UTF-8 width), and plain ASCII. Under a budget large
  1581. // enough to admit it, the line survives verbatim — proving the cost walker
  1582. // does not over- or under-charge and the string round-trips unescaped.
  1583. const { runtime } = await setup({ maxLogBytes: 4096 })
  1584. const result = await runtime.run({
  1585. program: ['import os', String.raw`os.write(1, "\ta\"b\\c\x01é\n".encode("utf-8"))`, 'return None'].join('\n'),
  1586. bindings: [],
  1587. })
  1588. expect(result.error).toBeUndefined()
  1589. expect(result.logs).toEqual(['\ta"b\\c\x01é'])
  1590. })
  1591. it('fails a completion dict with a non-string key as invalid-output (no key coercion)', async () => {
  1592. // json.dumps would coerce {1: "a", "1": "b"} to a single "1" key, silently
  1593. // dropping data. The shape validator rejects it before encoding.
  1594. const { runtime } = await setup()
  1595. const result = await runtime.run({
  1596. program: 'return {1: "first", "1": "second"}',
  1597. bindings: [],
  1598. })
  1599. expect(result.value).toBeUndefined()
  1600. expect(result.error?.kind).toBe('invalid-output')
  1601. expect(result.error?.message).toContain('non-string dict key')
  1602. })
  1603. it('rejects a binding argument with a non-string dict key before dispatch', async () => {
  1604. const { runtime } = await setup()
  1605. let called = false
  1606. const result = await runtime.run({
  1607. program: [
  1608. 'caught = ""',
  1609. 'try:',
  1610. ' await tools.sink({1: "x"})',
  1611. 'except RuntimeError as e:',
  1612. ' caught = str(e)',
  1613. 'return caught',
  1614. ].join('\n'),
  1615. bindings: tools({ sink: async () => { called = true; return null } }),
  1616. })
  1617. expect(result.error).toBeUndefined()
  1618. expect(result.value).toContain('lossless JSON')
  1619. expect(called).toBe(false)
  1620. })
  1621. it('fails a non-JSON completion value as invalid-output (no repr substitution)', async () => {
  1622. // A set is not lossless JSON. The old draft substituted repr(); the seam
  1623. // now requires refusing the run instead.
  1624. const { runtime } = await setup()
  1625. const result = await runtime.run({
  1626. program: 'return {1, 2, 3}',
  1627. bindings: [],
  1628. })
  1629. expect(result.value).toBeUndefined()
  1630. expect(result.error?.kind).toBe('invalid-output')
  1631. expect(result.error?.message).toContain('lossless JSON')
  1632. expect(result.error?.message).toContain('set')
  1633. })
  1634. it('fails a negative-zero completion value as invalid-output (sign bit is lossy over JSON)', async () => {
  1635. // JSON serialization turns -0.0 into 0 (or JS -0), silently changing the
  1636. // sign bit; the canonical lossless-JSON boundary rejects it, so the
  1637. // Python side must too — as a completion and as a binding argument.
  1638. const { runtime } = await setup()
  1639. const completion = await runtime.run({
  1640. program: 'return -0.0',
  1641. bindings: [],
  1642. })
  1643. expect(completion.error?.kind).toBe('invalid-output')
  1644. expect(completion.error?.message).toContain('negative zero')
  1645. const argument = await runtime.run({
  1646. program: [
  1647. 'try:',
  1648. ' await tools.echo(-0.0)',
  1649. ' return "accepted"',
  1650. 'except RuntimeError as e:',
  1651. ' return str(e)',
  1652. ].join('\n'),
  1653. bindings: tools({ echo: async args => args as never }),
  1654. })
  1655. expect(argument.error).toBeUndefined()
  1656. expect(argument.value).toContain('negative zero')
  1657. })
  1658. it('fails a NaN completion value as invalid-output (allow_nan=False)', async () => {
  1659. // json.dumps would happily emit NaN by default, but NaN is not JSON; the
  1660. // bootstrap passes allow_nan=False so it fails as invalid-output.
  1661. const { runtime } = await setup()
  1662. const result = await runtime.run({
  1663. program: 'return float("nan")',
  1664. bindings: [],
  1665. })
  1666. expect(result.error?.kind).toBe('invalid-output')
  1667. })
  1668. it('fails an over-budget completion value as output-limit (child-side check)', async () => {
  1669. const { runtime } = await setup({ maxValueBytes: 64 })
  1670. const result = await runtime.run({
  1671. program: 'return "V" * 5000',
  1672. bindings: [],
  1673. })
  1674. expect(result.value).toBeUndefined()
  1675. expect(result.error?.kind).toBe('output-limit')
  1676. expect(result.error?.message).toContain('exceeded 64 bytes')
  1677. })
  1678. it('meters a control-heavy completion value without materializing its escaped form', async () => {
  1679. // The child's lower bound admits a string by CHARACTER count, then the meter
  1680. // charged what `_dump_string(current).encode()` returned -- building the
  1681. // escaped copy plus its encode. Each NUL escapes to six bytes, so metering a
  1682. // value the budget then REJECTS allocated ~6x the original twice over:
  1683. // measured at 228.9 MiB of peak for a 20M-NUL string, against 19.1 MiB for
  1684. // the counting path that returns the identical 120,000,002 bytes. Past
  1685. // RLIMIT_AS the meter died as `exception: MemoryError`, inverting the
  1686. // `output-limit` this seam promises for an over-budget value.
  1687. //
  1688. // 8M NULs is 8,000,002 raw but 48,000,002 escaped: over the 16 MiB budget
  1689. // only when charged the escaped cost, so this also pins that the cheap
  1690. // character bound alone does not decide the verdict.
  1691. const { runtime } = await setup({ maxValueBytes: 16 * 1024 * 1024, maxWallMs: 60_000 })
  1692. const result = await runtime.run({
  1693. program: 'return "\\x00" * 8_000_000',
  1694. bindings: [],
  1695. })
  1696. expect(result.value).toBeUndefined()
  1697. expect(result.error?.kind).toBe('output-limit')
  1698. }, 90_000)
  1699. it('rejects a wide completion as output-limit before materializing its traversal state', async () => {
  1700. // `[0] * 2000000` sits far above maxValueBytes but well below the frame
  1701. // ceiling. The folded checker must reject it via the pre-enqueue bound —
  1702. // BEFORE pushing two million elements onto the walk — so a small
  1703. // addressSpaceMb does not turn the check itself into an RLIMIT_AS death.
  1704. const { runtime } = await setup({ maxValueBytes: 64, addressSpaceMb: 256, maxWallMs: 15_000 })
  1705. const result = await runtime.run({
  1706. program: 'return [0] * 2000000',
  1707. bindings: [],
  1708. })
  1709. expect(result.value).toBeUndefined()
  1710. expect(result.error?.kind).toBe('output-limit')
  1711. expect(result.error?.message).toContain('exceeded 64 bytes')
  1712. }, 20_000)
  1713. it('rejects a wide dict as output-limit without materializing its items list', async () => {
  1714. // Same pre-enqueue bound on the dict branch: `len(current)` replaces
  1715. // `list(current.items())`, which allocated one tuple per member before the
  1716. // bound could reject the value. Two million entries under a 64-byte cap
  1717. // fits the 256 MiB address space as a dict but not as a dict PLUS a
  1718. // two-million-tuple list.
  1719. const { runtime } = await setup({ maxValueBytes: 64, addressSpaceMb: 256, maxWallMs: 15_000 })
  1720. const result = await runtime.run({
  1721. program: 'return {str(i): 0 for i in range(2000000)}',
  1722. bindings: [],
  1723. })
  1724. expect(result.value).toBeUndefined()
  1725. expect(result.error?.kind).toBe('output-limit')
  1726. expect(result.error?.message).toContain('exceeded 64 bytes')
  1727. }, 20_000)
  1728. it('meters a float completion in the host\'s number spelling', async () => {
  1729. // CPython's repr disagrees with the host's String(number): `1.0` is three
  1730. // bytes here and one there, `1e-07` pads the exponent the host writes as
  1731. // `1e-7`. Both sides meter the SAME budget, so the child must count the
  1732. // bytes the host will receive — otherwise a boundary-sized value is
  1733. // falsely reported as output-limit.
  1734. const { runtime } = await setup({ maxValueBytes: 1 })
  1735. const integral = await runtime.run({ program: 'return 1.0', bindings: [] })
  1736. expect(integral.error).toBeUndefined()
  1737. expect(integral.value).toBe(1)
  1738. const exponent = await setup({ maxValueBytes: 4 })
  1739. const small = await exponent.runtime.run({ program: 'return 1e-7', bindings: [] })
  1740. expect(small.error).toBeUndefined()
  1741. expect(small.value).toBe(1e-7)
  1742. // The spelling is a meter input, not a licence to overshoot: `1.5` is three
  1743. // bytes on both sides and still fails a two-byte budget.
  1744. const tight = await setup({ maxValueBytes: 2 })
  1745. const over = await tight.runtime.run({ program: 'return 1.5', bindings: [] })
  1746. expect(over.error?.kind).toBe('output-limit')
  1747. })
  1748. it('carries floats across the wire in the host\'s number spelling', async () => {
  1749. // The child ENCODES with the same speller it meters with, so the frame the
  1750. // host parses must reproduce every double exactly — including the branches
  1751. // where CPython and ECMAScript disagree (integral floats, sub-1e-6
  1752. // exponents, >= 1e21, and beyond-safe-range integral doubles whose exact
  1753. // digits differ from the shortest round-trip form).
  1754. const { runtime } = await setup()
  1755. const result = await runtime.run({
  1756. program: 'return [1.0, 100.0, 1.5, 0.1, 1e-7, 1e-6, 1e-5, 123.456, -2.5e-8, 1e21, float(2**60), 5e-324, 1.7976931348623157e308]',
  1757. bindings: [],
  1758. })
  1759. expect(result.error).toBeUndefined()
  1760. expect(result.value).toEqual([1, 100, 1.5, 0.1, 1e-7, 1e-6, 1e-5, 123.456, -2.5e-8, 1e21, 2 ** 60, 5e-324, 1.7976931348623157e308])
  1761. })
  1762. it('rejects a forged non-lossless done value host-side as invalid-output', async () => {
  1763. // A forged done frame bypasses the child's _check_done_value. JSON.parse
  1764. // turns 1e400 into Infinity; validateChildFrame no longer scans done.value,
  1765. // so the host's own checkDoneValue must catch the non-lossless number.
  1766. const { runtime } = await setup()
  1767. const result = await runtime.run({
  1768. program: [
  1769. 'import os',
  1770. String.raw`os.write(3, b'{"type":"done","value":1e400}' + b'\n')`,
  1771. 'import time',
  1772. 'time.sleep(5)',
  1773. ].join('\n'),
  1774. bindings: [],
  1775. })
  1776. expect(result.value).toBeUndefined()
  1777. expect(result.error?.kind).toBe('invalid-output')
  1778. expect(result.error?.message).toContain('non-lossless number')
  1779. })
  1780. it('reports a syntax error as an exception without settling with a value', async () => {
  1781. const { runtime } = await setup()
  1782. const result = await runtime.run({
  1783. program: '$$invalid python$$',
  1784. bindings: [],
  1785. })
  1786. expect(result.error?.kind).toBe('exception')
  1787. expect(result.error?.message).toContain('SyntaxError')
  1788. // The parse-time diagnostic must carry the same source label as compile and
  1789. // runtime tracebacks (ast.parse passes filename="<model>"); a stale
  1790. // "<unknown>" label would leak an inconsistent origin to the model.
  1791. expect(result.error?.message).toContain('File \"<model>\"')
  1792. expect(result.value).toBeUndefined()
  1793. })
  1794. it('reports a runtime raise as an exception with the traceback', async () => {
  1795. const { runtime } = await setup()
  1796. const result = await runtime.run({
  1797. program: 'raise ValueError("intentional")',
  1798. bindings: [],
  1799. })
  1800. expect(result.error?.kind).toBe('exception')
  1801. expect(result.error?.message).toContain('ValueError')
  1802. expect(result.error?.message).toContain('intentional')
  1803. })
  1804. it('bounds a deep exception cause chain instead of burning the wall budget formatting it', async () => {
  1805. // A chain thousands of links deep would make the rendering walk and
  1806. // format() linear in its length, consuming maxWallMs. Rendering is capped
  1807. // at 100 links with a marker; the run reports the exception well within
  1808. // budget rather than timing out.
  1809. const { runtime } = await setup({ maxValueBytes: 1024 * 1024, maxWallMs: 20_000 })
  1810. const start = Date.now()
  1811. const result = await runtime.run({
  1812. program: [
  1813. 'err = None',
  1814. 'for i in range(3000):',
  1815. ' try:',
  1816. ' raise ValueError(i) from err',
  1817. ' except ValueError as e:',
  1818. ' err = e',
  1819. 'raise err',
  1820. ].join('\n'),
  1821. bindings: [],
  1822. })
  1823. expect(result.error?.kind).toBe('exception')
  1824. expect(result.error?.message).toContain('exception chain truncated at 100 links')
  1825. expect(Date.now() - start).toBeLessThan(15_000)
  1826. }, 25_000)
  1827. it('bounds an over-cap chain without assigning to the live exception', async () => {
  1828. // The cap used to be applied by severing the over-cap link ON the live
  1829. // exception. An exception class overriding __setattr__ to raise turned that
  1830. // assignment into model code running inside the bootstrap's failure
  1831. // handler; the throw skipped the `done` send that sits after the handler,
  1832. // so the host blocked on fd 3 and reported a maxWallMs timeout instead of
  1833. // the model's own exception. Cutting the chain on the TracebackException
  1834. // COPY touches no model hook, so the marker still appears and the run
  1835. // reports `exception`.
  1836. const { runtime } = await setup({ maxValueBytes: 1024 * 1024, maxWallMs: 15_000 })
  1837. const start = Date.now()
  1838. const result = await runtime.run({
  1839. program: [
  1840. 'class Sealed(Exception):',
  1841. ' def __setattr__(self, name, value):',
  1842. ' raise RuntimeError("live mutation refused")',
  1843. 'err = None',
  1844. 'for i in range(150):',
  1845. ' try:',
  1846. ' raise Sealed(i) from err',
  1847. ' except Sealed as e:',
  1848. ' err = e',
  1849. 'raise err',
  1850. ].join('\n'),
  1851. bindings: [],
  1852. })
  1853. expect(result.error?.kind).toBe('exception')
  1854. expect(result.error?.message).toContain('Sealed')
  1855. expect(result.error?.message).toContain('exception chain truncated at 100 links')
  1856. // The sever attempt is what used to leak: its message must not appear, and
  1857. // the run must settle well inside the wall budget rather than timing out.
  1858. expect(result.error?.message).not.toContain('live mutation refused')
  1859. expect(Date.now() - start).toBeLessThan(10_000)
  1860. }, 20_000)
  1861. it('still sends done when rendering the diagnostic itself raises', async () => {
  1862. // format() reaches the exception's own __str__, so a model class whose
  1863. // __str__ raises can throw from inside the failure handler. CPython's
  1864. // _safe_string absorbs a raising __str__ during formatting, but the
  1865. // fallback must hold for any throw on that path (a raising __repr__ of an
  1866. // argument, a MemoryError under RLIMIT_AS), so the assertion is the
  1867. // invariant that matters: a `done` frame carrying `exception`, never a
  1868. // timeout, and never the failing renderer's own message.
  1869. const { runtime } = await setup({ maxWallMs: 10_000 })
  1870. const result = await runtime.run({
  1871. program: [
  1872. 'class Unprintable(Exception):',
  1873. ' def __str__(self):',
  1874. ' raise RuntimeError("str refused")',
  1875. ' def __repr__(self):',
  1876. ' raise RuntimeError("repr refused")',
  1877. 'raise Unprintable()',
  1878. ].join('\n'),
  1879. bindings: [],
  1880. })
  1881. expect(result.error?.kind).toBe('exception')
  1882. expect(result.error?.message).toContain('Unprintable')
  1883. expect(result.error?.message).not.toContain('str refused')
  1884. expect(result.error?.message).not.toContain('repr refused')
  1885. }, 15_000)
  1886. it('sends done with an inert diagnostic when the whole rendering path raises', async () => {
  1887. // Drive the fallback itself. `TracebackException.format` reads the
  1888. // exception class's `__module__` to decide whether to qualify the name, and
  1889. // a metaclass property can raise there — a throw INSIDE the formatter,
  1890. // reached with no rebinding of anything the bootstrap owns. Without the
  1891. // wrapper it escapes the handler, the `done` send never runs, and the host
  1892. // times out at maxWallMs.
  1893. const { runtime } = await setup({ maxWallMs: 10_000 })
  1894. const result = await runtime.run({
  1895. program: [
  1896. 'class Meta(type):',
  1897. ' @property',
  1898. ' def __module__(cls):',
  1899. ' raise RuntimeError("renderer refused")',
  1900. 'class Hostile(ValueError, metaclass=Meta):',
  1901. ' pass',
  1902. 'raise Hostile("original failure")',
  1903. ].join('\n'),
  1904. bindings: [],
  1905. })
  1906. expect(result.error?.kind).toBe('exception')
  1907. // The inert fallback names the class and a fixed literal; it must not carry
  1908. // the renderer's message, and must not have become a timeout. `__name__` is
  1909. // still a plain str here, so the class name survives.
  1910. expect(result.error?.message).toBe('Hostile: <diagnostic rendering failed>')
  1911. }, 15_000)
  1912. it('falls back to a placeholder class name when __name__ itself raises', async () => {
  1913. // The fallback reads type(exc).__name__, which a metaclass property can
  1914. // hijack. It must neither run that override's failure into the handler nor
  1915. // format a non-str __name__ into the message. The hostile `__module__` is
  1916. // what drives execution into the fallback in the first place.
  1917. const { runtime } = await setup({ maxWallMs: 10_000 })
  1918. const result = await runtime.run({
  1919. program: [
  1920. 'class Meta(type):',
  1921. ' @property',
  1922. ' def __module__(cls):',
  1923. ' raise RuntimeError("renderer refused")',
  1924. ' @property',
  1925. ' def __name__(cls):',
  1926. ' raise RuntimeError("name refused")',
  1927. 'class Nameless(Exception, metaclass=Meta):',
  1928. ' pass',
  1929. 'raise Nameless()',
  1930. ].join('\n'),
  1931. bindings: [],
  1932. })
  1933. expect(result.error?.kind).toBe('exception')
  1934. expect(result.error?.message).toBe('<unknown>: <diagnostic rendering failed>')
  1935. }, 15_000)
  1936. it('reports the real exception when the program rebinds every name the failure path uses', async () => {
  1937. // The bootstrap IS __main__, so `import __main__; __main__._X = ...` reaches
  1938. // any module global a call-time lookup would read. The failure path is the
  1939. // worst place for that: the reporter, the byte cap, the traceback formatter,
  1940. // the settlement flush and the `done` send all run AFTER the `except` block,
  1941. // so a replacement that raises skips the send, leaves the host blocked on
  1942. // fd 3, and the run reports a maxWallMs timeout instead of the model's own
  1943. // exception. Rebind all of them at once; the run must still carry the real
  1944. // ValueError.
  1945. const { runtime } = await setup({ maxWallMs: 10_000 })
  1946. const result = await runtime.run({
  1947. program: [
  1948. 'import __main__',
  1949. 'def boom(*a, **k):',
  1950. ' raise RuntimeError("hijacked")',
  1951. '__main__._SAFE_MODEL_TRACEBACK = boom',
  1952. '__main__._cap_message = boom',
  1953. '__main__._model_traceback = boom',
  1954. '__main__._UNRENDERABLE_DIAGNOSTIC = boom',
  1955. '__main__._LogStream.flush_line = boom',
  1956. // `send_done` writes via LOCALLY-BOUND `_encode_json_plain` +
  1957. // `ProtocolChannel.write_encoded`; rebinding these at call time must not
  1958. // redirect the done frame (a late lookup would be `boom` -> worker-exit).
  1959. '__main__.ProtocolChannel.send_sync = boom',
  1960. '__main__.ProtocolChannel.write_encoded = boom',
  1961. '__main__._encode_json_plain = boom',
  1962. 'raise ValueError("real failure")',
  1963. ].join('\n'),
  1964. bindings: [],
  1965. })
  1966. expect(result.error?.kind).toBe('exception')
  1967. expect(result.error?.message).toContain('ValueError: real failure')
  1968. expect(result.error?.message).not.toContain('hijacked')
  1969. }, 15_000)
  1970. it('still delivers a done frame when a transitive encode name is rebound', async () => {
  1971. // `send_done` binds `_encode_json_plain` and `ProtocolChannel.write_encoded`
  1972. // into locals, but those callables' BODIES still resolve transitive module
  1973. // globals at call time: `_encode_json_plain` reaches `_dump_scalar`/`_dump_string`/
  1974. // `json.dumps`, and `write_encoded` reaches `os.write`. This bootstrap is
  1975. // `__main__`, so rebinding `__main__._dump_scalar` to a raising function makes
  1976. // the error-frame encode throw AFTER the `except` block. `send_done` catches
  1977. // that and writes a fixed literal done frame (kind `exception`) with the
  1978. // LOCALLY-BOUND `_os_write`/`_memoryview`/`_FALLBACK_DONE_FRAME` captured
  1979. // before the program runs, so the host still gets a verdict — the run must be an
  1980. // `exception`, never a `worker-exit`. The real message is lost (the literal
  1981. // carries a fixed `<unrenderable>` text), which is acceptable: the verdict
  1982. // outranks the diagnostic detail.
  1983. const { runtime } = await setup({ maxWallMs: 10_000 })
  1984. const result = await runtime.run({
  1985. program: [
  1986. 'import __main__',
  1987. 'def boom(*a, **k):',
  1988. ' raise RuntimeError("hijacked")',
  1989. '__main__._dump_scalar = boom',
  1990. '__main__.os = boom',
  1991. // The fallback must also survive a rebind of its own primitives.
  1992. '__main__._os_write = boom',
  1993. '__main__._memoryview = boom',
  1994. '__main__._FALLBACK_DONE_FRAME = boom',
  1995. 'raise ValueError("real failure")',
  1996. ].join('\n'),
  1997. bindings: [],
  1998. })
  1999. expect(result.error?.kind).toBe('exception')
  2000. expect(result.error?.kind).not.toBe('worker-exit')
  2001. }, 15_000)
  2002. it('still reports a model exception when the program rebinds BaseException', async () => {
  2003. // `_run`'s outer try/except catches the program's failure and builds a
  2004. // `done` frame. The clause previously used the module-global `BaseException`,
  2005. // which the program (running as `__main__`) can rebind: `__main__.BaseException
  2006. // = RuntimeError` makes the `except BaseException` resolve to `RuntimeError`,
  2007. // so a subsequent `ValueError` does not match and escapes `_run` with no
  2008. // `done` frame — misreporting the run as a `worker-exit`. The exception class
  2009. // is now bound into a `_run` LOCAL before the program runs, so the rebind
  2010. // cannot change which class the clause catches; the run must still report an
  2011. // `exception`, not a `worker-exit`.
  2012. const { runtime } = await setup({ maxWallMs: 10_000 })
  2013. const result = await runtime.run({
  2014. program: [
  2015. 'import __main__',
  2016. '__main__.BaseException = RuntimeError',
  2017. 'raise ValueError("real failure")',
  2018. ].join('\n'),
  2019. bindings: [],
  2020. })
  2021. expect(result.error?.kind).toBe('exception')
  2022. expect(result.error?.kind).not.toBe('worker-exit')
  2023. }, 15_000)
  2024. it('still reports the exception when BaseException and the traceback reporter are rebound together', async () => {
  2025. // The two rebind families compose: `__main__.BaseException = ValueError`
  2026. // must not change which class the `_run` catch resolves (it is a pre-program
  2027. // local), and a rebound reporter (`_SAFE_MODEL_TRACEBACK`/`_cap_message`/
  2028. // `_model_traceback`/`_UNRENDERABLE_DIAGNOSTIC`) must not break the done
  2029. // frame — `safe_model_traceback` holds its primitives as import-time closure
  2030. // cells. A `KeyError` (not a `ValueError` subclass) escapes a catch that
  2031. // resolves to the rebound class, so without the local binding the run would
  2032. // misreport as `worker-exit`; with it, the run reports the exception and the
  2033. // fallback reporter still produces the fixed literal.
  2034. const { runtime } = await setup({ maxWallMs: 10_000 })
  2035. const result = await runtime.run({
  2036. program: [
  2037. 'import __main__',
  2038. 'def boom(*a, **k):',
  2039. ' raise RuntimeError("hijacked")',
  2040. '__main__.BaseException = ValueError',
  2041. '__main__._SAFE_MODEL_TRACEBACK = boom',
  2042. '__main__._cap_message = boom',
  2043. '__main__._model_traceback = boom',
  2044. '__main__._UNRENDERABLE_DIAGNOSTIC = boom',
  2045. 'raise KeyError("real failure")',
  2046. ].join('\n'),
  2047. bindings: [],
  2048. })
  2049. expect(result.error?.kind).toBe('exception')
  2050. expect(result.error?.kind).not.toBe('worker-exit')
  2051. }, 15_000)
  2052. it('rejects an fd-3 frame whose raw length exceeds the parse cap before joining it', async () => {
  2053. // The 64 MiB frame parse cap bounds the RAW frame bytes, not the decoded
  2054. // structure; a compact wide frame near that ceiling could decode to far
  2055. // more host memory. The unframed-buffer counter is checked against
  2056. // FRAME_PARSE_CAP_BYTES BEFORE the Buffer.concat join, so an oversized
  2057. // frame is dropped at one copy of its wire bytes instead of being fully
  2058. // joined (a second copy) and only then discarded in the line loop — the
  2059. // peak-memory doubling the pre-join check exists to prevent. Fail-before:
  2060. // without the check the frame is joined whole and parsed (its log text
  2061. // admitted, truncating the ledger), and the run completes normally.
  2062. const { runtime } = await setup({ maxWallMs: 60_000 })
  2063. const result = await runtime.run({
  2064. program: [
  2065. 'import os',
  2066. // One frame just past the 64 MiB parse cap.
  2067. 'os.write(3, b"{\\"type\\":\\"log\\",\\"text\\":\\"" + b"a" * (65 * 1024 * 1024) + b"\\"}\\n")',
  2068. 'return "done"',
  2069. ].join('\n'),
  2070. bindings: [],
  2071. })
  2072. expect(result.error?.kind).toBe('worker-exit')
  2073. expect(result.error?.message).toContain('protocol frame exceeded')
  2074. }, 90_000)
  2075. it('caps an oversized rejection diagnostic so an invalid completion stays invalid-output', async () => {
  2076. // _done_with_value caps its rejection diagnostic through _cap_message: a
  2077. // hostile class name (huge type(value).__name__) would otherwise push the
  2078. // done frame past the host's 64 MiB parse cap, misreporting an
  2079. // invalid-output run as a worker-exit. The diagnostic is capped to the
  2080. // value budget, so the frame always crosses the parser.
  2081. const { runtime } = await setup({ maxWallMs: 60_000 })
  2082. const result = await runtime.run({
  2083. program: [
  2084. 'return type("N" * (70 * 1024 * 1024), (), {})()',
  2085. ].join('\n'),
  2086. bindings: [],
  2087. })
  2088. expect(result.error?.kind).toBe('invalid-output')
  2089. expect(result.error?.kind).not.toBe('worker-exit')
  2090. }, 90_000)
  2091. it('appends a flushed unterminated line to the next entry without a fake newline', async () => {
  2092. // An explicit flush of an unterminated line (print(..., end='', flush=True))
  2093. // used to push a full log frame, so the following print() landed in a
  2094. // SECOND entry and logs.join('\n') rendered 'a\nb' for what the program
  2095. // printed as one line. The flush frame now carries `open: true` and the
  2096. // host appends the next frame to the same entry.
  2097. const { runtime } = await setup()
  2098. const result = await runtime.run({
  2099. program: [
  2100. "print('a', end='', flush=True)",
  2101. "print('b')",
  2102. 'return "done"',
  2103. ].join('\n'),
  2104. bindings: [],
  2105. })
  2106. expect(result.error).toBeUndefined()
  2107. expect(result.logs).toEqual(['ab'])
  2108. }, 15_000)
  2109. it('keeps a SEALED open hold when the run ends with it still open', async () => {
  2110. // The finish-residual's sealed side: an open hold past MAX_PENDING_CHUNKS
  2111. // lands in openSealed, and the run ends without a closing frame — finish()
  2112. // must commit the SEALED prefix, not only the current fragments.
  2113. const { runtime } = await setup({ maxLogBytes: 65536 })
  2114. const result = await runtime.run({
  2115. program: [
  2116. 'import os',
  2117. "os.write(3, b'{\"type\":\"log\",\"text\":\"x\",\"open\":true}\\n')",
  2118. 'for _ in range(3000):',
  2119. " os.write(3, b'{\"type\":\"log\",\"text\":\"a\",\"open\":true}\\n')",
  2120. 'return "done"',
  2121. ].join('\n'),
  2122. bindings: [],
  2123. })
  2124. expect(result.error).toBeUndefined()
  2125. expect(result.logs).toEqual(['x' + 'a'.repeat(3000)])
  2126. }, 15_000)
  2127. it('keeps a flushed unterminated line when the run ends with it still open', async () => {
  2128. // The settlement flush pushes the residual with `open: true`; finish()
  2129. // admits it so a program that commits a partial line and returns does not
  2130. // lose it from logs.
  2131. const { runtime } = await setup()
  2132. const result = await runtime.run({
  2133. program: [
  2134. "print('committed', end='', flush=True)",
  2135. 'return "done"',
  2136. ].join('\n'),
  2137. bindings: [],
  2138. })
  2139. expect(result.error).toBeUndefined()
  2140. expect(result.logs).toEqual(['committed'])
  2141. }, 15_000)
  2142. it('skips the hold for a zero-content open continuation', async () => {
  2143. // An empty open continuation bills 0 and is NOT pushed into the held
  2144. // fragment array (an empty fragment contributes nothing to the merged
  2145. // entry, and holding it would let a forged empty-open flood grow host
  2146. // memory without touching the ledger).
  2147. const { runtime } = await setup({ maxLogBytes: 64 })
  2148. const result = await runtime.run({
  2149. program: [
  2150. 'import os',
  2151. "os.write(3, b'{\"type\":\"log\",\"text\":\"x\",\"open\":true}\\n')",
  2152. "os.write(3, b'{\"type\":\"log\",\"text\":\"\",\"open\":true}\\n')",
  2153. "os.write(3, b'{\"type\":\"log\",\"text\":\"y\"}\\n')",
  2154. 'return "done"',
  2155. ].join('\n'),
  2156. bindings: [],
  2157. })
  2158. expect(result.error).toBeUndefined()
  2159. expect(result.logs).toEqual(['xy'])
  2160. }, 15_000)
  2161. it('seals the open hold past MAX_PENDING_CHUNKS without changing the merged entry', async () => {
  2162. // A budget-sized single-character open flood would otherwise accumulate
  2163. // thousands of fragment array slots (each a slot plus string header, ~30x
  2164. // overhead the byte cap cannot see). The hold seals into one block past
  2165. // MAX_PENDING_CHUNKS; the merged entry is byte-identical.
  2166. const { runtime } = await setup({ maxLogBytes: 65536 })
  2167. const result = await runtime.run({
  2168. program: [
  2169. 'import os',
  2170. "os.write(3, b'{\"type\":\"log\",\"text\":\"x\",\"open\":true}\\n')",
  2171. // 3000 single-character open continuations (over MAX_PENDING_CHUNKS).
  2172. 'for _ in range(3000):',
  2173. " os.write(3, b'{\"type\":\"log\",\"text\":\"a\",\"open\":true}\\n')",
  2174. "os.write(3, b'{\"type\":\"log\",\"text\":\"y\"}\\n')",
  2175. 'return "done"',
  2176. ].join('\n'),
  2177. bindings: [],
  2178. })
  2179. expect(result.error).toBeUndefined()
  2180. expect(result.logs).toEqual(['x' + 'a'.repeat(3000) + 'y'])
  2181. }, 15_000)
  2182. it('bounds a forged open-frame flood against the log budget', async () => {
  2183. // The open hold must be bounded by the ledger: without the exact-cost check
  2184. // a forged open flood would grow the held fragment without touching
  2185. // logBudget — unbounded host retention under a small budget. The flood now
  2186. // truncates to the marker like any over-budget log traffic.
  2187. const { runtime } = await setup({ maxLogBytes: 64 })
  2188. const result = await runtime.run({
  2189. program: [
  2190. 'import os',
  2191. // 2000 forged open frames, each under the frame parse cap.
  2192. 'for _ in range(2000):',
  2193. " os.write(3, b'{\"type\":\"log\",\"text\":\"a\",\"open\":true}\\n')",
  2194. 'return "done"',
  2195. ].join('\n'),
  2196. bindings: [],
  2197. })
  2198. expect(result.error).toBeUndefined()
  2199. expect(result.logs).toEqual(['a'.repeat(60), logTruncationMarker(64)])
  2200. }, 15_000)
  2201. it('commits a sealed open hold before the truncation marker', async () => {
  2202. // The sealed variant of the prefix-commit case: an open flood past
  2203. // MAX_PENDING_CHUNKS lands in openSealed, then an over-budget line
  2204. // truncates — truncateLogs must commit the SEALED prefix (not only the
  2205. // current fragments) before the marker.
  2206. const { runtime } = await setup({ maxLogBytes: 65536 })
  2207. const result = await runtime.run({
  2208. program: [
  2209. 'import os',
  2210. "os.write(3, b'{\"type\":\"log\",\"text\":\"x\",\"open\":true}\\n')",
  2211. // 3000 single-character open continuations seal the hold, then a
  2212. // forged over-budget open frame trips the ledger: truncateLogs must
  2213. // commit the SEALED prefix before the marker.
  2214. 'for _ in range(3000):',
  2215. " os.write(3, b'{\"type\":\"log\",\"text\":\"a\",\"open\":true}\\n')",
  2216. "os.write(3, ('{\"type\":\"log\",\"text\":\"' + 'z' * 70000 + '\",\"open\":true}\\n').encode())",
  2217. 'return "done"',
  2218. ].join('\n'),
  2219. bindings: [],
  2220. })
  2221. expect(result.error).toBeUndefined()
  2222. expect(result.logs[0]).toBe('x' + 'a'.repeat(3000))
  2223. expect(result.logs[result.logs.length - 1]).toBe(logTruncationMarker(65536))
  2224. }, 15_000)
  2225. it('commits a flushed open prefix before the truncation marker', async () => {
  2226. // A flushed unterminated line is billed and committed; when a later
  2227. // over-budget write truncates, the committed prefix must appear BEFORE the
  2228. // marker — the ledger charged for it, so it cannot vanish.
  2229. const { runtime } = await setup({ maxLogBytes: 64 })
  2230. const result = await runtime.run({
  2231. program: [
  2232. "print('committed', end='', flush=True)",
  2233. "print('x' * 100)",
  2234. 'return "done"',
  2235. ].join('\n'),
  2236. bindings: [],
  2237. })
  2238. expect(result.error).toBeUndefined()
  2239. expect(result.logs).toEqual(['committed', logTruncationMarker(64)])
  2240. }, 15_000)
  2241. it('drops a forged fd-3 frame with illegal UTF-8 instead of accepting a mangled value', async () => {
  2242. // toString('utf8') would replace the illegal 0xFF with U+FFFD, so a forged
  2243. // done frame could land a corrupted completion value; the fatal decode
  2244. // throws and the frame is dropped. The program's real return still settles
  2245. // the run with the honest value.
  2246. const { runtime } = await setup()
  2247. const result = await runtime.run({
  2248. program: [
  2249. 'import os',
  2250. "os.write(3, b'{\"type\":\"done\",\"value\":\"bad' + bytes([0xFF]) + b'\"}\\n')",
  2251. 'return "ok"',
  2252. ].join('\n'),
  2253. bindings: [],
  2254. })
  2255. expect(result.error).toBeUndefined()
  2256. expect(result.value).toBe('ok')
  2257. }, 15_000)
  2258. it('no-ops a closing frame once an open flood already truncated the ledger', async () => {
  2259. // The closing-frame branch's post-truncation arm: an open flood exhausts
  2260. // the ledger (logsTruncated set, marker pushed), then a closing frame
  2261. // arrives — it must be a no-op, not append content past the marker.
  2262. const { runtime } = await setup({ maxLogBytes: 64 })
  2263. const result = await runtime.run({
  2264. program: [
  2265. 'import os',
  2266. 'for _ in range(2000):',
  2267. " os.write(3, b'{\"type\":\"log\",\"text\":\"a\",\"open\":true}\\n')",
  2268. "os.write(3, b'{\"type\":\"log\",\"text\":\"b\"}\\n')",
  2269. 'return "done"',
  2270. ].join('\n'),
  2271. bindings: [],
  2272. })
  2273. expect(result.error).toBeUndefined()
  2274. expect(result.logs).toEqual(['a'.repeat(60), logTruncationMarker(64)])
  2275. }, 15_000)
  2276. it('bills a merged open entry once, not per fragment', async () => {
  2277. // A merged entry's wire cost is billed ONCE, split across its fragments
  2278. // (first fragment pays quotes+separator, continuations pay only content).
  2279. // Under maxLogBytes: 64, 16 single-character flushes merge to one 16-char
  2280. // entry (2 quotes + 16 content + 1 separator = 19), which fits; per-
  2281. // fragment billing (each charged quotes+separator, ~4 bytes) would truncate
  2282. // at 16 x 4 = 64.
  2283. const { runtime } = await setup({ maxLogBytes: 64 })
  2284. const result = await runtime.run({
  2285. program: [
  2286. 'for _ in range(16):',
  2287. " print('x', end='', flush=True)",
  2288. "print('')",
  2289. 'return "done"',
  2290. ].join('\n'),
  2291. bindings: [],
  2292. })
  2293. expect(result.error).toBeUndefined()
  2294. expect(result.logs).toEqual(['x'.repeat(16)])
  2295. }, 15_000)
  2296. it('admits a compliant merged entry whose closing frame fits the remaining budget', async () => {
  2297. // The review's arithmetic check: print('a'*30, flush); print('b'*25) under
  2298. // maxLogBytes: 64 has a merged wire cost of 2 quotes + 55 content + 1
  2299. // separator = 58 <= 63, so it MUST be admitted as one entry. The earlier
  2300. // cap math (logBudget - openCost) made the closing frame's walk see a
  2301. // negative cap and truncate a compliant entry.
  2302. const { runtime } = await setup({ maxLogBytes: 64 })
  2303. const result = await runtime.run({
  2304. program: [
  2305. "print('a' * 30, end='', flush=True)",
  2306. "print('b' * 25)",
  2307. 'return "done"',
  2308. ].join('\n'),
  2309. bindings: [],
  2310. })
  2311. expect(result.error).toBeUndefined()
  2312. expect(result.logs).toEqual(['a'.repeat(30) + 'b'.repeat(25)])
  2313. }, 15_000)
  2314. it('rejects an open frame that would overflow the ledger by one byte', async () => {
  2315. // The review's arithmetic check: an open frame whose full JSON cost is 63
  2316. // (maxLogBytes: 64 -> ledger 63) must be rejected by the first-fragment
  2317. // cap logBudget - 1 (62), not admitted with a bill of 64 that pushes the
  2318. // ledger negative. The frame is FORGED on fd 3 so the child ledger cannot
  2319. // truncate first: a reverted cap of logBudget (63) would admit the frame,
  2320. // hold it, and flush it at settlement, so the marker assertion fails.
  2321. const { runtime } = await setup({ maxLogBytes: 64 })
  2322. const result = await runtime.run({
  2323. program: [
  2324. 'import os',
  2325. "os.write(3, ('{\"type\":\"log\",\"text\":\"' + 'x' * 61 + '\",\"open\":true}\\n').encode())",
  2326. 'return "done"',
  2327. ].join('\n'),
  2328. bindings: [],
  2329. })
  2330. expect(result.error).toBeUndefined()
  2331. expect(result.logs).toEqual([logTruncationMarker(64)])
  2332. }, 15_000)
  2333. it('bills the closing frame as the merged tail under an exact-fit budget', async () => {
  2334. // The child's split billing: a 30-char open + a 30-char closing frame cost
  2335. // 2 + 60 + 1 = 63 = ledger 63 exactly; the closing frame must be billed as
  2336. // the merged tail (content only), not as a fresh entry (which would
  2337. // double-charge the quotes+separator and truncate an exact-fit entry).
  2338. const { runtime } = await setup({ maxLogBytes: 64 })
  2339. const result = await runtime.run({
  2340. program: [
  2341. "print('a' * 30, end='', flush=True)",
  2342. "print('b' * 30)",
  2343. 'return "done"',
  2344. ].join('\n'),
  2345. bindings: [],
  2346. })
  2347. expect(result.error).toBeUndefined()
  2348. expect(result.logs).toEqual(['a'.repeat(30) + 'b'.repeat(30)])
  2349. }, 15_000)
  2350. it('does not over-reject an exact-fit closing line while an open entry accumulates', async () => {
  2351. // The write-path pre-check's cheap bound used +3 (quotes + separator) even
  2352. // while an open entry was accumulating, so an exact-fit merged TAIL was
  2353. // truncated. The recipe below goes through the SCAN pre-check (the
  2354. // newline-terminated write arrives with an empty pending buffer, so the
  2355. // buffered-chunks branch is skipped): 'a'*29 flush bills 32 (ledger 31
  2356. // left), then one write of 'b'*30 + newline merges 30 more chars whose
  2357. // cheap bound is 30, not 33 — the +3 form saw 30 + 3 = 33 > 31, sliced to
  2358. // a budget prefix, and pushed past the ledger, emitting the marker for a
  2359. // line that fits (merged cost 2 + 59 + 1 = 62 <= 63).
  2360. const { runtime } = await setup({ maxLogBytes: 64 })
  2361. const result = await runtime.run({
  2362. program: [
  2363. 'import sys',
  2364. "sys.stdout.write('a' * 29)",
  2365. 'sys.stdout.flush()',
  2366. "sys.stdout.write('b' * 30 + chr(10))",
  2367. 'return "done"',
  2368. ].join('\n'),
  2369. bindings: [],
  2370. })
  2371. expect(result.error).toBeUndefined()
  2372. expect(result.logs).toEqual(['a'.repeat(29) + 'b'.repeat(30)])
  2373. }, 15_000)
  2374. it('rejects a new open entry once the ledger has only two bytes left', async () => {
  2375. // The jsonStringCostUpTo sub-2-byte guard: forged open frames drive the
  2376. // host ledger down to 1 byte, then a new open entry's first-fragment cap
  2377. // (logBudget - 1 = 0) trips the guard and truncates.
  2378. const { runtime } = await setup({ maxLogBytes: 64 })
  2379. const result = await runtime.run({
  2380. program: [
  2381. 'import os',
  2382. "os.write(3, ('{\"type\":\"log\",\"text\":\"' + 'a' * 28 + '\",\"open\":true}\\n').encode())",
  2383. "os.write(3, ('{\"type\":\"log\",\"text\":\"' + 'a' * 31 + '\"}\\n').encode())",
  2384. "os.write(3, b'{\"type\":\"log\",\"text\":\"x\",\"open\":true}\\n')",
  2385. 'return "done"',
  2386. ].join('\n'),
  2387. bindings: [],
  2388. })
  2389. expect(result.error).toBeUndefined()
  2390. expect(result.logs).toEqual(['a'.repeat(59), logTruncationMarker(64)])
  2391. }, 15_000)
  2392. it('truncates when the closing frame of a merged entry overflows the budget', async () => {
  2393. // The merged entry's billed-once cost: an open fragment that nearly
  2394. // exhausts the budget, then a closing frame whose content no longer fits —
  2395. // the closing frame's exact-cost walk trips and the marker replaces the
  2396. // entry, exactly like any other over-budget log traffic.
  2397. const { runtime } = await setup({ maxLogBytes: 64 })
  2398. const result = await runtime.run({
  2399. program: [
  2400. "print('x' * 40, end='', flush=True)",
  2401. "print('y' * 40)",
  2402. 'return "done"',
  2403. ].join('\n'),
  2404. bindings: [],
  2405. })
  2406. expect(result.error).toBeUndefined()
  2407. expect(result.logs).toEqual(['x'.repeat(40), logTruncationMarker(64)])
  2408. }, 15_000)
  2409. it('keeps a float completion exact when the program mutates the decimal context', async () => {
  2410. // The float encoder's Decimal(repr(value)).normalize() used the process
  2411. // GLOBAL decimal context: a legitimate program setting
  2412. // `getcontext().prec = 2` silently rounded the completion value's digits,
  2413. // and `traps[Inexact] = True` made the encode raise, misclassifying a
  2414. // successful run as an exception. A fixed module-level Context(prec=28)
  2415. // makes the spelling decision context-independent.
  2416. const { runtime } = await setup()
  2417. const result = await runtime.run({
  2418. program: [
  2419. 'from decimal import getcontext',
  2420. 'getcontext().prec = 2',
  2421. 'getcontext().traps[__import__("decimal").Inexact] = True',
  2422. 'return 1.2345678901234567',
  2423. ].join('\n'),
  2424. bindings: [],
  2425. })
  2426. expect(result.error).toBeUndefined()
  2427. expect(result.value).toBe(1.2345678901234567)
  2428. }, 15_000)
  2429. it('bounds an over-cap exception-group nesting on the copy', async () => {
  2430. // Exception groups link through `exceptions`, not the cause/context
  2431. // dunders, so the cap has to count that edge too — otherwise a deeply
  2432. // nested group walks past the bound the marker claims to enforce.
  2433. const { runtime } = await setup({ maxValueBytes: 1024 * 1024, maxWallMs: 15_000 })
  2434. const result = await runtime.run({
  2435. program: [
  2436. 'import sys',
  2437. // ExceptionGroup is a 3.11+ builtin; on 3.10 the NameError is the
  2438. // failure mode being probed, so skip to keep the assertion meaningful.
  2439. 'if sys.version_info < (3, 11):',
  2440. ' raise ValueError("skip-old <model>")',
  2441. 'group = ValueError("leaf")',
  2442. 'for i in range(150):',
  2443. ' group = ExceptionGroup(f"g{i}", [group])',
  2444. 'raise group',
  2445. ].join('\n'),
  2446. bindings: [],
  2447. })
  2448. expect(result.error?.kind).toBe('exception')
  2449. // The version guard skips on Python < 3.11 (ExceptionGroup is a 3.11+
  2450. // builtin) with a distinct message; the truncation assertion applies on
  2451. // 3.11+ where the group nesting is what is being probed.
  2452. expect(result.error?.message).toMatch(/exception chain truncated at 100 links|skip-old/)
  2453. }, 20_000)
  2454. it('filters every bootstrap frame from the traceback of an uncaught binding rejection', async () => {
  2455. // A rejection re-raised by the bootstrap's dispatch adds bootstrap frames
  2456. // AFTER the model's own; only <model> frames may reach model-visible,
  2457. // durable output — a bootstrap.py path would leak host absolutes and make
  2458. // transcripts machine-dependent.
  2459. const { runtime } = await setup()
  2460. const result = await runtime.run({
  2461. program: 'await tools.boom({})',
  2462. bindings: tools({ boom: async () => { throw new Error('exploded') } }),
  2463. })
  2464. expect(result.error?.kind).toBe('exception')
  2465. expect(result.error?.message).toContain('exploded')
  2466. expect(result.error?.message).toContain('<model>')
  2467. expect(result.error?.message).not.toContain('bootstrap.py')
  2468. })
  2469. it('renders a non-Error thrown value from a host binding as its String form', async () => {
  2470. const { runtime } = await setup()
  2471. const result = await runtime.run({
  2472. program: [
  2473. 'caught = ""',
  2474. 'try:',
  2475. ' await tools.failRaw({})',
  2476. 'except RuntimeError as e:',
  2477. ' caught = str(e)',
  2478. 'return caught',
  2479. ].join('\n'),
  2480. bindings: tools({
  2481. failRaw: async () => { throw 'raw-nope' },
  2482. }),
  2483. })
  2484. expect(result.error).toBeUndefined()
  2485. expect(result.value).toContain('raw-nope')
  2486. })
  2487. it('reassembles a frame split across writes behind a completed one', async () => {
  2488. // One os.write carrying "<frame>\n<partial...>" leaves a non-empty
  2489. // residual after the newline loop; the tail must survive until its own
  2490. // newline arrives and then parse as a normal frame.
  2491. const { runtime } = await setup()
  2492. const result = await runtime.run({
  2493. program: [
  2494. 'import os, json',
  2495. 'head = json.dumps({"type":"log","text":"first"}).encode()',
  2496. 'tail = json.dumps({"type":"log","text":"second"}).encode()',
  2497. 'import time',
  2498. 'os.write(3, head + b"\\n" + tail[:5])',
  2499. 'time.sleep(0.2)',
  2500. 'os.write(3, tail[5:] + b"\\n")',
  2501. 'return "ok"',
  2502. ].join('\n'),
  2503. bindings: [],
  2504. })
  2505. expect(result.error).toBeUndefined()
  2506. expect(result.value).toBe('ok')
  2507. expect(result.logs).toContain('first')
  2508. expect(result.logs).toContain('second')
  2509. })
  2510. it('raises the declared errorClass with the member name on rejection', async () => {
  2511. // PTC mode declares { name: ToolCallError, memberNameProperty: toolName };
  2512. // a host rejection must surface as that class, carrying the failed tool.
  2513. const { runtime } = await setup()
  2514. const result = await runtime.run({
  2515. program: [
  2516. 'caught = ""',
  2517. 'try:',
  2518. ' await tools.fail({})',
  2519. 'except ToolCallError as e:',
  2520. ' caught = f"{type(e).__name__}:{e.toolName}:{e}"',
  2521. 'return caught',
  2522. ].join('\n'),
  2523. bindings: [{
  2524. global: 'tools',
  2525. functions: { fail: async () => { throw new Error('typed-nope') } },
  2526. errorClass: { name: 'ToolCallError', memberNameProperty: 'toolName' },
  2527. }],
  2528. })
  2529. expect(result.error).toBeUndefined()
  2530. expect(result.value).toBe('ToolCallError:fail:typed-nope')
  2531. })
  2532. it('keeps the declared error class catching when Exception and setattr are rebound', async () => {
  2533. // _make_error_class's minted __init__ def-time captures Exception and
  2534. // setattr, so a program rebinding __main__.Exception/__main__.setattr
  2535. // cannot break the rejection constructor: `except ToolCallError` must still
  2536. // catch and read the member property.
  2537. const { runtime } = await setup()
  2538. const result = await runtime.run({
  2539. program: [
  2540. 'import __main__',
  2541. 'def boom(*a, **k):',
  2542. ' raise RuntimeError("hijacked")',
  2543. '__main__.Exception = boom',
  2544. '__main__.setattr = boom',
  2545. 'caught = ""',
  2546. 'try:',
  2547. ' await tools.fail({})',
  2548. 'except ToolCallError as e:',
  2549. ' caught = f"{type(e).__name__}:{e.toolName}"',
  2550. 'return caught',
  2551. ].join('\n'),
  2552. bindings: [{
  2553. global: 'tools',
  2554. functions: { fail: async () => { throw new Error('typed-nope') } },
  2555. errorClass: { name: 'ToolCallError', memberNameProperty: 'toolName' },
  2556. }],
  2557. })
  2558. expect(result.error).toBeUndefined()
  2559. expect(result.value).toBe('ToolCallError:fail')
  2560. }, 15_000)
  2561. it('runs when errorClass metadata is exposed through one-read getters', async () => {
  2562. // Validation reads errorClass.name and errorClass.memberNameProperty, and
  2563. // the ORIGINAL object used to ride along to the boot frame, whose
  2564. // JSON.stringify re-read it after validation: a getter that throws or
  2565. // changes on a second read turned the seam-misuse rejection into a
  2566. // worker-exit (or injected a different name than validation approved).
  2567. // The snapshot reads each field exactly once into a plain copy, so a
  2568. // getter that only tolerates one read must boot and run cleanly.
  2569. let nameReads = 0
  2570. let memberReads = 0
  2571. const errorClass = {
  2572. get name(): string {
  2573. nameReads += 1
  2574. if (nameReads > 1) throw new Error(`errorClass.name read ${nameReads} times`)
  2575. return 'ToolCallError'
  2576. },
  2577. get memberNameProperty(): string {
  2578. memberReads += 1
  2579. if (memberReads > 1) throw new Error(`errorClass.memberNameProperty read ${memberReads} times`)
  2580. return 'toolName'
  2581. },
  2582. }
  2583. const { runtime } = await setup()
  2584. const result = await runtime.run({
  2585. program: 'return "ok"',
  2586. bindings: [{ global: 'tools', functions: {}, errorClass }],
  2587. })
  2588. expect(result.error).toBeUndefined()
  2589. expect(result.value).toBe('ok')
  2590. expect(nameReads).toBe(1)
  2591. expect(memberReads).toBe(1)
  2592. }, 15_000)
  2593. it('runs when the binding global is exposed through a one-read getter', async () => {
  2594. // Validation reads namespace.global several times (identifier check, map
  2595. // key, claim, boot frame), and the map key came from a fresh read each
  2596. // time: a getter returning a different name on a later read injected a
  2597. // global validation never approved, and the program referencing the
  2598. // approved name died with NameError. Snapshotting reads it exactly once,
  2599. // so the child must receive the name the program was written against.
  2600. let globalReads = 0
  2601. const namespace = {
  2602. get global(): string {
  2603. globalReads += 1
  2604. return globalReads === 1 ? 'tools' : 'evil'
  2605. },
  2606. functions: { echo: async (args: unknown) => args as CodeJsonValue },
  2607. }
  2608. const { runtime } = await setup()
  2609. const result = await runtime.run({
  2610. program: 'return await tools.echo(41)',
  2611. bindings: [namespace],
  2612. })
  2613. expect(result.error).toBeUndefined()
  2614. expect(result.value).toBe(41)
  2615. expect(globalReads).toBe(1)
  2616. }, 15_000)
  2617. it('rejects an errorClass name colliding with its namespace global at the seam', async () => {
  2618. const { runtime } = await setup()
  2619. await expect(runtime.run({
  2620. program: 'return 1',
  2621. bindings: [{ global: 'tools', functions: {}, errorClass: { name: 'tools', memberNameProperty: 'toolName' } }],
  2622. })).rejects.toThrow(/collides with another injected global/)
  2623. })
  2624. it('rejects a namespace global colliding with a runtime-owned name at the seam', async () => {
  2625. // `__dsh_main__` passes the identifier check, but exec()ing the generated
  2626. // wrapper would silently overwrite the binding after injection. `console`
  2627. // is the WORKER backend's slot — refused here too so a namespace list
  2628. // valid on one backend is valid on all.
  2629. const { runtime } = await setup()
  2630. // `__debug__` is refused for a different reason than a collision: CPython
  2631. // compiles a bare `__debug__` reference to the constant True and refuses to
  2632. // assign the name at compile time, so an injected global under it is
  2633. // unreachable from the program — accepted by the seam, unusable here.
  2634. for (const global of ['__dsh_main__', 'console', '__debug__']) {
  2635. await expect(runtime.run({
  2636. program: 'x = 1',
  2637. bindings: [{ global, functions: {} }],
  2638. })).rejects.toThrow(/collides with a runtime-owned global/)
  2639. }
  2640. })
  2641. it('accepts a non-identifier memberNameProperty and rejects only an empty one', async () => {
  2642. // The seam permits any non-empty own property except the reserved
  2643. // members; Python setattr/getattr carry exotic names like `tool-name`,
  2644. // and the worker backend accepts them, so this backend must too.
  2645. const { runtime } = await setup()
  2646. const result = await runtime.run({
  2647. program: [
  2648. 'try:',
  2649. ' await tools.boom({})',
  2650. 'except ToolCallError as e:',
  2651. ' return getattr(e, "tool-name")',
  2652. ].join('\n'),
  2653. bindings: [{
  2654. global: 'tools',
  2655. functions: { boom: async () => { throw new Error('nope') } },
  2656. errorClass: { name: 'ToolCallError', memberNameProperty: 'tool-name' },
  2657. }],
  2658. })
  2659. expect(result.error).toBeUndefined()
  2660. expect(result.value).toBe('boom')
  2661. await expect(runtime.run({
  2662. program: 'return 1',
  2663. bindings: [{ global: 'tools', functions: {}, errorClass: { name: 'ToolCallError', memberNameProperty: '' } }],
  2664. })).rejects.toThrow(/memberNameProperty must be a non-empty attribute name/)
  2665. })
  2666. it('resolves a basename pythonBin to an absolute path (runs a real program)', async () => {
  2667. // A bare `python3` basename must resolve against PATH and actually launch
  2668. // under the empty-env spawn — exercises the accessSync success branch.
  2669. const { runtime } = await setup({ pythonBin: 'python3' })
  2670. const result = await runtime.run({ program: 'return 7', bindings: [] })
  2671. expect(result.error).toBeUndefined()
  2672. expect(result.value).toBe(7)
  2673. })
  2674. it('rejects at load a basename pythonBin with no PATH match', async () => {
  2675. // resolvePythonBin turns a basename into an absolute path before the
  2676. // empty-env spawn; a basename with no PATH match must fail at load (like an
  2677. // empty or NUL pythonBin) rather than silently falling to execvp's
  2678. // platform default PATH and starting a system interpreter the caller never
  2679. // asked for.
  2680. const ctx = new Context()
  2681. await expect(ctx.plugin(PythonCodeRuntime, { pythonBin: 'definitely-no-such-python-xyz' }))
  2682. .rejects.toThrow(/does not resolve on PATH/)
  2683. })
  2684. it('rejects a memberNameProperty naming a constrained BaseException attribute', async () => {
  2685. // `__dict__`/`__class__` are constrained descriptors alongside
  2686. // `__traceback__` — setattr of a string raises TypeError while
  2687. // constructing the rejection — so every dunder is refused at the seam.
  2688. const { runtime } = await setup()
  2689. // name/message/stack are the seam's own exclusions (CodeBindingErrorClass
  2690. // forbids replacing them; the worker backend rejects them identically).
  2691. for (const member of ['__traceback__', '__dict__', '__class__', 'args', 'name', 'message', 'stack']) {
  2692. await expect(runtime.run({
  2693. program: 'return 1',
  2694. bindings: [{ global: 'tools', functions: {}, errorClass: { name: 'ToolCallError', memberNameProperty: member } }],
  2695. })).rejects.toThrow(/reserved error member/)
  2696. }
  2697. })
  2698. it('rejects a lossy binding resolution (NaN) instead of coercing it to null', async () => {
  2699. // JSON.stringify would turn NaN into null and drop undefined fields; the
  2700. // seam requires a descriptive rejection so data cannot silently corrupt.
  2701. const { runtime } = await setup()
  2702. const result = await runtime.run({
  2703. program: [
  2704. 'caught = ""',
  2705. 'try:',
  2706. ' await tools.bad({})',
  2707. 'except RuntimeError as e:',
  2708. ' caught = str(e)',
  2709. 'return caught',
  2710. ].join('\n'),
  2711. bindings: tools({ bad: async () => Number.NaN }),
  2712. })
  2713. expect(result.error).toBeUndefined()
  2714. expect(result.value).toContain('lossless JSON')
  2715. })
  2716. it('contains a forged pathological done value without crashing the host', async () => {
  2717. // A ~20k-deep nested array forged onto fd 3 would overflow a recursive
  2718. // JSON.stringify; the host's iterative encoder measures it stack-safely
  2719. // and fails it deterministically on the byte budget (40 kB > 32 KiB).
  2720. const { runtime } = await setup()
  2721. const result = await runtime.run({
  2722. program: [
  2723. 'import os, json',
  2724. 'depth = 20000',
  2725. 'payload = "[" * depth + "]" * depth',
  2726. 'os.write(3, b\'{"type":"done","value":\' + payload.encode() + b\'}\\n\')',
  2727. 'import time',
  2728. 'time.sleep(5)',
  2729. ].join('\n'),
  2730. bindings: [],
  2731. })
  2732. expect(result.value).toBeUndefined()
  2733. expect(result.error?.kind).toBe('output-limit')
  2734. })
  2735. it('preserves a deeply nested completion value below the byte budget', async () => {
  2736. // CodeJsonValue has no depth limit: a 10000-deep nested list is only
  2737. // ~20 kB — under maxValueBytes — and must cross intact. That depth
  2738. // overflows BOTH recursive serializers the pipeline used to rely on
  2739. // (CPython's json.dumps recursion limit ~1000s, V8's JSON.stringify), so
  2740. // it proves the child-side _encode_json_plain and the host-side
  2741. // encodeJsonPlain together. The host JSON.parse of the frame is iterative
  2742. // in V8 for arrays, so only the two encoders were at risk.
  2743. const { runtime } = await setup()
  2744. const result = await runtime.run({
  2745. program: [
  2746. 'v = None',
  2747. 'for _ in range(10000):',
  2748. ' v = [v]',
  2749. 'return v',
  2750. ].join('\n'),
  2751. bindings: [],
  2752. })
  2753. expect(result.error).toBeUndefined()
  2754. // Walk down iteratively (a recursive toEqual would itself overflow).
  2755. let depth = 0
  2756. let cursor: unknown = result.value
  2757. while (Array.isArray(cursor)) {
  2758. expect(cursor).toHaveLength(1)
  2759. cursor = cursor[0]
  2760. depth++
  2761. }
  2762. expect(depth).toBe(10000)
  2763. expect(cursor).toBeNull()
  2764. })
  2765. it('bridges a deeply nested binding resolution back into the program stack-safely', async () => {
  2766. // A binding resolution has no seam-level depth or byte cap; neither the
  2767. // host's reply serialization nor the CHILD's reply decode may die on
  2768. // recursion (json.loads raises RecursionError ~10k levels deep; the
  2769. // bootstrap decodes frames iteratively). 12000 levels sits past that
  2770. // limit while staying tiny in bytes.
  2771. const { runtime } = await setup()
  2772. const deep = ((): unknown => {
  2773. let v: unknown = null
  2774. for (let i = 0; i < 12000; i++) v = [v]
  2775. return v
  2776. })()
  2777. const result = await runtime.run({
  2778. program: [
  2779. 'v = await tools.deep({})',
  2780. 'depth = 0',
  2781. 'while isinstance(v, list):',
  2782. ' v = v[0]',
  2783. ' depth += 1',
  2784. 'return depth',
  2785. ].join('\n'),
  2786. bindings: tools({ deep: async () => deep as never }),
  2787. })
  2788. expect(result.error).toBeUndefined()
  2789. expect(result.value).toBe(12000)
  2790. })
  2791. it('rejects a reserved errorClass name at the seam', async () => {
  2792. const { runtime } = await setup()
  2793. await expect(runtime.run({
  2794. program: 'return 1',
  2795. bindings: [{
  2796. global: 'tools',
  2797. functions: {},
  2798. errorClass: { name: 'class', memberNameProperty: 'toolName' },
  2799. }],
  2800. })).rejects.toThrow(/errorClass.name "class" is not a usable Python identifier/)
  2801. })
  2802. it('routes a declared inherited-attribute name through the bridge via subscript', async () => {
  2803. // __class__ resolves on `object` before any fallback hook; the proxy's
  2804. // __getattribute__ intercepts declared names first, and subscript access
  2805. // is the SDK-advertised route for underscore names.
  2806. const { runtime } = await setup()
  2807. const seen: string[] = []
  2808. const result = await runtime.run({
  2809. program: [
  2810. 'a = await tools["__class__"]({"via": "subscript"})',
  2811. 'b = await tools.__class__({"via": "dot"})',
  2812. 'return [a, b]',
  2813. ].join('\n'),
  2814. bindings: tools({
  2815. '__class__': async () => { seen.push('called'); return 'bridged' },
  2816. }),
  2817. })
  2818. expect(result.error).toBeUndefined()
  2819. expect(result.value).toEqual(['bridged', 'bridged'])
  2820. expect(seen).toEqual(['called', 'called'])
  2821. })
  2822. it('rejects NaN binding arguments immediately instead of hanging', async () => {
  2823. // Default json.dumps would emit a non-standard NaN token that the host
  2824. // JSON.parse drops silently, hanging the call until the wall clock;
  2825. // allow_nan=False raises in-program right away.
  2826. const { runtime } = await setup({ maxWallMs: 8000 })
  2827. const start = Date.now()
  2828. const result = await runtime.run({
  2829. program: [
  2830. 'caught = ""',
  2831. 'try:',
  2832. ' await tools.echo({"x": float("nan")})',
  2833. 'except RuntimeError as e:',
  2834. ' caught = str(e)',
  2835. 'return caught',
  2836. ].join('\n'),
  2837. bindings: tools({ echo: async args => args as CodeJsonValue }),
  2838. })
  2839. expect(result.error).toBeUndefined()
  2840. expect(result.value).toContain('lossless JSON')
  2841. expect(Date.now() - start).toBeLessThan(5000)
  2842. })
  2843. it('carries large binding arguments well past maxValueBytes', async () => {
  2844. // Binding traffic has no seam byte cap: a call frame far larger than the
  2845. // completion budget must reach the host intact (the fd-3 ceiling is a
  2846. // fixed memory-safety bound, not an output budget).
  2847. const maxValueBytes = 4096
  2848. const { runtime } = await setup({ maxValueBytes })
  2849. let receivedLength = 0
  2850. const result = await runtime.run({
  2851. program: [
  2852. `big = "B" * ${maxValueBytes * 50}`,
  2853. 'r = await tools.measure({"payload": big})',
  2854. 'return r',
  2855. ].join('\n'),
  2856. bindings: tools({
  2857. measure: async (args) => {
  2858. receivedLength = ((args as { payload: string }).payload).length
  2859. return receivedLength
  2860. },
  2861. }),
  2862. })
  2863. expect(result.error).toBeUndefined()
  2864. expect(receivedLength).toBe(maxValueBytes * 50)
  2865. expect(result.value).toBe(maxValueBytes * 50)
  2866. })
  2867. it('rejects an unknown binding name inside the program with a matching error', async () => {
  2868. const { runtime } = await setup()
  2869. const result = await runtime.run({
  2870. program: [
  2871. 'caught = ""',
  2872. 'try:',
  2873. ' await tools.nope({})',
  2874. 'except (AttributeError, RuntimeError) as e:',
  2875. ' caught = str(e)',
  2876. 'return caught',
  2877. ].join('\n'),
  2878. bindings: tools({ known: async () => 'ok' }),
  2879. })
  2880. expect(result.error).toBeUndefined()
  2881. expect(result.value).toContain('nope')
  2882. })
  2883. it('bounds an unknown-binding diagnostic built from a forged call frame', async () => {
  2884. // `call.global` and `call.name` carry no byte cap of their own, only the
  2885. // 64 MiB fd-3 frame parse cap, and the reply interpolated them raw: one copy
  2886. // into the template result, one into the `JSON.stringify` escape, one into
  2887. // the `encodeJsonPlain` frame, one into the pipe write. Slicing each field
  2888. // to `maxValueBytes` code units first makes an 8 MiB forged name a
  2889. // 128-byte reply. The observable effect is the reply the child then has to
  2890. // READ: its fd-3 reader is unbuffered, so `readline` consumes an oversized
  2891. // reply one `read(2)` per byte and the run's own legitimate call never gets
  2892. // answered — measured under a 60 s ceiling, the 8 MiB case timed out and a
  2893. // 64 MiB case cost the host 509.9 MiB of heap against 120.3 MiB with the
  2894. // slices in place. The child's address space stays generous enough to BUILD
  2895. // the forgery, which is not what is under test.
  2896. const { runtime } = await setup({ maxValueBytes: 128, addressSpaceMb: 1024, maxWallMs: 20_000 })
  2897. const result = await runtime.run({
  2898. program: [
  2899. 'import os',
  2900. 'frame = b\'{"type":"call","id":9001,"global":"tools","name":"\' + b"n" * (8 * 1024 * 1024) + b\'","args":{}}\\n\'',
  2901. // One os.write returns short past the pipe buffer, and a partial frame
  2902. // would glue itself to the next one and be dropped as malformed, so the
  2903. // forgery goes out through a drain loop.
  2904. 'view = memoryview(frame)',
  2905. 'while view:',
  2906. ' view = view[os.write(3, view):]',
  2907. // A legitimate call after the forgery: its reply can only arrive once
  2908. // the child has read past whatever the forged frame was answered with.
  2909. 'await tools.known({})',
  2910. 'return "settled"',
  2911. ].join('\n'),
  2912. bindings: tools({ known: async () => 'ok' }),
  2913. })
  2914. expect(result.error).toBeUndefined()
  2915. expect(result.value).toBe('settled')
  2916. }, 40_000)
  2917. it('bridges a binding call reached via subscript access (tools["name"])', async () => {
  2918. // The SDK tells the model `await tools["my-tool"](args)` works for exotic
  2919. // names; the proxy's __getitem__ must route it through the bridge.
  2920. const { runtime } = await setup()
  2921. const result = await runtime.run({
  2922. program: [
  2923. 'r = await tools["my-tool"]({"n": 7})',
  2924. 'return r',
  2925. ].join('\n'),
  2926. bindings: tools({ 'my-tool': async args => ({ got: args as CodeJsonValue }) }),
  2927. })
  2928. expect(result.error).toBeUndefined()
  2929. expect(result.value).toEqual({ got: { n: 7 } })
  2930. })
  2931. it('raises KeyError for an undeclared subscript name', async () => {
  2932. const { runtime } = await setup()
  2933. const result = await runtime.run({
  2934. program: [
  2935. 'caught = ""',
  2936. 'try:',
  2937. ' await tools["absent"]({})',
  2938. 'except KeyError as e:',
  2939. ' caught = str(e)',
  2940. 'return caught',
  2941. ].join('\n'),
  2942. bindings: tools({ known: async () => 'ok' }),
  2943. })
  2944. expect(result.error).toBeUndefined()
  2945. expect(result.value).toContain('absent')
  2946. })
  2947. })
  2948. describe('PythonCodeRuntime — budgets, termination, disposal', () => {
  2949. it('kills a wall-clock runaway program via SIGTERM/SIGKILL and reports timeout', async () => {
  2950. const { runtime } = await setup({ maxWallMs: 500, graceMs: 200 })
  2951. const start = Date.now()
  2952. const result = await runtime.run({
  2953. program: 'import time\nwhile True: time.sleep(1)',
  2954. bindings: [],
  2955. })
  2956. const elapsed = Date.now() - start
  2957. // The wall timer may fire first or the exit-after-signal may resolve; both are ok.
  2958. expect(['timeout', 'worker-exit']).toContain(result.error?.kind)
  2959. // We got somewhere in the neighborhood of maxWallMs, not the underlying `sleep(1)`.
  2960. expect(elapsed).toBeLessThan(2000)
  2961. }, 5000)
  2962. it('aborts a run when the outer signal fires mid-flight', async () => {
  2963. const { runtime } = await setup({ maxWallMs: 10_000 })
  2964. const controller = new AbortController()
  2965. const settled: Promise<CodeRunResult> = runtime.run({
  2966. program: 'import time\nwhile True: time.sleep(0.1)',
  2967. bindings: [],
  2968. signal: controller.signal,
  2969. })
  2970. setTimeout(() => { controller.abort('outer-abort') }, 200)
  2971. const result = await settled
  2972. expect(['abort', 'worker-exit']).toContain(result.error?.kind)
  2973. }, 5000)
  2974. it('settles the run when a mid-flight abort reason cannot be converted', async () => {
  2975. // The listener converted the reason before calling `finish()`, so a hostile
  2976. // reason threw from inside an `AbortSignal` listener. Node reports that as an
  2977. // uncaught exception — it can terminate the host — and `finish()` never ran,
  2978. // so the run stayed live until the wall ceiling and misreported as `timeout`
  2979. // (observed) instead of the caller's cancellation. `maxWallMs` is short so
  2980. // that misreport is a fast assertion failure rather than a suite timeout.
  2981. const uncaught: unknown[] = []
  2982. const record = (error: unknown): void => { uncaught.push(error) }
  2983. process.on('uncaughtException', record)
  2984. try {
  2985. const { runtime } = await setup({ maxWallMs: 4_000, graceMs: 200 })
  2986. const controller = new AbortController()
  2987. const settled: Promise<CodeRunResult> = runtime.run({
  2988. program: 'import time\nwhile True: time.sleep(0.1)',
  2989. bindings: [],
  2990. signal: controller.signal,
  2991. })
  2992. setTimeout(() => {
  2993. controller.abort({ [Symbol.toPrimitive]() { throw new Error('reason blew up') } })
  2994. }, 200)
  2995. const result = await settled
  2996. expect(result.error?.kind).toBe('abort')
  2997. expect(result.error?.message).toBe('<unrenderable rejection value>')
  2998. expect(uncaught).toEqual([])
  2999. } finally {
  3000. process.off('uncaughtException', record)
  3001. }
  3002. }, 15_000)
  3003. it('disposes to quiescence: an in-flight run resolves as abort and the child exits', async () => {
  3004. const { fiber, runtime } = await setup({ maxWallMs: 10_000 })
  3005. const pending = runtime.run({
  3006. program: 'import time\nwhile True: time.sleep(0.1)',
  3007. bindings: [],
  3008. })
  3009. // Give the process time to spawn and start running.
  3010. await new Promise(resolve => setTimeout(resolve, 200))
  3011. await fiber.dispose()
  3012. const result = await pending
  3013. expect(['abort', 'worker-exit']).toContain(result.error?.kind)
  3014. }, 5000)
  3015. it('reports an interpreter removed after load as worker-exit', async () => {
  3016. const dir = await mkdtemp(join(tmpdir(), 'dsh-python-removed-'))
  3017. const pythonBin = join(dir, 'python3')
  3018. await writeFile(pythonBin, `#!/bin/sh\nexec "${PYABS}" "$@"\n`, { mode: 0o755 })
  3019. const { runtime, fiber } = await setup({ pythonBin, maxWallMs: 3000 })
  3020. rmSync(pythonBin)
  3021. try {
  3022. const result = await runtime.run({ program: 'return 1', bindings: [] })
  3023. expect(result.error?.kind).toBe('worker-exit')
  3024. } finally {
  3025. await fiber.dispose()
  3026. rmSync(dir, { recursive: true, force: true })
  3027. }
  3028. }, 8000)
  3029. it('applies the strictest of the configured and inherited resource limits', async () => {
  3030. // This case used to drive the bootstrap's `applying resource limits failed`
  3031. // handler with `cpuSeconds: 2 ** 63`, asserting that a cap the child cannot
  3032. // apply fails the run rather than running it uncapped. That premise no longer
  3033. // holds, for two independent reasons, so the test now pins what is actually
  3034. // guaranteed instead of a path no admissible input reaches.
  3035. //
  3036. // First, `2 ** 63` is not a safe integer, so it is now rejected at LOAD as a
  3037. // configuration error — it can never reach the child at all. Second, even the
  3038. // largest admissible values are applied successfully, because `_clamped`
  3039. // bounds every requested pair by the inherited hard limit: an unprivileged
  3040. // process may lower a hard limit but never raise one, so the child keeps the
  3041. // stricter of the two rather than asking for something `setrlimit` refuses.
  3042. // The failure handler remains as a substrate guard (a platform whose kernel
  3043. // refuses the call for its own reasons), but it is no longer reachable from
  3044. // configuration, and a test that pretends otherwise documents a contract the
  3045. // code does not have.
  3046. //
  3047. // What is observable: a very large cap still yields a working run, and the
  3048. // containment it promises is met by the inherited ceiling.
  3049. const { runtime } = await setup({ cpuSeconds: Number.MAX_SAFE_INTEGER - 1, maxWallMs: 10_000 })
  3050. const result = await runtime.run({ program: 'return 1', bindings: [] })
  3051. expect(result.error).toBeUndefined()
  3052. expect(result.value).toBe(1)
  3053. }, 20_000)
  3054. it('settles as worker-exit when the child exits before sending done (no hang)', async () => {
  3055. // Regression: settlement must key off `close` (process reaped AND stdio
  3056. // drained), not `exit`. With `exit`, finish() re-armed a second exit
  3057. // listener that never fired — run() hung forever whenever the exit event
  3058. // beat the final fd-3 data (deterministic on macOS, a lost race elsewhere).
  3059. const { runtime } = await setup({ maxWallMs: 10_000 })
  3060. const result = await runtime.run({
  3061. program: 'import os\nos._exit(7)',
  3062. bindings: [],
  3063. })
  3064. expect(result.error?.kind).toBe('worker-exit')
  3065. expect(result.error?.message).toContain('code=7')
  3066. }, 5000)
  3067. it('classifies RLIMIT_CPU soft-limit expiry (SIGXCPU) as a timeout', async () => {
  3068. // A CPU hot loop burns the soft limit; the kernel delivers SIGXCPU, whose
  3069. // close signal the host maps to `timeout`. macOS re-delivers SIGXCPU
  3070. // differently, so we assert only kind/message here — CI's darwin leg
  3071. // validates real delivery. cpuSeconds must be an integer for setrlimit.
  3072. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 20_000 })
  3073. const result = await runtime.run({
  3074. program: 'while True: pass',
  3075. bindings: [],
  3076. })
  3077. expect(result.error?.kind).toBe('timeout')
  3078. expect(result.error?.message).toContain('CPU time exhausted')
  3079. }, 8000)
  3080. it('keeps an early self-inflicted SIGKILL a worker-exit, not a CPU timeout', async () => {
  3081. // The unsolicited-SIGKILL-as-timeout classification applies only when the
  3082. // CPU budget could have expired (wall time >= cpuSeconds). A SIGKILL
  3083. // seconds before that (cgroup OOM, an operator, os.kill) is substrate
  3084. // death and stays worker-exit per the orthogonal taxonomy.
  3085. const { runtime } = await setup({ cpuSeconds: 60, maxWallMs: 10_000 })
  3086. const result = await runtime.run({
  3087. program: [
  3088. 'import os, signal',
  3089. 'os.kill(os.getpid(), signal.SIGKILL)',
  3090. ].join('\n'),
  3091. bindings: [],
  3092. })
  3093. expect(result.error?.kind).toBe('worker-exit')
  3094. expect(result.error?.message).toContain('SIGKILL')
  3095. })
  3096. it('charges a forked descendant against the run CPU budget', async () => {
  3097. // RLIMIT_CPU is per-process and every child inherits a FRESH budget, so a
  3098. // program that shells out multiplies `cpuSeconds` by the number of
  3099. // descendants it starts. Measured before the aggregate meter existed: with
  3100. // cpuSeconds 1, two sequential busy children burned 2.0 CPU-seconds
  3101. // (RUSAGE_CHILDREN) and the run still returned a SUCCESS completion. The
  3102. // settle-time check meters RUSAGE_SELF + RUSAGE_CHILDREN and converts the
  3103. // overrun into the same SIGXCPU the untrapped soft limit sends.
  3104. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 30_000 })
  3105. const result = await runtime.run({
  3106. program: [
  3107. 'import subprocess, sys',
  3108. 'for _ in range(2):',
  3109. ' subprocess.run([sys.executable, "-c", "import time\\nt=time.time()\\nwhile time.time()-t<1.2: pass"])',
  3110. 'return "escaped the cpu budget"',
  3111. ].join('\n'),
  3112. bindings: [],
  3113. })
  3114. // Darwin's SIGXCPU re-delivery differs, so accept either terminal
  3115. // classification; what must NOT happen is the completion crossing.
  3116. expect(['timeout', 'worker-exit']).toContain(result.error?.kind)
  3117. expect(result.value).toBeUndefined()
  3118. }, 40_000)
  3119. it('does not charge wall time or a cheap descendant against the CPU budget', async () => {
  3120. // The meter is CPU, not wall clock, and it must not fire on a child that
  3121. // burns almost nothing: a sleeping program and a trivial subprocess both
  3122. // have to complete normally, or the check would reject every program that
  3123. // shells out.
  3124. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 30_000 })
  3125. const slept = await runtime.run({
  3126. program: 'import time\ntime.sleep(1.5)\nreturn "slept"',
  3127. bindings: [],
  3128. })
  3129. expect(slept.error).toBeUndefined()
  3130. expect(slept.value).toBe('slept')
  3131. const cheap = await runtime.run({
  3132. program: [
  3133. 'import subprocess, sys',
  3134. 'subprocess.run([sys.executable, "-c", "pass"])',
  3135. 'return "cheap child"',
  3136. ].join('\n'),
  3137. bindings: [],
  3138. })
  3139. expect(cheap.error).toBeUndefined()
  3140. expect(cheap.value).toBe('cheap child')
  3141. }, 40_000)
  3142. it('spends no part of addressSpaceMb on bootstrap machinery', async () => {
  3143. // RLIMIT_AS counts RESERVED address space, so anything the bootstrap maps
  3144. // for its own accounting is subtracted from the program's `addressSpaceMb`.
  3145. // A sampling thread for the descendant-CPU meter cost 72 MiB here (an 8 MiB
  3146. // stack plus a 64 MiB glibc per-thread malloc arena reservation) and turned
  3147. // the 2-million-entry dict rejection below into a MemoryError under a
  3148. // 256 MiB cap on a slower runner. Assert the child's own mappings directly
  3149. // rather than inferring the budget from a near-cap allocation, so the bound
  3150. // is read from /proc instead of from how much headroom one machine happens
  3151. // to have; 48 MiB is well above the ~30 MiB a bare interpreter maps and
  3152. // well below the 102 MiB the thread produced. `addressSpaceMb` itself is
  3153. // skipped on darwin (the dyld shared cache makes any practical cap
  3154. // unsettable) and /proc/self/maps does not exist there, so the mapping
  3155. // assertion is Linux-only; the completion path is checked everywhere.
  3156. const { runtime } = await setup({ maxValueBytes: 4096, addressSpaceMb: 256 })
  3157. const mapped = await runtime.run({
  3158. program: [
  3159. 'import sys',
  3160. 'if sys.platform != "linux":',
  3161. ' return 0',
  3162. 'total = 0',
  3163. 'with open("/proc/self/maps") as handle:',
  3164. ' for line in handle:',
  3165. ' low, high = (int(part, 16) for part in line.split(" ", 1)[0].split("-"))',
  3166. ' total += high - low',
  3167. 'return total // (1024 * 1024)',
  3168. ].join('\n'),
  3169. bindings: [],
  3170. })
  3171. expect(mapped.error).toBeUndefined()
  3172. expect(mapped.value).toBeLessThan(48)
  3173. }, 20_000)
  3174. it('spends no part of addressSpaceMb on the reply pump, across a binding await', async () => {
  3175. // The test above measures BEFORE the program yields, so it could not see the
  3176. // reply pump's cost: `loop.run_in_executor(None, read_frame)` created the
  3177. // default executor's first thread on the first `await tools.*`, and that
  3178. // thread's 8 MiB stack plus a 64 MiB glibc per-thread malloc arena are
  3179. // charged to RLIMIT_AS while the limit is already in force — measured, the
  3180. // child went from 30.34 MiB to 102.39 MiB across one binding call. Under a
  3181. // small `addressSpaceMb` the thread cannot start and a legitimate call hangs
  3182. // to `maxWallMs`; under a larger one an allocation that should have fit dies
  3183. // as MemoryError. `loop.add_reader` watches the fd with no thread at all.
  3184. //
  3185. // Measuring both sides inside one run is what discriminates: a single
  3186. // after-the-fact number cannot separate the pump's cost from the
  3187. // interpreter's own footprint. Linux-only for the same reason as above.
  3188. const { runtime } = await setup({ addressSpaceMb: 256, maxWallMs: 20_000 })
  3189. const result = await runtime.run({
  3190. program: [
  3191. 'import sys',
  3192. 'def mapped():',
  3193. ' if sys.platform != "linux":',
  3194. ' return 0',
  3195. ' total = 0',
  3196. ' with open("/proc/self/maps") as handle:',
  3197. ' for line in handle:',
  3198. ' low, high = (int(part, 16) for part in line.split(" ", 1)[0].split("-"))',
  3199. ' total += high - low',
  3200. ' return total // (1024 * 1024)',
  3201. 'before = mapped()',
  3202. 'echoed = await tools.echo({"ping": True})',
  3203. 'return {"before": before, "after": mapped(), "echoed": echoed}',
  3204. ].join('\n'),
  3205. bindings: tools({ echo: async args => args as CodeJsonValue }),
  3206. })
  3207. expect(result.error).toBeUndefined()
  3208. const value = result.value as { before: number; after: number; echoed: unknown }
  3209. // The binding call really happened, so the pump really ran.
  3210. expect(value.echoed).toEqual({ ping: true })
  3211. // Awaiting a binding maps nothing extra. The 8 MiB allowance absorbs ordinary
  3212. // heap growth while staying far below the 72 MiB a pump thread cost.
  3213. expect(value.after - value.before).toBeLessThan(8)
  3214. }, 30_000)
  3215. it('still terminates a program that ignores SIGXCPU (hard-limit backstop)', async () => {
  3216. // A hot loop under SIG_IGN burns through the soft limit; the kernel's
  3217. // hard limit (cpuSeconds + 1) SIGKILLs it. Only a kernel-authoritative
  3218. // SIGXCPU close classifies as the CPU timeout — a bare SIGKILL is
  3219. // indistinguishable from a cgroup OOM kill, so it reports worker-exit
  3220. // (Darwin re-delivers SIGXCPU instead, where the wall clock settles it
  3221. // as timeout). Either way the run TERMINATES within the budget — the
  3222. // backstop holds even when the classification is the opaque one.
  3223. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 6_000 })
  3224. const result = await runtime.run({
  3225. program: [
  3226. 'import signal',
  3227. 'signal.signal(signal.SIGXCPU, signal.SIG_IGN)',
  3228. 'while True: pass',
  3229. ].join('\n'),
  3230. bindings: [],
  3231. })
  3232. expect(['timeout', 'worker-exit']).toContain(result.error?.kind)
  3233. }, 12_000)
  3234. it('enforces the CPU budget even when the program monkeypatches the enforcement primitives', async () => {
  3235. // The check uses import-time-captured references, so replacing
  3236. // resource.getrusage / signal.signal / os.kill on the modules cannot
  3237. // defang it: a trapping program that also swaps the callables and burns
  3238. // past the budget still dies by the authoritative SIGXCPU.
  3239. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 15_000 })
  3240. const result = await runtime.run({
  3241. program: [
  3242. 'import signal, os, resource, time',
  3243. 'signal.signal(signal.SIGXCPU, lambda *a: None)',
  3244. 'resource.getrusage = lambda *a: (_ for _ in ()).throw(RuntimeError("nope"))',
  3245. 'os.kill = lambda *a: None',
  3246. 'signal.signal = lambda *a: None',
  3247. 'deadline = time.process_time() + 1.05',
  3248. 'while time.process_time() < deadline: pass',
  3249. 'return "escaped"',
  3250. ].join('\n'),
  3251. bindings: [],
  3252. })
  3253. expect(result.error?.kind).toBe('timeout')
  3254. expect(result.value).toBeUndefined()
  3255. }, 15_000)
  3256. it('re-delivers SIGXCPU when a trapping program returns inside the soft-to-hard gap', async () => {
  3257. // A program can trap SIGXCPU and settle during the one-second gap; the
  3258. // bootstrap re-checks the kernel CPU meter (getrusage) after settlement
  3259. // and dies by SIGXCPU with the default disposition restored, so the host
  3260. // still classifies the exhausted budget as a timeout instead of success.
  3261. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 15_000 })
  3262. const result = await runtime.run({
  3263. program: [
  3264. 'import signal, time',
  3265. 'fired = []',
  3266. 'signal.signal(signal.SIGXCPU, lambda *a: fired.append(1))',
  3267. 'deadline = time.process_time() + 1.05',
  3268. 'while time.process_time() < deadline: pass',
  3269. 'return "escaped"',
  3270. ].join('\n'),
  3271. bindings: [],
  3272. })
  3273. expect(result.error?.kind).toBe('timeout')
  3274. expect(result.error?.message).toContain('CPU time exhausted')
  3275. expect(result.value).toBeUndefined()
  3276. }, 15_000)
  3277. it('enforces the CPU budget when the program rebinds the enforcer on __main__', async () => {
  3278. // The bootstrap IS `__main__`, so `import __main__` reaches its globals.
  3279. // The enforcement callable holds its primitives in closure cells (not
  3280. // module attributes) and `_run` reads the callable into a frame local
  3281. // before the program starts, so neither replacing the global nor swapping
  3282. // the module's captured names changes what runs after settlement.
  3283. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 15_000 })
  3284. const result = await runtime.run({
  3285. program: [
  3286. 'import signal, time, __main__',
  3287. 'signal.signal(signal.SIGXCPU, lambda *a: None)',
  3288. '__main__._DIE_IF_CPU_EXHAUSTED = lambda *_: None',
  3289. 'deadline = time.process_time() + 1.05',
  3290. 'while time.process_time() < deadline: pass',
  3291. 'return "escaped"',
  3292. ].join('\n'),
  3293. bindings: [],
  3294. })
  3295. expect(result.error?.kind).toBe('timeout')
  3296. expect(result.value).toBeUndefined()
  3297. }, 15_000)
  3298. it('bounds a program that defeats the post-check by writing its closure cell', async () => {
  3299. // The closure-cell capture raises the cost of defeating the post-check; it
  3300. // does NOT make it unreachable, and nothing in-process could: a cell is
  3301. // writable through `fn.__closure__[i].cell_contents`, and `sys._getframe`
  3302. // reads _run's frame locals. This program does exactly that — walks to
  3303. // _run's frame, takes the enforcement callable, and replaces its captured
  3304. // `getrusage` with one reporting zero CPU used — then burns past cpuSeconds
  3305. // with SIGXCPU trapped. The run must still fail, because the bound that
  3306. // model code cannot forge is outside the interpreter: the RLIMIT_CPU HARD
  3307. // limit at cpuSeconds + 1, whose SIGKILL admits no handler. No success is
  3308. // reportable either way.
  3309. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 20_000 })
  3310. const start = Date.now()
  3311. const result = await runtime.run({
  3312. program: [
  3313. 'import signal, sys, time',
  3314. 'signal.signal(signal.SIGXCPU, lambda *a: None)',
  3315. // Walk out of __dsh_main__ to _run's frame and take its local.
  3316. 'die = None',
  3317. 'depth = 1',
  3318. 'while depth < 12:',
  3319. ' frame = sys._getframe(depth)',
  3320. ' if "die_if_cpu_exhausted" in frame.f_locals:',
  3321. ' die = frame.f_locals["die_if_cpu_exhausted"]',
  3322. ' break',
  3323. ' depth += 1',
  3324. 'assert die is not None, "enforcer not reachable from the frame chain"',
  3325. 'class Zero:',
  3326. ' ru_utime = 0.0',
  3327. ' ru_stime = 0.0',
  3328. 'names = die.__code__.co_freevars',
  3329. 'die.__closure__[names.index("getrusage")].cell_contents = lambda *a: Zero()',
  3330. // Burn well past the soft limit into the hard limit's SIGKILL.
  3331. 'while True: pass',
  3332. ].join('\n'),
  3333. bindings: [],
  3334. })
  3335. // What holds on EVERY platform: the tampering bought no success. The run
  3336. // failed, carried no value, and the reported kind is one of the two
  3337. // kernel-level outcomes — never a completion.
  3338. expect(result.value).toBeUndefined()
  3339. expect(result.error?.kind === 'worker-exit' || result.error?.kind === 'timeout').toBe(true)
  3340. if (process.platform === 'linux') {
  3341. // Linux enforces the RLIMIT_CPU HARD limit at cpuSeconds + 1 promptly, so
  3342. // the CPU bound — not the 20 s wall ceiling — is what stops the program.
  3343. // Its SIGKILL is not SIGXCPU, so the orthogonal-failure taxonomy reports
  3344. // `worker-exit`: a bare SIGKILL is not evidence of CPU burn.
  3345. expect(result.error?.kind).toBe('worker-exit')
  3346. expect(Date.now() - start).toBeLessThan(15_000)
  3347. } else {
  3348. // Darwin does not deliver the hard limit's SIGKILL on the same schedule;
  3349. // observed on the macOS lane, a program that patches the post-check runs
  3350. // to the WALL ceiling instead. The CPU budget is therefore not the
  3351. // binding constraint against a tampering program there — the wall clock
  3352. // is. Asserted rather than skipped so the difference stays visible.
  3353. expect(result.error?.kind).toBe('timeout')
  3354. }
  3355. }, 30_000)
  3356. it('keeps a finished-but-not-closed run live so dispose awaits the child\'s death', async () => {
  3357. // finish() no longer drops the run from `live`; settle() (at close) does.
  3358. // A SIGTERM-trapping program with a small graceMs sits in the grace window
  3359. // after finish() fires — dispose() must not resolve until the SIGKILL
  3360. // backstop actually reaps the child. The program prints its pid (captured
  3361. // as a log even on abort); once dispose() resolves, that pid must be dead
  3362. // (process.kill(pid, 0) throws ESRCH).
  3363. const { fiber, runtime } = await setup({ maxWallMs: 10_000, graceMs: 400 })
  3364. // Deterministic readiness: the program reports its pid through a binding
  3365. // AFTER installing the trap, so dispose cannot race the spawn (a fixed
  3366. // sleep lost that race on slow CI runners — SIGTERM landed pre-trap).
  3367. let reportedPid!: (pid: number) => void
  3368. const trapReady = new Promise<number>((resolve) => { reportedPid = resolve })
  3369. const pending = runtime.run({
  3370. program: [
  3371. 'import signal, time, os',
  3372. 'signal.signal(signal.SIGTERM, lambda *a: None)',
  3373. 'await tools.ready({"pid": os.getpid()})',
  3374. 'while True: time.sleep(0.05)',
  3375. ].join('\n'),
  3376. bindings: tools({
  3377. ready: async (args) => {
  3378. reportedPid((args as { pid: number }).pid)
  3379. return 'ok'
  3380. },
  3381. }),
  3382. })
  3383. const pid = await trapReady
  3384. const start = Date.now()
  3385. await fiber.dispose()
  3386. const elapsed = Date.now() - start
  3387. const result = await pending
  3388. expect(['abort', 'worker-exit', 'timeout']).toContain(result.error?.kind)
  3389. // dispose() returned only after the grace window elapsed (the SIGTERM trap
  3390. // forces the SIGKILL backstop path), proving the run stayed live past finish().
  3391. expect(elapsed).toBeGreaterThanOrEqual(300)
  3392. expect(Number.isInteger(pid) && pid > 0).toBe(true)
  3393. // The child is fully reaped by the time dispose() resolved.
  3394. expect(() => process.kill(pid, 0)).toThrow(/ESRCH/)
  3395. }, 8000)
  3396. it('settles on the decided result even when a setsid-escaped orphan holds stdio open past close', async () => {
  3397. // `close` only fires once every inherited stdio stream drains. A descendant
  3398. // started with start_new_session=True escapes the child's process group, so
  3399. // the SIGTERM/SIGKILL aimed at that group never reaches it; if it inherited
  3400. // our stdout/stderr/fd 3 and outlives the run, `close` would never fire and
  3401. // run() would hang forever. The close-deadline backstop (graceMs + margin)
  3402. // must force settlement on the value the `done` frame already decided.
  3403. const { runtime } = await setup({ graceMs: 100 })
  3404. const start = Date.now()
  3405. const result = await runtime.run({
  3406. program: [
  3407. 'import subprocess, sys',
  3408. // Orphan in a fresh session, inheriting our stdout/stderr/fd 3, alive
  3409. // past the close-deadline so `close` cannot fire on its own. Its own
  3410. // 5 s self-exit is the leak ceiling AND the discriminator: it must stay
  3411. // ABOVE the < 4000 ms upper-bound assertion below, so if the deadline
  3412. // backstop failed to settle, settlement could only come from this
  3413. // self-exit at ~5 s and blow the bound — a sharper signal than the wall
  3414. // ceiling would give.
  3415. 'subprocess.Popen([sys.executable, "-c", "import time; time.sleep(5)"],',
  3416. ' start_new_session=True)',
  3417. 'return "escaped"',
  3418. ].join('\n'),
  3419. bindings: [],
  3420. })
  3421. const elapsed = Date.now() - start
  3422. // The done frame decided the value; the deadline settled it despite the
  3423. // orphan pinning the pipes open.
  3424. expect(result.error).toBeUndefined()
  3425. expect(result.value).toBe('escaped')
  3426. // Settlement waited for the backstop (graceMs + CLOSE_REAP_MARGIN_MS ≈ 2.1s),
  3427. // not the orphan's 5 s self-exit — proving the deadline, not a fallback, fired.
  3428. expect(elapsed).toBeGreaterThanOrEqual(1_500)
  3429. expect(elapsed).toBeLessThan(4_000)
  3430. }, 8000)
  3431. it('flushes a newline-free diagnostic when the closeDeadline forces settlement', async () => {
  3432. // A leader that writes an unterminated diagnostic via `os.write(1, ...)` and
  3433. // then exits, leaving a setsid orphan holding the pipes open, settles through
  3434. // the closeDeadline destroy() path — which fires no `end`. The residual must
  3435. // be flushed before destroy() drops it, or the diagnostic is lost from
  3436. // `logs`. The value is decided by the done frame; the diagnostic must survive.
  3437. const { runtime } = await setup({ graceMs: 100 })
  3438. const result = await runtime.run({
  3439. program: [
  3440. 'import os, subprocess, sys',
  3441. 'os.write(1, b"leader-diagnostic-no-newline")',
  3442. 'subprocess.Popen([sys.executable, "-c", "import time; time.sleep(5)"],',
  3443. ' start_new_session=True)',
  3444. 'return "escaped"',
  3445. ].join('\n'),
  3446. bindings: [],
  3447. })
  3448. expect(result.error).toBeUndefined()
  3449. expect(result.value).toBe('escaped')
  3450. expect(result.logs).toContain('leader-diagnostic-no-newline')
  3451. }, 8000)
  3452. it('closes the child stdin so a program read sees EOF instead of blocking', async () => {
  3453. // The host closes the child's stdin write handle immediately after spawn
  3454. // (the program is an async body that reads nothing from fd 0; a live pipe
  3455. // would hold a host-side handle open past the run). A program that DOES
  3456. // read fd 0 therefore sees EOF at once. Fail-before: with the handle left
  3457. // open and no data written, `sys.stdin.read()` blocks and the run would
  3458. // hang to maxWallMs as a timeout.
  3459. const { runtime } = await setup({ maxWallMs: 8_000 })
  3460. const result = await runtime.run({
  3461. program: [
  3462. 'import sys',
  3463. 'data = sys.stdin.read()',
  3464. 'return "read: " + repr(data)',
  3465. ].join('\n'),
  3466. bindings: [],
  3467. })
  3468. expect(result.error).toBeUndefined()
  3469. expect(result.value).toBe("read: ''")
  3470. }, 15_000)
  3471. it('keeps runtime type annotations as live classes, not PEP 563 strings, when the program reads them', async () => {
  3472. // bootstrap.py imports `from __future__ import annotations`; without
  3473. // dont_inherit=True on compile(), that PEP 563 flag leaks into the program's
  3474. // compiled code and stringifies its type annotations, changing the semantics
  3475. // of a legal program that reads `f.__annotations__` at runtime.
  3476. const { runtime } = await setup()
  3477. const result = await runtime.run({
  3478. program: [
  3479. 'def f(x: int) -> int:',
  3480. ' return x',
  3481. 'return f.__annotations__["x"].__name__',
  3482. ].join('\n'),
  3483. bindings: [],
  3484. })
  3485. expect(result.error).toBeUndefined()
  3486. expect(result.value).toBe('int')
  3487. }, 15_000)
  3488. it('reaps a same-group child that ignores SIGTERM and releases the pipes before close', async () => {
  3489. // The same-group counterpart to the setsid-orphan case above. A descendant
  3490. // left in the child's OWN process group (no setsid, so `kill(-pid)` reaches
  3491. // it) can ignore SIGTERM yet still release the inherited stdout/stderr/fd 3
  3492. // it does not hold — here by giving the Popen child DEVNULL streams and
  3493. // letting close_fds drop fd 3. The leader then writes `done` and exits, its
  3494. // `close` fires because the pipes drained, and settle() runs while that
  3495. // descendant is still alive. settle() then keeps a REF'd poll alive until the
  3496. // grace-window SIGKILL has emptied the whole process group, so the host cannot
  3497. // exit and reparent the survivor to init: no subprocess outlives the fiber.
  3498. //
  3499. // The descendant must have SIG_IGN installed BEFORE the host sends SIGTERM,
  3500. // or it dies from the default SIGTERM whether the fix is present or not — so
  3501. // it writes a readiness marker after trapping and the leader waits for that
  3502. // marker before returning. While alive it bumps a heartbeat file every 50 ms;
  3503. // the test asserts the heartbeat STOPS, which is what "no longer executing"
  3504. // means whether the killed descendant is reaped or lingers as a zombie (a
  3505. // SIGKILL'd process runs no more code either way). It sleeps 30 s as a safety
  3506. // net so a broken fix cannot leak it forever.
  3507. const handoff = await makeTempDir('dsh-samegroup-')
  3508. const readyMarker = join(handoff, 'ready')
  3509. const heartbeat = join(handoff, 'heartbeat')
  3510. const { runtime } = await setup({ maxWallMs: 10_000, graceMs: 300 })
  3511. const result = await runtime.run({
  3512. program: [
  3513. 'import subprocess, sys, os, time',
  3514. `marker = ${JSON.stringify(readyMarker)}`,
  3515. `heartbeat = ${JSON.stringify(heartbeat)}`,
  3516. // Same group (no start_new_session); ignores SIGTERM; holds none of the
  3517. // leader's pipes (DEVNULL std streams, close_fds drops fd 3). It writes
  3518. // the marker (argv[1]) only AFTER the trap is installed — so the leader
  3519. // cannot return, and the host cannot send SIGTERM, before it is ignored —
  3520. // then rewrites the heartbeat (argv[2]) every 50 ms for up to 30 s.
  3521. 'code = ("import signal, sys, time\\n"',
  3522. ' "signal.signal(signal.SIGTERM, signal.SIG_IGN)\\n"',
  3523. ' "open(sys.argv[1], \'w\').close()\\n"',
  3524. ' "end = time.time() + 30\\n"',
  3525. ' "while time.time() < end:\\n"',
  3526. ' " open(sys.argv[2], \'w\').close()\\n"',
  3527. ' " time.sleep(0.05)\\n")',
  3528. 'child = subprocess.Popen([sys.executable, "-c", code, marker, heartbeat],',
  3529. ' stdin=subprocess.DEVNULL,',
  3530. ' stdout=subprocess.DEVNULL,',
  3531. ' stderr=subprocess.DEVNULL)',
  3532. 'deadline = time.time() + 5',
  3533. 'while not os.path.exists(marker) and time.time() < deadline:',
  3534. ' time.sleep(0.02)',
  3535. 'return "spawned"',
  3536. ].join('\n'),
  3537. bindings: [],
  3538. })
  3539. expect(result.error).toBeUndefined()
  3540. expect(result.value).toBe('spawned')
  3541. // The trap really installed before the leader returned, so this is the
  3542. // SIGTERM-ignoring descendant, not one that would have died to the default.
  3543. expect(existsSync(readyMarker)).toBe(true)
  3544. // The grace-window SIGKILL (graceMs 300 + reap margin) empties the group. Once
  3545. // it has, the descendant stops bumping the heartbeat. Poll the heartbeat's
  3546. // mtime: two consecutive reads far enough apart with no change means it is no
  3547. // longer executing — true whether it was reaped or lingers as a zombie, so
  3548. // the assertion holds in a container whose init does not wait() orphans. The
  3549. // window (well under the 30 s self-timeout) proves the SIGKILL did the work.
  3550. const mtime = (): number => { try { return statSync(heartbeat).mtimeMs } catch { return 0 } }
  3551. const stopDeadline = Date.now() + 8_000
  3552. let last = mtime()
  3553. let still = false
  3554. while (Date.now() < stopDeadline) {
  3555. await new Promise(resolve => setTimeout(resolve, 400))
  3556. const now = mtime()
  3557. if (now === last && now !== 0) { still = true; break }
  3558. last = now
  3559. }
  3560. expect(still).toBe(true)
  3561. }, 20_000)
  3562. it('dispose awaits reaping of a same-group survivor from a completed run', async () => {
  3563. // The quiescence contract also holds for a run that ALREADY resolved: the run
  3564. // stays tracked in `live` until its process group is reaped, so a `dispose()`
  3565. // that races a just-returned run() still awaits the survivor rather than
  3566. // snapshotting an empty `live` and returning while it lives. Here the run
  3567. // completes (leaving a SIGTERM-ignoring same-group descendant), then dispose()
  3568. // is called; the heartbeat must be stale BY THE TIME dispose() resolves —
  3569. // proving teardown waited for the reap, not merely that the reap eventually
  3570. // happened.
  3571. const handoff = await makeTempDir('dsh-dispose-quiesce-')
  3572. const readyMarker = join(handoff, 'ready')
  3573. const heartbeat = join(handoff, 'heartbeat')
  3574. const { runtime, fiber } = await setup({ maxWallMs: 10_000, graceMs: 300 })
  3575. const result = await runtime.run({
  3576. program: [
  3577. 'import subprocess, sys, os, time',
  3578. `marker = ${JSON.stringify(readyMarker)}`,
  3579. `heartbeat = ${JSON.stringify(heartbeat)}`,
  3580. 'code = ("import signal, sys, time\\n"',
  3581. ' "signal.signal(signal.SIGTERM, signal.SIG_IGN)\\n"',
  3582. ' "open(sys.argv[1], \'w\').close()\\n"',
  3583. ' "end = time.time() + 30\\n"',
  3584. ' "while time.time() < end:\\n"',
  3585. ' " open(sys.argv[2], \'w\').close()\\n"',
  3586. ' " time.sleep(0.05)\\n")',
  3587. 'child = subprocess.Popen([sys.executable, "-c", code, marker, heartbeat],',
  3588. ' stdin=subprocess.DEVNULL,',
  3589. ' stdout=subprocess.DEVNULL,',
  3590. ' stderr=subprocess.DEVNULL)',
  3591. 'deadline = time.time() + 5',
  3592. 'while not os.path.exists(marker) and time.time() < deadline:',
  3593. ' time.sleep(0.02)',
  3594. 'return "spawned"',
  3595. ].join('\n'),
  3596. bindings: [],
  3597. })
  3598. expect(result.error).toBeUndefined()
  3599. expect(existsSync(readyMarker)).toBe(true)
  3600. // dispose() must not return until the group is reaped. After it resolves, the
  3601. // heartbeat must already be stale: read its mtime, wait past the heartbeat
  3602. // interval, and confirm it did not advance — the descendant is no longer
  3603. // executing (reaped or zombie), so teardown was genuinely quiescent.
  3604. await fiber.dispose()
  3605. const mtime = (): number => { try { return statSync(heartbeat).mtimeMs } catch { return 0 } }
  3606. const afterDispose = mtime()
  3607. // Pin the assertion to a heartbeat that actually ran: mtime() returns 0 when
  3608. // the file never existed, so without this the `toBe` below would pass
  3609. // vacuously (0 === 0) if the survivor never wrote a heartbeat at all.
  3610. expect(afterDispose).toBeGreaterThan(0)
  3611. await new Promise(resolve => setTimeout(resolve, 500))
  3612. expect(mtime()).toBe(afterDispose)
  3613. }, 20_000)
  3614. it('sends SIGKILL at the poll deadline when the event loop was blocked past both timers', async () => {
  3615. // If the host event loop is blocked (a big synchronous computation) from
  3616. // before the group-reap poll was scheduled until after the deadline, both the
  3617. // poll timer and the grace-window SIGKILL timer are overdue when the loop
  3618. // resumes. Node runs the earlier-scheduled poll first, so the SIGKILL timer
  3619. // may not have fired yet. The deadline arm must then send SIGKILL ITSELF
  3620. // rather than cancel the unfired escalation — otherwise a SIGTERM-ignoring
  3621. // same-group survivor is released for good. A synchronous busy-loop after
  3622. // run() resolves reproduces the block deterministically.
  3623. const handoff = await makeTempDir('dsh-deadline-')
  3624. const readyMarker = join(handoff, 'ready')
  3625. const heartbeat = join(handoff, 'heartbeat')
  3626. const graceMs = 300
  3627. const { runtime } = await setup({ maxWallMs: 10_000, graceMs })
  3628. const result = await runtime.run({
  3629. program: [
  3630. 'import subprocess, sys, os, time',
  3631. `marker = ${JSON.stringify(readyMarker)}`,
  3632. `heartbeat = ${JSON.stringify(heartbeat)}`,
  3633. 'code = ("import signal, sys, time\\n"',
  3634. ' "signal.signal(signal.SIGTERM, signal.SIG_IGN)\\n"',
  3635. ' "open(sys.argv[1], \'w\').close()\\n"',
  3636. ' "end = time.time() + 30\\n"',
  3637. ' "while time.time() < end:\\n"',
  3638. ' " open(sys.argv[2], \'w\').close()\\n"',
  3639. ' " time.sleep(0.05)\\n")',
  3640. 'child = subprocess.Popen([sys.executable, "-c", code, marker, heartbeat],',
  3641. ' stdin=subprocess.DEVNULL,',
  3642. ' stdout=subprocess.DEVNULL,',
  3643. ' stderr=subprocess.DEVNULL)',
  3644. 'deadline = time.time() + 5',
  3645. 'while not os.path.exists(marker) and time.time() < deadline:',
  3646. ' time.sleep(0.02)',
  3647. 'return "spawned"',
  3648. ].join('\n'),
  3649. bindings: [],
  3650. })
  3651. expect(result.error).toBeUndefined()
  3652. expect(existsSync(readyMarker)).toBe(true)
  3653. // Block the event loop synchronously past graceMs + CLOSE_REAP_MARGIN_MS
  3654. // (2000) with margin, so both timers are overdue when the loop resumes.
  3655. const blockUntil = Date.now() + graceMs + 2_000 + 800
  3656. while (Date.now() < blockUntil) { /* busy-wait, no yield */ }
  3657. // Yield: the overdue poll runs (group still non-empty, deadline passed) and
  3658. // must send SIGKILL itself. The survivor then stops bumping the heartbeat.
  3659. const mtime = (): number => { try { return statSync(heartbeat).mtimeMs } catch { return 0 } }
  3660. const stopDeadline = Date.now() + 5_000
  3661. let last = mtime()
  3662. let stopped = false
  3663. while (Date.now() < stopDeadline) {
  3664. await new Promise(resolve => setTimeout(resolve, 400))
  3665. const now = mtime()
  3666. if (now === last && now !== 0) { stopped = true; break }
  3667. last = now
  3668. }
  3669. expect(stopped).toBe(true)
  3670. }, 20_000)
  3671. })
  3672. describe('PythonCodeRuntime — hostile peer', () => {
  3673. it('drops garbage bytes and unknown-shape frames posted directly to fd 3', async () => {
  3674. // The model program can reach fd 3 and write anything. We inject a
  3675. // non-JSON line, a valid JSON but unknown-shape frame, and a broken done
  3676. // frame; the host must not crash, and the real `done` still settles the run.
  3677. const { runtime } = await setup()
  3678. const result = await runtime.run({
  3679. program: [
  3680. 'import os',
  3681. 'os.write(3, b"not-json\\n")',
  3682. 'os.write(3, b\'{"type":"unknown"}\\n\')',
  3683. 'os.write(3, b\'{"type":"done","error":{"message":42}}\\n\')',
  3684. 'return "survived"',
  3685. ].join('\n'),
  3686. bindings: [],
  3687. })
  3688. expect(result.error).toBeUndefined()
  3689. expect(result.value).toBe('survived')
  3690. })
  3691. it('answers a forged call frame for an unknown binding and never crashes', async () => {
  3692. // The unknown-binding reply path, driven through the id the host expects:
  3693. // the program lets its own first call claim id 0 and forges id 1, which the
  3694. // host answers with the `unknown binding` rejection the honest call would
  3695. // have received. A forged id out of sequence is dropped instead — that is
  3696. // the id-bound test below, not this one.
  3697. const { runtime } = await setup({ maxWallMs: 8_000 })
  3698. let seenLegitCall = false
  3699. const result = await runtime.run({
  3700. program: [
  3701. 'import os, json',
  3702. 'x = await tools.echo({"ping": True})',
  3703. 'os.write(3, json.dumps({"type":"call","id":1,"global":"tools","name":"forged","args":{}}).encode() + b"\\n")',
  3704. // The forged frame is answered, but nothing in the child awaits id 1, so
  3705. // the reply is ignored and the run completes on its own value.
  3706. 'return x',
  3707. ].join('\n'),
  3708. bindings: tools({
  3709. echo: async (args) => { seenLegitCall = true; return args as CodeJsonValue },
  3710. }),
  3711. })
  3712. expect(result.error).toBeUndefined()
  3713. expect(result.value).toEqual({ ping: true })
  3714. expect(seenLegitCall).toBe(true)
  3715. }, 15_000)
  3716. it('caps the unknown-binding preview for a huge forged name', async () => {
  3717. // The unknown-binding reply's JSON.stringify ran on the WHOLE capped
  3718. // target, allocating the escaped form — up to ~6x under control-heavy
  3719. // input. The preview is now built from a 1 KiB prefix, so a forged call
  3720. // with a huge global/name cannot spike host memory near the value ceiling;
  3721. // the reply still identifies the binding.
  3722. const { runtime } = await setup({ maxWallMs: 8_000, maxValueBytes: 1024 * 1024 })
  3723. const result = await runtime.run({
  3724. program: [
  3725. 'import os, json',
  3726. 'x = await tools.echo({"ping": True})',
  3727. 'name = "n" * 100000',
  3728. 'os.write(3, json.dumps({"type":"call","id":1,"global":"tools","name":name,"args":{}}).encode() + b"\\n")',
  3729. 'return x',
  3730. ].join('\n'),
  3731. bindings: tools({
  3732. echo: async args => args as CodeJsonValue,
  3733. }),
  3734. })
  3735. expect(result.error).toBeUndefined()
  3736. expect(result.value).toEqual({ ping: true })
  3737. }, 15_000)
  3738. it('drops forged call frames whose ids are not the next in sequence, retaining no per-id state', async () => {
  3739. // The host used to remember every answered id in a Set, so a program could
  3740. // write an unbounded run of unique forged ids — each frame far below the
  3741. // 64 MiB cap, so nothing rejected them — and grow host memory for the
  3742. // whole run. Ids are consecutive from 0, so one counter replaces the set.
  3743. //
  3744. // The discriminator is that the forgeries must not be answered. Each names a
  3745. // binding that does exist, so a host answering them would run `echo` once
  3746. // per forgery; the count proves only the legitimate call was dispatched.
  3747. // Ids also run DESCENDING, so a high-water-mark test would drop the honest
  3748. // call that follows rather than the forgeries.
  3749. const { runtime } = await setup()
  3750. let echoCalls = 0
  3751. const result = await runtime.run({
  3752. program: [
  3753. 'import os, json',
  3754. 'for i in range(2000, 0, -1):',
  3755. ' os.write(3, json.dumps({"type":"call","id":i,"global":"tools","name":"echo","args":{"forged":i}}).encode() + b"\\n")',
  3756. 'x = await tools.echo({"ping": True})',
  3757. 'return x',
  3758. ].join('\n'),
  3759. bindings: tools({
  3760. echo: async (args) => { echoCalls += 1; return args as CodeJsonValue },
  3761. }),
  3762. })
  3763. expect(result.error).toBeUndefined()
  3764. expect(result.value).toEqual({ ping: true })
  3765. expect(echoCalls).toBe(1)
  3766. }, 15_000)
  3767. it('keeps answering calls a program makes after one with unserializable arguments', async () => {
  3768. // The child claims an id only once its write succeeds, so a call rejected
  3769. // child-side for non-lossless arguments leaves no gap. Were a gap possible,
  3770. // the host's exact-successor test would drop every later call and the run
  3771. // would hang to the wall ceiling instead of completing.
  3772. const { runtime } = await setup({ maxWallMs: 8_000 })
  3773. const seen: unknown[] = []
  3774. const result = await runtime.run({
  3775. program: [
  3776. 'caught = ""',
  3777. 'try:',
  3778. ' await tools.echo({"bad": float("inf")})',
  3779. 'except RuntimeError as e:',
  3780. ' caught = str(e)',
  3781. 'after = await tools.echo({"ok": True})',
  3782. 'return {"caught": caught, "after": after}',
  3783. ].join('\n'),
  3784. bindings: tools({
  3785. echo: async (args) => { seen.push(args); return args as CodeJsonValue },
  3786. }),
  3787. })
  3788. expect(result.error).toBeUndefined()
  3789. const value = result.value as { caught: string; after: unknown }
  3790. expect(value.caught).toContain('lossless JSON')
  3791. expect(value.after).toEqual({ ok: true })
  3792. // The rejected call never reached the host; the one after it did.
  3793. expect(seen).toEqual([{ ok: true }])
  3794. }, 15_000)
  3795. it('drops a forged frame carrying an integer outside JavaScript safe range', async () => {
  3796. // JSON.parse would silently round 9007199254740993 to ...992 BEFORE any
  3797. // validation, corrupting a dispatched argument or completion. The host
  3798. // scans the raw line and drops such frames as hostile traffic; the honest
  3799. // child cannot produce one (its validator rejects unsafe ints).
  3800. const { runtime } = await setup()
  3801. let dispatched: unknown
  3802. const result = await runtime.run({
  3803. program: [
  3804. 'import os',
  3805. // Forged call frame with an unsafe int argument, then a forged done
  3806. // frame with an unsafe int value — both must be dropped whole.
  3807. 'os.write(3, b\'{"type":"call","id":7,"global":"tools","name":"echo","args":9007199254740993}\\n\')',
  3808. 'os.write(3, b\'{"type":"done","value":9007199254740993}\\n\')',
  3809. 'x = await tools.echo({"ok": True})',
  3810. 'return x',
  3811. ].join('\n'),
  3812. bindings: tools({
  3813. echo: async (args) => { dispatched = args; return args as CodeJsonValue },
  3814. }),
  3815. })
  3816. expect(result.error).toBeUndefined()
  3817. // The forged done did not settle the run; the legit call and completion did.
  3818. expect(result.value).toEqual({ ok: true })
  3819. expect(dispatched).toEqual({ ok: true })
  3820. })
  3821. it('truncates host-side logs once the budget is exhausted and emits the marker', async () => {
  3822. // Set a tiny host-side budget; the Python side has a much larger one, so
  3823. // its LogBuffer will not truncate — the host ledger fires first.
  3824. const { runtime } = await setup({ maxLogBytes: 128 })
  3825. const result = await runtime.run({
  3826. program: [
  3827. 'for _ in range(50):',
  3828. ' print("aaaaaaaaaa")',
  3829. 'return "done"',
  3830. ].join('\n'),
  3831. bindings: [],
  3832. })
  3833. expect(result.error).toBeUndefined()
  3834. const markers = result.logs.filter(line => line.includes('log capture truncated at 128 bytes'))
  3835. expect(markers.length).toBeGreaterThanOrEqual(1)
  3836. })
  3837. it('reports an exception whose message holds an unpaired surrogate instead of stranding to the wall clock', async () => {
  3838. // A strict UTF-8 encode of "\ud800" throws while BUILDING the failure
  3839. // frame; the run would then hang to maxWallMs and misreport as timeout.
  3840. const { runtime } = await setup({ maxWallMs: 8_000 })
  3841. const result = await runtime.run({
  3842. program: String.raw`raise Exception("bad \ud800 surrogate")`,
  3843. bindings: [],
  3844. })
  3845. expect(result.error?.kind).toBe('exception')
  3846. expect(result.error?.message).toContain('bad')
  3847. expect(result.error?.message).toContain('surrogate')
  3848. })
  3849. it('carries a lone-surrogate completion string across the wire as its JSON escape', async () => {
  3850. // UTF-8 has no encoding for a lone surrogate, but JSON does: the ASCII
  3851. // `\ud800` escape, which JSON.parse reads back as the same UTF-16 code
  3852. // unit. `CodeJsonValue`, `snapshotJsonValue`, and the worker backend all
  3853. // accept such a string, so this backend must not narrow the shared seam.
  3854. const { runtime } = await setup()
  3855. const result = await runtime.run({
  3856. program: String.raw`return {"lone": "a\ud800b", "spelled": "😀"}`,
  3857. bindings: [],
  3858. })
  3859. expect(result.error).toBeUndefined()
  3860. // The lone half survives as the code unit itself; a spelled-out high-low
  3861. // PAIR folds into the astral character the host would hold for it.
  3862. expect(result.value).toEqual({ lone: 'a\ud800b', spelled: '\u{1f600}' })
  3863. })
  3864. it('meters a lone surrogate at its six escaped bytes, matching the host', async () => {
  3865. // The child and the host share maxValueBytes, so the child must charge the
  3866. // escape's six ASCII bytes (plus two quotes): eight fits, nine does not.
  3867. const { runtime } = await setup({ maxValueBytes: 8 })
  3868. const ok = await runtime.run({ program: String.raw`return "\ud800"`, bindings: [] })
  3869. expect(ok.error).toBeUndefined()
  3870. expect(ok.value).toBe('\ud800')
  3871. const over = await setup({ maxValueBytes: 7 })
  3872. const result = await over.runtime.run({ program: String.raw`return "\ud800"`, bindings: [] })
  3873. expect(result.error?.kind).toBe('output-limit')
  3874. })
  3875. it('meters a surrogate-dense completion by counting, not by materializing a match list', async () => {
  3876. // `_json_str_cost` counted lone surrogates with `_SURROGATE.findall`,
  3877. // which materializes one single-character string PER surrogate: a
  3878. // surrogate-dense value near the budget (each surrogate serializes to six
  3879. // bytes, so a budget-sized value holds millions of them) would allocate
  3880. // millions of objects before the meter returned — an O(N)-objects spike
  3881. // that defeats the meter's documented contract of counting without
  3882. // building. The count is now a length difference over the removal `sub`
  3883. // already performs. Three million lone surrogates pin the boundary at
  3884. // scale: 18,000,002 serialized bytes succeed at an 18,000,002 budget and
  3885. // report output-limit one byte under, proving the meter counts every
  3886. // surrogate exactly rather than dropping or over-charging any.
  3887. const { runtime } = await setup({ maxValueBytes: 18_000_002 })
  3888. const ok = await runtime.run({ program: 'return "\\ud800" * 3000000', bindings: [] })
  3889. expect(ok.error).toBeUndefined()
  3890. expect(ok.value).toBe('\ud800'.repeat(3_000_000))
  3891. const over = await setup({ maxValueBytes: 18_000_001 })
  3892. const result = await over.runtime.run({ program: 'return "\\ud800" * 3000000', bindings: [] })
  3893. expect(result.error?.kind).toBe('output-limit')
  3894. }, 60_000)
  3895. it('passes a lone-surrogate binding argument through instead of failing the call', async () => {
  3896. // The argument validator shared the same over-narrow rejection; a host
  3897. // binding must receive the code unit the program passed.
  3898. const seen: unknown[] = []
  3899. const { runtime } = await setup()
  3900. const result = await runtime.run({
  3901. program: String.raw`return await tools.echo({"text": "x\udfff"})`,
  3902. bindings: tools({ echo: async (args: unknown) => { seen.push(args); return args as CodeJsonValue } }),
  3903. })
  3904. expect(result.error).toBeUndefined()
  3905. expect(seen).toEqual([{ text: 'x\udfff' }])
  3906. expect(result.value).toEqual({ text: 'x\udfff' })
  3907. })
  3908. it('meters a non-ASCII completion in UTF-8 JSON bytes, matching the host', async () => {
  3909. // json.dumps' default \uXXXX escaping would count "é" as 8 bytes while
  3910. // the host meter counts its UTF-8 JSON form (4); the shared budget must
  3911. // agree, so a 4-byte-fitting value passes a maxValueBytes of 4.
  3912. const { runtime } = await setup({ maxValueBytes: 4 })
  3913. const ok = await runtime.run({ program: 'return "é"', bindings: [] })
  3914. expect(ok.error).toBeUndefined()
  3915. expect(ok.value).toBe('é')
  3916. const over = await runtime.run({ program: 'return "éx"', bindings: [] })
  3917. expect(over.error?.kind).toBe('output-limit')
  3918. })
  3919. it('filters bootstrap frames from exception-group members (TaskGroup)', async () => {
  3920. // Python 3.11+ stores member stacks under TracebackException.exceptions;
  3921. // the <model>-frame filter must recurse into them too.
  3922. const { runtime } = await setup()
  3923. const result = await runtime.run({
  3924. program: [
  3925. 'import asyncio, sys',
  3926. 'if sys.version_info < (3, 11):',
  3927. ' raise ValueError("skip-old <model>")',
  3928. 'async def boom():',
  3929. ' raise ValueError("group-member")',
  3930. 'async with asyncio.TaskGroup() as tg:',
  3931. ' tg.create_task(boom())',
  3932. ].join('\n'),
  3933. bindings: [],
  3934. })
  3935. expect(result.error?.kind).toBe('exception')
  3936. expect(result.error?.message).toContain('<model>')
  3937. expect(result.error?.message).not.toContain('bootstrap.py')
  3938. })
  3939. it('keeps frames intact when a model thread floods logs while a large frame drains', async () => {
  3940. // os.write releases the GIL and a frame beyond PIPE_BUF is not atomic:
  3941. // without the writer lock + full-write loop, the printing thread could
  3942. // interleave bytes mid-frame and the host would drop the malformed JSON,
  3943. // hanging the run to the wall clock (or losing the completion).
  3944. const { runtime } = await setup({ maxValueBytes: 1024 * 1024, maxLogBytes: 4 * 1024 * 1024, maxWallMs: 15_000 })
  3945. const result = await runtime.run({
  3946. program: [
  3947. 'import threading',
  3948. 'stop = False',
  3949. 'def spam():',
  3950. ' while not stop:',
  3951. ' print("spam-line-" + "y" * 100)',
  3952. 't = threading.Thread(target=spam)',
  3953. 't.start()',
  3954. // A ~300 KiB completion — several PIPE_BUF units — while spam runs.
  3955. 'big = "x" * (300 * 1024)',
  3956. 'stop = True',
  3957. 't.join()',
  3958. 'return big',
  3959. ].join('\n'),
  3960. bindings: [],
  3961. })
  3962. expect(result.error).toBeUndefined()
  3963. expect(result.value).toBe('x'.repeat(300 * 1024))
  3964. }, 20_000)
  3965. it('settles cleanly while a daemon thread keeps writing unterminated log text', async () => {
  3966. // WARNING regression: the settlement `flush_out()/flush_err()` on the main
  3967. // coroutine read and clear `_LogStream._pending` and the shared LogBuffer
  3968. // ledger with NO lock, while a model daemon thread's `print`/`write` mutate
  3969. // the same state. Capturing the bound method (`out_stream.flush_line`) only
  3970. // fixes WHICH callable runs, not what it reads mid-flight: the flush could
  3971. // interleave with a concurrent write and join a `_pending` list being
  3972. // mutated under it, corrupting the ledger and costing the `done` frame — the
  3973. // run would then strand to the wall clock instead of completing. The shared
  3974. // re-entrant lock serializes them.
  3975. //
  3976. // A pure data race has no single bad input to reject deterministically, so
  3977. // this maximizes overlap: daemon threads emit UNTERMINATED writes (which
  3978. // pile into `_pending` rather than flushing per line) right up to the moment
  3979. // the body returns and settlement flushes. Repeated so the interleave lands.
  3980. for (let attempt = 0; attempt < 5; attempt++) {
  3981. const { runtime, fiber } = await setup({ maxLogBytes: 4 * 1024 * 1024, maxWallMs: 15_000 })
  3982. const result = await runtime.run({
  3983. program: [
  3984. 'import sys, threading',
  3985. 'stop = False',
  3986. 'def spam():',
  3987. ' while not stop:',
  3988. // No newline: the text accumulates in the stream's `_pending`, which is
  3989. // exactly the state the settlement flush also touches.
  3990. ' sys.stdout.write("tail-fragment-" + "z" * 64)',
  3991. 'workers = [threading.Thread(target=spam, daemon=True) for _ in range(4)]',
  3992. 'for t in workers: t.start()',
  3993. // Let the daemons build up pending writes, then return so settlement
  3994. // flushes while they are still mid-write.
  3995. 'import time; time.sleep(0.05)',
  3996. 'return "settled"',
  3997. ].join('\n'),
  3998. bindings: [],
  3999. })
  4000. expect(result.error).toBeUndefined()
  4001. expect(result.value).toBe('settled')
  4002. await fiber.dispose()
  4003. }
  4004. }, 30_000)
  4005. it('completes a binding called from a worker thread on its own event loop', async () => {
  4006. // A binding reply Future is created on the loop that ran `dispatch`. When the
  4007. // model calls a binding from a worker THREAD via `asyncio.run(tools.x(...))`,
  4008. // that Future belongs to the thread's loop, not the main loop where
  4009. // `_pump_replies` reads the reply. `asyncio.Future` is not thread-safe:
  4010. // completing it from another thread does not wake its own loop, so a direct
  4011. // `set_result` would strand the awaiting thread and the run would degrade to a
  4012. // wall-clock timeout. The pump must schedule completion on the Future's own
  4013. // loop via `call_soon_threadsafe`. The tight maxWallMs makes the pre-fix
  4014. // failure a fast timeout rather than a hang.
  4015. //
  4016. // The main coroutine yields with `await asyncio.sleep` while the worker runs,
  4017. // rather than a synchronous `t.join()`: joining would block the main thread,
  4018. // so the main loop could not run `_pump_replies` and the call would deadlock
  4019. // regardless of the fix — that blocks the pump, not the cross-loop delivery
  4020. // this test pins.
  4021. const { runtime } = await setup({ maxWallMs: 8_000 })
  4022. const seen: unknown[] = []
  4023. const result = await runtime.run({
  4024. program: [
  4025. 'import asyncio, threading',
  4026. 'result = {}',
  4027. 'def worker():',
  4028. // A fresh loop in this thread; the binding Future is created here.
  4029. ' result["value"] = asyncio.run(tools.echo({"from": "thread"}))',
  4030. 't = threading.Thread(target=worker)',
  4031. 't.start()',
  4032. 'while t.is_alive():',
  4033. ' await asyncio.sleep(0.02)',
  4034. 'return result["value"]',
  4035. ].join('\n'),
  4036. bindings: tools({
  4037. echo: async (args) => { seen.push(args); return args as CodeJsonValue },
  4038. }),
  4039. })
  4040. expect(result.error).toBeUndefined()
  4041. expect(result.value).toEqual({ from: 'thread' })
  4042. // The host binding actually ran (the reply round-tripped), not a timeout.
  4043. expect(seen).toEqual([{ from: 'thread' }])
  4044. }, 15_000)
  4045. it('keeps the reply pump alive when a late reply targets a closed thread loop', async () => {
  4046. // A binding called from a worker thread that ABANDONS the call (its
  4047. // `asyncio.run` is cancelled) leaves the pending entry holding that thread's
  4048. // loop, which `asyncio.run` closes on return. When the host later answers
  4049. // that call, `_pump_replies` schedules the completion onto the closed loop —
  4050. // `call_soon_threadsafe` raises `RuntimeError('Event loop is closed')`.
  4051. // Unguarded, that RuntimeError ends the pump task and strands every later
  4052. // reply; the guard drops the moot reply and keeps the pump serving.
  4053. //
  4054. // The ordering is a STRUCTURAL guarantee, not a timing window: the worker
  4055. // closes its loop before the main coroutine signals `closed`; the host
  4056. // answers the abandoned `slow` call (hitting the closed loop) before it
  4057. // answers `release`, because `release`'s handler only resolves `slow` first
  4058. // and then yields a microtask. So the pump provably meets the closed loop on
  4059. // `slow`'s reply before it must deliver `release`'s. Fail-before: the pump
  4060. // dies on `slow`, `release`'s reply is never read, and `await tools.release`
  4061. // hangs to the (small) maxWallMs as a timeout.
  4062. let releaseSlow!: () => void
  4063. const slowGate = new Promise<void>((resolve) => { releaseSlow = resolve })
  4064. const { runtime } = await setup({ maxWallMs: 6_000 })
  4065. const result = await runtime.run({
  4066. program: [
  4067. 'import asyncio, threading',
  4068. 'closed = threading.Event()',
  4069. 'def worker():',
  4070. ' async def body():',
  4071. // Abandon the call: wait_for cancels it, but the pending host-side entry
  4072. // survives (dispatch does not pop on cancellation), holding this loop.
  4073. ' try:',
  4074. ' await asyncio.wait_for(tools.slow({}), timeout=0.1)',
  4075. ' except asyncio.TimeoutError:',
  4076. ' pass',
  4077. ' asyncio.run(body())', // closes the thread's loop on return
  4078. ' closed.set()',
  4079. 't = threading.Thread(target=worker)',
  4080. 't.start()',
  4081. 'while not closed.is_set():',
  4082. ' await asyncio.sleep(0.02)',
  4083. // The loop is closed. Now the host answers slow (dead-loop reply) then
  4084. // release; the pump must survive the first to deliver the second.
  4085. 'after = await tools.release({})',
  4086. 'return after',
  4087. ].join('\n'),
  4088. bindings: tools({
  4089. slow: async () => {
  4090. // Answer only once the worker has closed its loop AND the main
  4091. // coroutine is awaiting release, so this reply reaches the pump against
  4092. // the closed loop.
  4093. await slowGate
  4094. return 'late'
  4095. },
  4096. release: async () => {
  4097. // Let slow's reply be written first, then yield a microtask so the
  4098. // pump processes the dead-loop reply before release's own reply lands.
  4099. releaseSlow()
  4100. await new Promise(resolve => setImmediate(resolve))
  4101. return 'released'
  4102. },
  4103. }),
  4104. })
  4105. expect(result.error).toBeUndefined()
  4106. // The pump survived the closed-loop reply and delivered the later binding.
  4107. expect(result.value).toBe('released')
  4108. }, 15_000)
  4109. it('keeps the reply pump alive when RuntimeError is rebound before the program runs', async () => {
  4110. // `_pump_replies` catches a closed-loop scheduling failure with `except
  4111. // _RuntimeError`. If that name were bound as a pump BODY local, it would be
  4112. // captured at pump-start — but `_run` reaches the model's top-level
  4113. // statements (which run before the pump's first step, since there is no
  4114. // suspension point between `create_task` and `await __dsh_main__`) with the
  4115. // rebind already applied, so `_RuntimeError` would capture the REBOUND class
  4116. // and the closed-loop `RuntimeError` would escape, killing the pump. Binding
  4117. // it as a DEF-TIME default argument captures the original before any model
  4118. // code runs. This rebinds `__main__.RuntimeError` as the very first program
  4119. // statement and drives the closed-loop worker pattern: the pump must survive
  4120. // the dead-loop reply and deliver the later binding.
  4121. let releaseSlow!: () => void
  4122. const slowGate = new Promise<void>((resolve) => { releaseSlow = resolve })
  4123. const { runtime } = await setup({ maxWallMs: 6_000 })
  4124. const result = await runtime.run({
  4125. program: [
  4126. 'import __main__',
  4127. '__main__.RuntimeError = ValueError',
  4128. 'import asyncio, threading',
  4129. 'closed = threading.Event()',
  4130. 'def worker():',
  4131. ' async def body():',
  4132. ' try:',
  4133. ' await asyncio.wait_for(tools.slow({}), timeout=0.1)',
  4134. ' except asyncio.TimeoutError:',
  4135. ' pass',
  4136. ' asyncio.run(body())',
  4137. ' closed.set()',
  4138. 't = threading.Thread(target=worker)',
  4139. 't.start()',
  4140. 'while not closed.is_set():',
  4141. ' await asyncio.sleep(0.02)',
  4142. 'after = await tools.release({})',
  4143. 'return after',
  4144. ].join('\n'),
  4145. bindings: tools({
  4146. slow: async () => { await slowGate; return 'late' },
  4147. release: async () => { releaseSlow(); await new Promise(resolve => setImmediate(resolve)); return 'released' },
  4148. }),
  4149. })
  4150. expect(result.error).toBeUndefined()
  4151. expect(result.value).toBe('released')
  4152. }, 15_000)
  4153. it('keeps a successful completion when _done_with_value is rebound', async () => {
  4154. // `_run` calls `_done_with_value(value, max_value_bytes)` after the program
  4155. // returns. The name is a module global, and this bootstrap IS `__main__`, so
  4156. // `__main__._done_with_value = boom` as a program statement would otherwise
  4157. // be resolved at call time and a legitimate success would be rewritten into
  4158. // an `exception`. `_run` now binds `done_with_value_bound = _done_with_value`
  4159. // before the program runs, so the entry name is immune; the run must still
  4160. // report the success value.
  4161. const { runtime } = await setup({ maxWallMs: 10_000 })
  4162. const result = await runtime.run({
  4163. program: [
  4164. 'import __main__',
  4165. 'def boom(*a, **k):',
  4166. ' raise RuntimeError("hijacked")',
  4167. '__main__._done_with_value = boom',
  4168. 'return 1',
  4169. ].join('\n'),
  4170. bindings: [],
  4171. })
  4172. expect(result.error).toBeUndefined()
  4173. expect(result.value).toBe(1)
  4174. }, 15_000)
  4175. it('round-trips an exactly representable large integer through a binding echo', async () => {
  4176. // The reply serializer must print BigInt digits for a beyond-safe
  4177. // integral double: String(2**60) emits a rounded form, and the child
  4178. // would receive a DIFFERENT integer than the binding resolved.
  4179. const { runtime } = await setup()
  4180. const result = await runtime.run({
  4181. program: [
  4182. 'v = await tools.echo(2**60)',
  4183. 'return v == 2**60',
  4184. ].join('\n'),
  4185. bindings: tools({ echo: async args => args as never }),
  4186. })
  4187. expect(result.error).toBeUndefined()
  4188. expect(result.value).toBe(true)
  4189. })
  4190. it('preserves an exactly representable large integer and rejects a rounding one', async () => {
  4191. // The canonical boundary accepts every JS-double-exact value: 2**53 and
  4192. // 2**60 round-trip exactly and must cross (matching the worker backend);
  4193. // 2**53+1 rounds and must fail as invalid-output.
  4194. const { runtime } = await setup()
  4195. const exact = await runtime.run({ program: 'return [2**53, 2**60]', bindings: [] })
  4196. expect(exact.error).toBeUndefined()
  4197. expect(exact.value).toEqual([2 ** 53, 2 ** 60])
  4198. const lossy = await runtime.run({ program: 'return 2**53 + 1', bindings: [] })
  4199. expect(lossy.error?.kind).toBe('invalid-output')
  4200. expect(lossy.error?.message).toContain('not exactly representable')
  4201. })
  4202. it('rejects a container subclass whose overridden methods hide its contents', async () => {
  4203. // A dict subclass returning [] from items() passes an isinstance check but
  4204. // serializes as {}, so the host would receive a value the program did not
  4205. // compute. Exact-type matching fails it as invalid-output instead. The
  4206. // worker backend rejects the prototype-equivalent shapes the same way.
  4207. const { runtime } = await setup()
  4208. const hidden = await runtime.run({
  4209. program: [
  4210. 'class Sneaky(dict):',
  4211. ' def items(self): return []',
  4212. ' def keys(self): return []',
  4213. ' def __iter__(self): return iter([])',
  4214. ' def __len__(self): return 0',
  4215. 'return Sneaky(secret="kept")',
  4216. ].join('\n'),
  4217. bindings: [],
  4218. })
  4219. expect(hidden.error?.kind).toBe('invalid-output')
  4220. expect(hidden.error?.message).toContain('unsupported type (Sneaky)')
  4221. // A list subclass is refused on the same rule.
  4222. const listish = await runtime.run({
  4223. program: ['class L(list):', ' def __iter__(self): return iter([])', 'return L([1, 2, 3])'].join('\n'),
  4224. bindings: [],
  4225. })
  4226. expect(listish.error?.kind).toBe('invalid-output')
  4227. expect(listish.error?.message).toContain('unsupported type (L)')
  4228. // The exact built-in containers still cross unchanged.
  4229. const plain = await runtime.run({ program: 'return {"secret": [1, 2]}', bindings: [] })
  4230. expect(plain.error).toBeUndefined()
  4231. expect(plain.value).toEqual({ secret: [1, 2] })
  4232. })
  4233. it('rejects a scalar subclass whose overrides disagree with what gets serialized', async () => {
  4234. // The validators checked scalars with isinstance, so a subclass passed
  4235. // every check by its real value while the ENCODER read an override — the
  4236. // host then received a value the walk never approved. Each case below is a
  4237. // distinct override reaching a distinct reader.
  4238. const { runtime } = await setup()
  4239. // _dump_float spells a float from repr(value), so an overridden __repr__
  4240. // decides the digits: F(2.5) serialized as 1.
  4241. const floated = await runtime.run({
  4242. program: [
  4243. 'class F(float):',
  4244. ' def __repr__(self): return "1.0"',
  4245. 'return F(2.5)',
  4246. ].join('\n'),
  4247. bindings: [],
  4248. })
  4249. expect(floated.error?.kind).toBe('invalid-output')
  4250. expect(floated.error?.message).toContain('unsupported type (F)')
  4251. // The JS-safe-range bound is two comparisons, so overriding them admits an
  4252. // int whose true digits (json.dumps reads the C-level value) the host's
  4253. // JSON.parse rounds: 9007199254740993 arrives as ...992.
  4254. const inted = await runtime.run({
  4255. program: [
  4256. 'class I(int):',
  4257. ' def __gt__(self, other): return False',
  4258. ' def __lt__(self, other): return False',
  4259. 'return I(2 ** 53 + 1)',
  4260. ].join('\n'),
  4261. bindings: [],
  4262. })
  4263. expect(inted.error?.kind).toBe('invalid-output')
  4264. expect(inted.error?.message).toContain('unsupported type (I)')
  4265. // The pre-encode size bound reads len(), so overriding it to 0 admits a
  4266. // string of any length past maxValueBytes.
  4267. const stringed = await runtime.run({
  4268. program: [
  4269. 'class S(str):',
  4270. ' def __len__(self): return 0',
  4271. 'return S("Q" * 100000)',
  4272. ].join('\n'),
  4273. bindings: [],
  4274. })
  4275. expect(stringed.error?.kind).toBe('invalid-output')
  4276. expect(stringed.error?.message).toContain('unsupported type (S)')
  4277. // A str-subclass dict KEY reaches the same len() bound.
  4278. const keyed = await runtime.run({
  4279. program: [
  4280. 'class S(str):',
  4281. ' def __len__(self): return 0',
  4282. 'return {S("Q" * 100000): 1}',
  4283. ].join('\n'),
  4284. bindings: [],
  4285. })
  4286. expect(keyed.error?.kind).toBe('invalid-output')
  4287. expect(keyed.error?.message).toContain('non-string dict key (S)')
  4288. // bool is an int subclass that IS lossless JSON, and the exact scalars all
  4289. // still cross unchanged.
  4290. const plain = await runtime.run({
  4291. program: 'return {"t": True, "f": False, "n": None, "i": 7, "d": 2.5, "s": "ok"}',
  4292. bindings: [],
  4293. })
  4294. expect(plain.error).toBeUndefined()
  4295. expect(plain.value).toEqual({ t: true, f: false, n: null, i: 7, d: 2.5, s: 'ok' })
  4296. })
  4297. it('rejects a scalar subclass passed as a binding argument', async () => {
  4298. // The uncapped binding-argument validator shares the exact-type rule, so
  4299. // the call fails through its rejection contract instead of dispatching a
  4300. // float whose digits come from an override.
  4301. const { runtime } = await setup()
  4302. const seen: CodeJsonValue[] = []
  4303. const result = await runtime.run({
  4304. program: [
  4305. 'class F(float):',
  4306. ' def __repr__(self): return "1.0"',
  4307. 'try:',
  4308. ' await tools.echo({"v": F(2.5)})',
  4309. 'except Exception as exc:',
  4310. ' return str(exc)',
  4311. ].join('\n'),
  4312. bindings: tools({ echo: async (args) => {
  4313. seen.push(args as CodeJsonValue)
  4314. return null
  4315. } }),
  4316. })
  4317. expect(result.error).toBeUndefined()
  4318. expect(result.value).toContain('unsupported type (F)')
  4319. expect(seen).toEqual([])
  4320. })
  4321. it('rejects a container subclass passed as a binding argument', async () => {
  4322. // Binding arguments run the uncapped validator, which must apply the same
  4323. // exact-type rule: the call fails descriptively instead of dispatching a
  4324. // value whose serialization disagrees with what was validated.
  4325. const { runtime } = await setup()
  4326. const seen: CodeJsonValue[] = []
  4327. const result = await runtime.run({
  4328. program: [
  4329. 'class Sneaky(dict):',
  4330. ' def items(self): return []',
  4331. 'try:',
  4332. ' await tools.echo(Sneaky(secret="kept"))',
  4333. 'except Exception as exc:',
  4334. ' return str(exc)',
  4335. ].join('\n'),
  4336. bindings: tools({ echo: async (args) => {
  4337. seen.push(args as CodeJsonValue)
  4338. return null
  4339. } }),
  4340. })
  4341. expect(result.error).toBeUndefined()
  4342. expect(result.value).toContain('unsupported type (Sneaky)')
  4343. expect(seen).toEqual([])
  4344. })
  4345. it('fails an oversized completion as output-limit without materializing its encoding', async () => {
  4346. // A 100 MiB string under maxValueBytes: 1024 must fail as output-limit.
  4347. // The address-space cap leaves room for the program to BUILD the string
  4348. // (one copy + interpreter) but not for the old full pre-check encode,
  4349. // which materialized chunk fragments plus the joined copy (~2 more
  4350. // copies) and died on RLIMIT_AS as MemoryError/worker-exit.
  4351. const { runtime } = await setup({ maxValueBytes: 1024, addressSpaceMb: 384, maxWallMs: 15_000 })
  4352. const result = await runtime.run({
  4353. program: 'return "x" * (100 * 1024 * 1024)',
  4354. bindings: [],
  4355. })
  4356. expect(result.error?.kind).toBe('output-limit')
  4357. expect(result.error?.message).toContain('exceeded 1024 bytes')
  4358. }, 20_000)
  4359. it('rejects a control-heavy oversized completion on its length, not its escaped copy', async () => {
  4360. // Every "\x00" escapes to the six bytes "\u0000", so the escaped form of a
  4361. // 40 MB string is ~240 MB. The walk must refuse on the cheap
  4362. // `len(current) + 2` lower bound; the 384 MiB address space holds the raw
  4363. // string but not its escaped expansion, so a pre-escape check dies on
  4364. // RLIMIT_AS instead of returning output-limit.
  4365. const { runtime } = await setup({ maxValueBytes: 1024, addressSpaceMb: 384, maxWallMs: 15_000 })
  4366. const result = await runtime.run({
  4367. program: 'return "\\x00" * (40 * 1024 * 1024)',
  4368. bindings: [],
  4369. })
  4370. expect(result.error?.kind).toBe('output-limit')
  4371. expect(result.error?.message).toContain('exceeded 1024 bytes')
  4372. }, 20_000)
  4373. it('truncates a single print far above maxLogBytes instead of dying on the encode', async () => {
  4374. // LogBuffer must reject via the cheap char-count lower bound BEFORE
  4375. // UTF-8-encoding the whole string: the full encode of a ~100 MB line
  4376. // would double the allocation and can breach RLIMIT_AS. 256 MiB
  4377. // address space comfortably holds one copy of the 100 MB string but
  4378. // not the pre-fix double allocation plus interpreter overhead spikes.
  4379. const { runtime } = await setup({ maxLogBytes: 1024, addressSpaceMb: 256, maxWallMs: 15_000 })
  4380. const result = await runtime.run({
  4381. program: [
  4382. 'print("x" * (100 * 1024 * 1024))',
  4383. 'return "done"',
  4384. ].join('\n'),
  4385. bindings: [],
  4386. })
  4387. expect(result.error).toBeUndefined()
  4388. expect(result.value).toBe('done')
  4389. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  4390. }, 20_000)
  4391. it('stops host capture at the child ledger truncation, keeping exactly one marker', async () => {
  4392. // The two ledgers exhaust independently. One child entry larger than
  4393. // `maxLogBytes` sends ONLY the marker, so the host budget is still nearly
  4394. // untouched — and the marker used to arrive as an ordinary `log` frame the
  4395. // host could not tell from program output. Text written afterwards was
  4396. // therefore retained AFTER the marker, contradicting the stop-after-
  4397. // truncation contract, and a later host-side exhaustion could append a
  4398. // second marker. The frame now carries `truncated: true`.
  4399. //
  4400. // `os.write(1, ...)` bypasses the child's own stream, so those bytes reach
  4401. // the host as stray stdout and take the host ledger path rather than the
  4402. // child's — which is exactly the route that leaked past the marker.
  4403. const { runtime } = await setup({ maxLogBytes: 64, maxWallMs: 10_000 })
  4404. const result = await runtime.run({
  4405. program: [
  4406. 'import os',
  4407. 'print("y" * 70000)',
  4408. 'os.write(1, b"AFTER")',
  4409. 'return "done"',
  4410. ].join('\n'),
  4411. bindings: [],
  4412. })
  4413. expect(result.error).toBeUndefined()
  4414. expect(result.value).toBe('done')
  4415. const markers = result.logs.filter(line => line.includes('log capture truncated'))
  4416. expect(markers).toHaveLength(1)
  4417. // The marker is the LAST entry: nothing was retained after truncation.
  4418. expect(result.logs.at(-1)).toBe(markers[0])
  4419. expect(result.logs.join('\n')).not.toContain('AFTER')
  4420. }, 20_000)
  4421. it('keeps one marker when a program forges repeated truncation frames', async () => {
  4422. // `truncated` is attacker-reachable: the program owns fd 3 and can write the
  4423. // flag itself, so the field is a hostile input rather than a trusted signal.
  4424. // Repeats must collapse to the single marker the contract promises, and only
  4425. // the literal `true` counts — a forged `"yes"` is rebuilt away by
  4426. // validateChildFrame, so that frame stays ordinary text.
  4427. const { runtime } = await setup({ maxLogBytes: 4096, maxWallMs: 10_000 })
  4428. const result = await runtime.run({
  4429. program: [
  4430. 'import os, json',
  4431. 'os.write(3, json.dumps({"type":"log","text":"first","truncated":"yes"}).encode() + b"\\n")',
  4432. 'os.write(3, json.dumps({"type":"log","text":"MARK-A","truncated":True}).encode() + b"\\n")',
  4433. 'os.write(3, json.dumps({"type":"log","text":"MARK-B","truncated":True}).encode() + b"\\n")',
  4434. 'return "done"',
  4435. ].join('\n'),
  4436. bindings: [],
  4437. })
  4438. expect(result.error).toBeUndefined()
  4439. expect(result.value).toBe('done')
  4440. // The non-boolean flag did not truncate, so its text was captured normally.
  4441. expect(result.logs).toContain('first')
  4442. // The first genuine flag stopped capture and emitted the HOST's own marker;
  4443. // the frame's own text is discarded, so neither payload appears.
  4444. expect(result.logs).not.toContain('MARK-A')
  4445. expect(result.logs).not.toContain('MARK-B')
  4446. expect(result.logs.at(-1)).toBe(logTruncationMarker(4096))
  4447. expect(result.logs.filter(line => line.includes('log capture truncated'))).toHaveLength(1)
  4448. }, 20_000)
  4449. it('discards the text of a forged truncation frame instead of retaining it', async () => {
  4450. // The marker branch bypasses `admit`, so retaining the frame's own text put
  4451. // attacker-controlled bytes into `logs` with no cap at all: measured, a 1 MiB
  4452. // forged text was retained whole under `maxLogBytes: 64`, and the only bound
  4453. // left was the 64 MiB frame parse cap. The host emits its own marker instead,
  4454. // so the retained size is fixed regardless of what the program sent.
  4455. const forgedBytes = 1024 * 1024
  4456. const { runtime } = await setup({ maxLogBytes: 64, maxWallMs: 20_000 })
  4457. const result = await runtime.run({
  4458. program: [
  4459. 'import os, json',
  4460. `big = "A" * ${forgedBytes}`,
  4461. 'os.write(3, json.dumps({"type":"log","truncated":True,"text":big}).encode() + b"\\n")',
  4462. 'return "done"',
  4463. ].join('\n'),
  4464. bindings: [],
  4465. })
  4466. expect(result.error).toBeUndefined()
  4467. expect(result.value).toBe('done')
  4468. // Only the host marker is kept, so the total stays orders of magnitude below
  4469. // what the forgery carried — and below the cap it was trying to escape.
  4470. expect(result.logs).toEqual([logTruncationMarker(64)])
  4471. expect(result.logs.join('').length).toBeLessThan(forgedBytes / 1000)
  4472. }, 30_000)
  4473. it('coalesces unframed fd-3 fragments without recopying the sealed prefix', async () => {
  4474. // The frame ceiling meters payload BYTES, but each retained chunk is its own
  4475. // Buffer with object and backing-store overhead the byte count cannot see:
  4476. // 5000 single-byte newline-free writes produced 5000 chunks holding 5031
  4477. // bytes, so a program pacing such writes could accumulate millions of objects
  4478. // inside the wall budget and exhaust the host heap far below 256 MiB.
  4479. //
  4480. // The observable behavior is that the run still completes normally: the
  4481. // fragments are coalesced rather than rejected, since a slow trickle of bytes
  4482. // is not itself a protocol violation.
  4483. //
  4484. // `Buffer.concat` is wrapped for the duration so the cumulative copy volume
  4485. // is measured rather than inferred: that total is what separates sealing into
  4486. // blocks from re-merging the whole buffer, and both shapes pass every
  4487. // behavioral assertion below.
  4488. //
  4489. // The trickle is terminated with its own newline before the real frame is
  4490. // written. Without that, those 5000 bytes prefix the frame on the SAME line,
  4491. // which then parses as junk and is dropped — correct framing behavior, but it
  4492. // would leave this test asserting the wrong thing.
  4493. // Bound at capture: `Buffer.concat` is a static method, and taking a bare
  4494. // reference to one trips no-unbound-method.
  4495. const realConcat = Buffer.concat.bind(Buffer)
  4496. let copied = 0
  4497. Buffer.concat = (list: readonly Uint8Array[], total?: number): Buffer<ArrayBuffer> => {
  4498. for (const part of list) copied += part.length
  4499. return realConcat(list, total)
  4500. }
  4501. const program = [
  4502. 'import os',
  4503. // Newline-free single-byte writes, spaced so each lands as its own read.
  4504. // 60000 rather than 5000: the trickle has to cross the seal threshold
  4505. // enough times for the two shapes to separate. At 5000 writes there are
  4506. // only four seals, so even the quadratic form copies well under a
  4507. // megabyte and the budget below could not tell them apart.
  4508. 'for _ in range(60000):',
  4509. ' os.write(3, b"x")',
  4510. ' os.sched_yield()',
  4511. 'os.write(3, b"\\n")',
  4512. // A real frame after the trickle proves framing still works on the
  4513. // coalesced residual.
  4514. 'print("after-trickle")',
  4515. 'return "done"',
  4516. ].join('\n')
  4517. let result: CodeRunResult
  4518. try {
  4519. const { runtime } = await setup({ maxWallMs: 30_000 })
  4520. result = await runtime.run({ program, bindings: [] })
  4521. } finally {
  4522. Buffer.concat = realConcat
  4523. }
  4524. expect(result.error).toBeUndefined()
  4525. expect(result.value).toBe('done')
  4526. expect(result.logs).toContain('after-trickle')
  4527. // Sealing appends a finished block rather than re-merging everything held, so
  4528. // each byte is copied once. Re-concatenating the whole buffer at every
  4529. // threshold made the cumulative copy volume quadratic — 10 MiB trickled a
  4530. // byte at a time copies 53.7 GB that way. A per-byte-copied budget is the
  4531. // discriminator, and it is measured rather than reasoned about: this shape
  4532. // copies about 119 KB for 60000 trickled bytes, the re-merging shape about
  4533. // 540 KB. 256 KiB sits between them with margin on both sides — most writes
  4534. // are coalesced by the pipe before they reach us, so the observed ratio is
  4535. // smaller than the asymptotic one, and the threshold has to sit where a real
  4536. // measurement lands rather than where the asymptote suggests.
  4537. expect(copied).toBeLessThan(256 * 1024)
  4538. }, 40_000)
  4539. it('seals trickled stray fragments into blocks without recopying the sealed prefix', async () => {
  4540. // The stray-capture buffer has the same object-overhead exposure as the fd-3
  4541. // reader above: each newline-free `data` chunk is its own Buffer, so a
  4542. // program pacing single-byte `os.write(1, ...)` accumulates one object per
  4543. // write, which the serialized-cost counter cannot see. Past MAX_PENDING_CHUNKS
  4544. // the fragments seal into a finished block; re-merging the whole residual at
  4545. // each threshold instead would copy the sealed prefix again and again, making
  4546. // the cumulative copy volume quadratic. `Buffer.concat` is wrapped to measure
  4547. // that volume — both shapes admit the same final log entry, so the copy total
  4548. // is the discriminator. maxLogBytes is raised so the trickle is retained,
  4549. // not truncated, which is what forces the fragments to accumulate and seal.
  4550. const realConcat = Buffer.concat.bind(Buffer)
  4551. let copied = 0
  4552. Buffer.concat = (list: readonly Uint8Array[], total?: number): Buffer<ArrayBuffer> => {
  4553. for (const part of list) copied += part.length
  4554. return realConcat(list, total)
  4555. }
  4556. let result: CodeRunResult
  4557. try {
  4558. const { runtime } = await setup({ maxLogBytes: 200_000, maxWallMs: 30_000 })
  4559. result = await runtime.run({
  4560. program: [
  4561. 'import os',
  4562. 'for _ in range(60000):',
  4563. ' os.write(1, b"x")',
  4564. ' os.sched_yield()',
  4565. 'os.write(1, b"\\n")',
  4566. 'return "done"',
  4567. ].join('\n'),
  4568. bindings: [],
  4569. })
  4570. } finally {
  4571. Buffer.concat = realConcat
  4572. }
  4573. expect(result.error).toBeUndefined()
  4574. expect(result.value).toBe('done')
  4575. // The trickle coalesces into one log line (no interior newlines). Its exact
  4576. // length depends on pipe coalescing, but it is one entry and non-empty.
  4577. expect(result.logs.length).toBe(1)
  4578. expect((result.logs[0] as string).length).toBeGreaterThan(0)
  4579. // Sealing appends a finished block rather than re-merging everything held, so
  4580. // each byte is copied a bounded number of times. Re-merging the whole
  4581. // residual at every seal threshold instead makes the cumulative copy volume
  4582. // quadratic. Measured like the fd-3 sibling above rather than reasoned about:
  4583. // this sealed shape copies about 120 KB for 60000 trickled bytes, the
  4584. // re-merging shape about 538 KB (the stray path adds one whole-residual
  4585. // concat at the terminating newline over the fd-3 sibling's 119/540, landing
  4586. // at the same order). 256 KiB sits between them with margin on both sides, so
  4587. // reverting the seal to a re-merge turns this assertion red.
  4588. expect(copied).toBeLessThan(256 * 1024)
  4589. }, 40_000)
  4590. it('caps a huge exception diagnostic child-side before it crosses the wire', async () => {
  4591. // A program can raise with a multi-megabyte message; the child must cap
  4592. // it at maxValueBytes before formatting/sending, not ship the whole
  4593. // payload for the host to truncate after parsing.
  4594. const { runtime } = await setup({ maxValueBytes: 1024 })
  4595. const result = await runtime.run({
  4596. program: 'raise ValueError("boom-" + "x" * (8 * 1024 * 1024))',
  4597. bindings: [],
  4598. })
  4599. expect(result.error?.kind).toBe('exception')
  4600. expect(result.error?.message).toContain('boom-')
  4601. expect(result.error?.message.endsWith('… [truncated]')).toBe(true)
  4602. expect(Buffer.byteLength(result.error?.message ?? '', 'utf8')).toBeLessThan(2048)
  4603. })
  4604. it('caps a control-heavy exception diagnostic by its serialized cost, not raw bytes', async () => {
  4605. // The diagnostic crosses fd 3 inside a JSON frame where a control character
  4606. // escapes sixfold (a NUL is one raw byte, six as `\u0000`). Capping by raw
  4607. // UTF-8 length would let a NUL-heavy message near maxValueBytes serialize to
  4608. // ~6x that and breach the frame ceiling — the silent worker-exit inversion
  4609. // the load-time cap check exists to prevent. The child meters the diagnostic
  4610. // by its serialized cost, so a NUL flood is truncated to fit the frame and
  4611. // the run still reports the exception rather than a worker-exit.
  4612. const { runtime } = await setup({ maxValueBytes: 4096 })
  4613. const result = await runtime.run({
  4614. // 512 KiB of NUL: ~3 MiB once escaped, far past the 4 KiB cap.
  4615. program: 'raise ValueError("\\x00" * (512 * 1024))',
  4616. bindings: [],
  4617. })
  4618. expect(result.error?.kind).toBe('exception')
  4619. expect(result.error?.message.endsWith('… [truncated]')).toBe(true)
  4620. // The SERIALIZED form (what the frame carried) fits the budget, so its raw
  4621. // length is well under it too — a raw-byte cap would have admitted ~4 KiB of
  4622. // NULs that serialize to ~24 KiB.
  4623. const serialized = JSON.stringify(result.error?.message ?? '')
  4624. expect(Buffer.byteLength(serialized, 'utf8')).toBeLessThanOrEqual(4096 + 8)
  4625. })
  4626. it('bounds a newline-free partial-line flood while the program is still running', async () => {
  4627. // print("x", end="") never completes a line, so nothing reaches the
  4628. // Python LogBuffer until settlement — the buffered tail must still hit
  4629. // the budget mid-run instead of growing without bound to RLIMIT/timeout.
  4630. const { runtime } = await setup({ maxLogBytes: 1024, maxWallMs: 15_000 })
  4631. const result = await runtime.run({
  4632. program: [
  4633. 'for _ in range(100000):',
  4634. ' print("xxxxxxxxxx", end="")',
  4635. 'return "done"',
  4636. ].join('\n'),
  4637. bindings: [],
  4638. })
  4639. expect(result.error).toBeUndefined()
  4640. expect(result.value).toBe('done')
  4641. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  4642. // The retained text is bounded by the budget, not the 1 MB the program wrote.
  4643. expect(result.logs.join('\n').length).toBeLessThan(4096)
  4644. }, 20_000)
  4645. it('discards empty writes instead of buffering one list slot each', async () => {
  4646. // An empty chunk adds no character, so the mid-run budget check (which
  4647. // compares buffered CHARS against the remaining ledger) can never fire on
  4648. // it. Buffering empty strings therefore grew `_pending` without bound —
  4649. // millions of slots per CPU second — until RLIMIT_AS turned an append into
  4650. // a MemoryError, long after the log ledger was exhausted. Two million
  4651. // empty writes must instead settle normally and contribute NO log entry,
  4652. // proving the chunk was dropped rather than joined at flush_line.
  4653. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 256, maxWallMs: 20_000 })
  4654. const result = await runtime.run({
  4655. program: [
  4656. 'import sys',
  4657. 'for _ in range(2000000):',
  4658. ' sys.stdout.write("")',
  4659. 'return "done"',
  4660. ].join('\n'),
  4661. bindings: [],
  4662. })
  4663. expect(result.error).toBeUndefined()
  4664. expect(result.value).toBe('done')
  4665. expect(result.logs).toEqual([])
  4666. }, 30_000)
  4667. it('stops scanning a single-write newline flood once the log ledger truncates', async () => {
  4668. // One write carrying half a million newlines: the offset scan must exit the
  4669. // instant LogBuffer truncates rather than re-slicing and pushing every
  4670. // remaining line. If it kept scanning it would exhaust the CPU/wall budget;
  4671. // the run instead settles quickly with exactly one truncation marker.
  4672. const { runtime } = await setup({ maxLogBytes: 256, maxWallMs: 10_000 })
  4673. const start = Date.now()
  4674. const result = await runtime.run({
  4675. program: ['print("x\\n" * 500000, end="")', 'return "done"'].join('\n'),
  4676. bindings: [],
  4677. })
  4678. expect(result.error).toBeUndefined()
  4679. expect(result.value).toBe('done')
  4680. expect(result.logs.filter(line => line.includes('log capture truncated'))).toHaveLength(1)
  4681. expect(Date.now() - start).toBeLessThan(8_000)
  4682. }, 15_000)
  4683. it('bounds an oversized newline-terminated write before joining and slicing it', async () => {
  4684. // The newline branch slices the first line out of the write before
  4685. // `LogBuffer.push` can apply its cheap budget rejection, so a single
  4686. // over-budget write cost a full extra copy of itself in peak address space —
  4687. // the amplification that bound exists to avoid, applied one layer too late.
  4688. // Measured under a 400 MiB addressSpaceMb with the slice unbounded: writes
  4689. // of 200 MiB and up died on MemoryError inside `sys.stdout.write`, reported
  4690. // as the PROGRAM's own exception rather than the promised truncation marker.
  4691. // `"\\n".rjust(n, "A")` is a single allocation ending in the newline, so the
  4692. // payload itself fits and the only remaining allocation is the stream's own
  4693. // slice; 340 MiB of a 400 MiB cap cannot survive one more copy of it.
  4694. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 400, maxWallMs: 30_000 })
  4695. const result = await runtime.run({
  4696. program: [
  4697. 'import sys',
  4698. 'payload = "\\n".rjust(340 * 1024 * 1024, "A")',
  4699. 'sys.stdout.write(payload)',
  4700. 'return "done"',
  4701. ].join('\n'),
  4702. bindings: [],
  4703. })
  4704. expect(result.error).toBeUndefined()
  4705. expect(result.value).toBe('done')
  4706. expect(result.logs).toEqual([logTruncationMarker(256)])
  4707. }, 40_000)
  4708. it('bounds a newline-free write against the already-buffered chunks before joining them', async () => {
  4709. // The newline-free arm buffers the write and then compared the buffered
  4710. // CHARACTER COUNT against the ledger — correct — but paid for the comparison
  4711. // with `"".join(self._pending)`, a second full copy of everything held. One
  4712. // buffered character is enough to make that join a copy of the whole
  4713. // following write. Measured under a 400 MiB addressSpaceMb with a 340 MiB
  4714. // second write: the join raised MemoryError inside `sys.stdout.write`, and
  4715. // because the oversized chunks stayed in `_pending` the settlement
  4716. // `flush_line` raised it again — that throw sits after the `except
  4717. // BaseException` block, so it costs the `done` frame and the run came back
  4718. // `timeout: wall-clock ceiling reached (30000ms)` with no logs at all. The
  4719. // bound must be applied BEFORE the join and the chunks dropped on that path,
  4720. // so the run settles with the truncation marker it promises.
  4721. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 400, maxWallMs: 30_000 })
  4722. const result = await runtime.run({
  4723. program: [
  4724. 'import sys',
  4725. // One unterminated character first, so `_pending` is non-empty and the
  4726. // large write cannot take the "buffered text IS the write" shortcut.
  4727. 'sys.stdout.write("x")',
  4728. 'payload = "A" * (340 * 1024 * 1024)',
  4729. 'sys.stdout.write(payload)',
  4730. 'return "done"',
  4731. ].join('\n'),
  4732. bindings: [],
  4733. })
  4734. expect(result.error).toBeUndefined()
  4735. expect(result.value).toBe('done')
  4736. expect(result.logs).toEqual([logTruncationMarker(256)])
  4737. }, 40_000)
  4738. it('bounds a newline-terminated write against the already-buffered chunks before joining them', async () => {
  4739. // Same allocation, reached through the newline arm: with chunks pending, the
  4740. // whole write used to be appended and joined so the offset scan could run
  4741. // over one string. Only the FIRST line needs those chunks, so a pending
  4742. // chunk plus a 340 MiB newline-terminated write under a 400 MiB
  4743. // addressSpaceMb died on MemoryError in the join before the per-line bound
  4744. // could reject anything, and the retained chunks made the settlement flush
  4745. // die the same way: measured, `timeout: wall-clock ceiling reached
  4746. // (30000ms)`. The reconstructed first line is now checked against the ledger
  4747. // and only a budget-sized prefix of it is copied; the rest of the write is
  4748. // scanned in place.
  4749. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 400, maxWallMs: 30_000 })
  4750. const result = await runtime.run({
  4751. program: [
  4752. 'import sys',
  4753. 'sys.stdout.write("x")',
  4754. 'payload = "\\n".rjust(340 * 1024 * 1024, "A")',
  4755. 'sys.stdout.write(payload)',
  4756. 'return "done"',
  4757. ].join('\n'),
  4758. bindings: [],
  4759. })
  4760. expect(result.error).toBeUndefined()
  4761. expect(result.value).toBe('done')
  4762. expect(result.logs).toEqual([logTruncationMarker(256)])
  4763. }, 40_000)
  4764. it('emits pending text on an explicit flush, before the run can be killed', async () => {
  4765. // `_LogStream` inherits TextIOBase's no-op `flush()`, so an explicit
  4766. // `print(..., flush=True)` or `sys.stdout.flush()` left the text in
  4767. // `_pending` with nothing to drain it but `flush_line` after settlement — a
  4768. // call a hanging or killed run never reaches. Measured: printing
  4769. // "before hang" with flush=True ahead of an infinite loop returned
  4770. // `logs: []`, losing the one diagnostic the program deliberately committed.
  4771. const { runtime } = await setup({ maxWallMs: 4_000 })
  4772. const result = await runtime.run({
  4773. program: [
  4774. 'import sys',
  4775. 'print("before hang", end="", flush=True)',
  4776. 'while True: pass',
  4777. ].join('\n'),
  4778. bindings: [],
  4779. })
  4780. expect(result.error?.kind).toBe('timeout')
  4781. expect(result.logs).toContain('before hang')
  4782. }, 15_000)
  4783. it('marks a dropped tail when the ledger lands on exactly zero remaining', async () => {
  4784. // One 100-character line costs 103 serialized bytes (quotes + separator),
  4785. // consuming a 104-byte budget minus the 1-byte array-envelope reservation
  4786. // (104 - 1 = 103) EXACTLY. Landing on zero never trips
  4787. // LogBuffer's "cost > remaining" branch, so `_truncated` stays unset and the
  4788. // stream's own `remaining > 0` guard silently discarded the unscanned tail —
  4789. // the run reported a complete log while dropping text. The tail must be
  4790. // pushed so the marker is emitted. (This surfaced only after empty writes
  4791. // stopped being buffered: `print` issues a trailing `write("")` whose
  4792. // buffered-empty path used to force the marker out incidentally.) A single
  4793. // wide line is used rather than many narrow ones so the CHILD ledger is the
  4794. // one that lands on zero: the host's identical ledger truncates first when
  4795. // many small entries precede the long marker text. `["y"*100]` serializes to
  4796. // exactly 104 bytes (103 payload + 1 envelope), so 104 is the smallest
  4797. // budget that admits the entry.
  4798. const { runtime } = await setup({ maxLogBytes: 104, maxWallMs: 10_000 })
  4799. const result = await runtime.run({
  4800. program: ['print("y" * 100 + "\\n" + "z" * 10, end="")', 'return "done"'].join('\n'),
  4801. bindings: [],
  4802. })
  4803. expect(result.error).toBeUndefined()
  4804. expect(result.value).toBe('done')
  4805. expect(result.logs).toContain('y'.repeat(100))
  4806. expect(result.logs.filter(line => line.includes('log capture truncated'))).toHaveLength(1)
  4807. // The dropped tail is not retained, but its loss is now reported.
  4808. expect(result.logs.some(line => line.includes('z'))).toBe(false)
  4809. }, 15_000)
  4810. it('keeps an admitted log within the serialized array envelope at the exact limit', async () => {
  4811. // Each entry is charged its JSON-string cost plus one separator byte, and
  4812. // the serialized outer logs array adds one more byte of envelope (two
  4813. // brackets and n-1 commas). The ledgers reserve that byte, so a result that
  4814. // exactly exhausts the ledger still serializes within the configured cap.
  4815. // At the 64-byte floor (the smallest admissible maxLogBytes): ledger 63,
  4816. // a 60-character line serializes as `"aaa...a"` (62 bytes) + 1 separator
  4817. // = 63, exactly exhausting the ledger and serializing as `["aaa...a"]`
  4818. // = 64 = the cap; a 61-character line costs 64 > 63 and truncates. The
  4819. // marker rides envelope, so the serialized logs run to cap + marker.
  4820. const { runtime } = await setup({ maxLogBytes: 64, maxWallMs: 10_000 })
  4821. const result = await runtime.run({
  4822. program: ['print("a" * 60 + "\\n" + "b" * 61, end="")', 'return "done"'].join('\n'),
  4823. bindings: [],
  4824. })
  4825. expect(result.error).toBeUndefined()
  4826. expect(result.value).toBe('done')
  4827. // The 60-character line was admitted; the 61-character line was not (a
  4828. // single 'b' would also match the marker's "bytes", so check for the line).
  4829. expect(result.logs).toContain('a'.repeat(60))
  4830. expect(result.logs.some(line => line.includes('b'.repeat(61)))).toBe(false)
  4831. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  4832. }, 15_000)
  4833. it('rejects a log budget too small to serialize the truncation marker', async () => {
  4834. // A maxLogBytes below 64 cannot serialize the truncation marker itself;
  4835. // it is rejected at construction so a marker-only truncated run cannot
  4836. // report more than the public cap. maxValueBytes keeps no floor beyond the
  4837. // positive-integer requirement (a completion can be 1 byte).
  4838. await expect(setup({ maxLogBytes: 63, maxWallMs: 10_000 })).rejects.toThrow(/must be at least 64/)
  4839. }, 15_000)
  4840. it('charges the JSON-escaped cost of control characters against the log ledger', async () => {
  4841. // A NUL renders as \u0000 (6 bytes) in the serialized outer logs; the
  4842. // ledger must charge that expansion, or a control-character flood admits
  4843. // 6x the configured cap.
  4844. const { runtime } = await setup({ maxLogBytes: 256 })
  4845. const result = await runtime.run({
  4846. program: [
  4847. 'for _ in range(500):',
  4848. ' print("\\x00" * 10)',
  4849. 'return "done"',
  4850. ].join('\n'),
  4851. bindings: [],
  4852. })
  4853. expect(result.error).toBeUndefined()
  4854. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  4855. // Serialized (escaped) size of retained entries stays in the budget's
  4856. // neighborhood: well under the ~30 kB an uncharged flood would retain.
  4857. const serialized = Buffer.byteLength(JSON.stringify(result.logs), 'utf8')
  4858. expect(serialized).toBeLessThan(1024)
  4859. })
  4860. it('charges the serialized cost child-side, so a control-heavy line truncates instead of being admitted whole', async () => {
  4861. // The child's ledger must charge what the entry costs on the wire, not its
  4862. // raw UTF-8 length: a NUL is one raw byte but six as its escape. A 24 MiB NUL
  4863. // line clears the cheap char-count lower bound (24 MiB < 32 MiB budget), so
  4864. // charging raw bytes would ADMIT it and emit a ~144 MiB escaped entry;
  4865. // charging the serialized cost (~144 MiB > the 32 MiB budget) rejects it
  4866. // before any encode and emits the marker instead. The address space (512 MiB,
  4867. // clearing the 12x load gate for a 32 MiB budget) is sized so the run loads;
  4868. // the gate separately guarantees a correctly-charged near-budget entry fits.
  4869. const { runtime } = await setup({ maxLogBytes: 32 * 1024 * 1024, addressSpaceMb: 512, maxWallMs: 20_000 })
  4870. const result = await runtime.run({
  4871. program: [
  4872. 'print("\\x00" * (24 * 1024 * 1024))',
  4873. 'return "done"',
  4874. ].join('\n'),
  4875. bindings: [],
  4876. })
  4877. expect(result.error).toBeUndefined()
  4878. expect(result.value).toBe('done')
  4879. expect(result.logs.filter(line => line.includes('log capture truncated'))).toHaveLength(1)
  4880. // Nothing of the line itself was retained: the ledger refused the whole entry.
  4881. expect(result.logs.every(line => !line.includes(String.fromCharCode(0)))).toBe(true)
  4882. }, 30_000)
  4883. it('bounds a huge unterminated tail after an early newline without copying it whole', async () => {
  4884. // The newline branch of _LogStream.write buffered the whole unterminated
  4885. // tail after the last newline into `_pending` before the flush trigger could
  4886. // bound it, so an early newline followed by a huge tail made a second full
  4887. // copy of the model's own string — a MemoryError the config gate cannot
  4888. // catch (the tail far exceeds maxLogBytes). The tail is now sliced to a
  4889. // budget-sized prefix, so the run truncates and completes. Linux-only RLIMIT_AS
  4890. // repro (Darwin skips the limit); on macOS this asserts the happy path.
  4891. //
  4892. // Sizing: the model builds `tail` (N) then the `"\n" + tail` write argument
  4893. // (another ~N), so construction peaks at ~2N — kept under the 384 MiB address
  4894. // space at N = 150 MiB (~300 MiB). The pre-fix code then buffered the whole
  4895. // ~150 MiB tail again, pushing past 384 MiB; the sliced prefix does not.
  4896. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 384, maxWallMs: 20_000 })
  4897. const result = await runtime.run({
  4898. program: [
  4899. 'import sys',
  4900. 'tail = "A" * (150 * 1024 * 1024)',
  4901. 'sys.stdout.write("\\n" + tail)',
  4902. 'return "done"',
  4903. ].join('\n'),
  4904. bindings: [],
  4905. })
  4906. expect(result.error).toBeUndefined()
  4907. expect(result.value).toBe('done')
  4908. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  4909. }, 30_000)
  4910. it('flushes logs before framing the value so their peaks do not add against RLIMIT_AS', async () => {
  4911. // The load gate bounds maxLogBytes and maxValueBytes INDEPENDENTLY against the
  4912. // address space, each at the 12x worst case. But the child framed the
  4913. // completion value (materializing its escaped form to meter it, then encoding
  4914. // the frame) while a newline-free log tail still sat unflushed in _pending.
  4915. // Those two peaks added: two budgets each admitted alone could together breach
  4916. // RLIMIT_AS, dying as worker-exit instead of settling. The flush now runs
  4917. // before the value is framed, so the log pending is freed first.
  4918. //
  4919. // Config: 32 MiB each against 512 MiB (each 32*12 = 384 MiB < 448 MiB
  4920. // budgetable, so both load). The program writes ~33M astral chars with no
  4921. // newline (buffered ~132 MB, under the char-count flush trigger) then returns
  4922. // ~33M astral chars — a ~132 MB serialized value that is itself OVER the 32 MiB
  4923. // maxValueBytes, so the correct outcome is `output-limit`. Pre-fix the
  4924. // unflushed 132 MB plus the value's build-and-encode (~396 MB) exceeded 512 MiB
  4925. // and OOM'd (reported as exception/worker-exit); flushing first lets the value
  4926. // check complete (~460 MB alone) and report output-limit. On Darwin (no
  4927. // RLIMIT_AS) the value is over budget too, so output-limit holds either way;
  4928. // the OOM the reorder prevents is the Linux-only failure.
  4929. const { runtime } = await setup({
  4930. maxLogBytes: 32 * 1024 * 1024,
  4931. maxValueBytes: 32 * 1024 * 1024,
  4932. addressSpaceMb: 512,
  4933. maxWallMs: 20_000,
  4934. })
  4935. const result = await runtime.run({
  4936. program: [
  4937. 'import sys',
  4938. 'sys.stdout.write("\\U0001F600" * 33_000_000)',
  4939. 'return "\\U0001F600" * 33_000_000',
  4940. ].join('\n'),
  4941. bindings: [],
  4942. })
  4943. expect(result.error?.kind).toBe('output-limit')
  4944. }, 30_000)
  4945. it('checks and encodes a wide completion value in O(depth), not O(width)', async () => {
  4946. // A wide flat list serializes to ~2 bytes per element but the pre-fix walk
  4947. // enqueued one traversal tuple per element (_check_done_value) and one stack
  4948. // entry plus a separator marker per element (_encode_json_plain) — ~56 bytes
  4949. // per element, ~28x the serialized size. A value the byte meter admits could
  4950. // therefore OOM on the checker's or encoder's own bookkeeping, the inversion
  4951. // the load gate exists to prevent (the gate reserves 12x, not 28x). Both now
  4952. // walk with an O(depth) cursor that pulls one child at a time, so the only
  4953. // width-proportional allocation is the output string the meter bounded.
  4954. //
  4955. // Config: maxValueBytes 20 MiB against 384 MiB (20*12 = 240 MiB < 320 MiB
  4956. // budgetable, so it loads). `[0] * 6_000_000` is ~12 MB of JSON, under the
  4957. // 20 MiB budget, so it must round-trip. Pre-fix the ~400 MB of per-element
  4958. // frames plus the interpreter exceeded 384 MiB and returned MemoryError as an
  4959. // exception. Linux-only RLIMIT_AS repro; on macOS the value round-trips
  4960. // either way, but the fixture stays within the address space so it is honest.
  4961. //
  4962. // `maxWallMs` is 60s, not the 20s the memory assertion alone needs: the O(depth)
  4963. // cursor pulls 6M elements one at a time through Python-level frames, which costs
  4964. // ~11s on an idle machine and more under the coverage lane's V8 instrumentation
  4965. // with several workers sharing a box. This budget bounds the run without letting a
  4966. // loaded runner's scheduling latency read as a `timeout` — what this test asserts
  4967. // is the O(depth) memory shape, not a speed claim.
  4968. const { runtime } = await setup({ maxValueBytes: 20 * 1024 * 1024, addressSpaceMb: 384, maxWallMs: 60_000 })
  4969. const result = await runtime.run({ program: 'return [0] * 6_000_000', bindings: [] })
  4970. expect(result.error).toBeUndefined()
  4971. expect(Array.isArray(result.value)).toBe(true)
  4972. expect((result.value as number[]).length).toBe(6_000_000)
  4973. }, 90_000)
  4974. it('validates wide binding arguments in O(depth), not O(width)', async () => {
  4975. // The completion-value walks are budgeted; this one is not. `dispatch` runs
  4976. // `_lossless_json_violation` on the arguments the MODEL built, and no
  4977. // child-side byte budget bounds them first: the frame ceiling is the host's
  4978. // and applies only after this validation returns. A per-member traversal
  4979. // frame therefore turned a legitimate call into the program's own
  4980. // MemoryError. Measured with tracemalloc on the two walk shapes over this
  4981. // exact argument (JSON ~17 MB): the cursor peaks at 0.0 MiB of auxiliary
  4982. // state, the pre-fix `stack.extend` at 459.1 MiB -- past the 384 MiB
  4983. // configured below, so the discriminating failure is real. It is Linux-only:
  4984. // Darwin skips RLIMIT_AS, so this case round-trips there either way.
  4985. //
  4986. // The binding echoes its argument's length back, so the assertion proves the
  4987. // call actually round-tripped rather than merely avoiding a crash.
  4988. const { runtime } = await setup({ addressSpaceMb: 384, maxWallMs: 60_000 })
  4989. const result = await runtime.run({
  4990. program: 'return await tools.width([0] * 6_000_000)',
  4991. bindings: [{
  4992. global: 'tools',
  4993. functions: { width: async (items: unknown) => (items as number[]).length },
  4994. }],
  4995. })
  4996. expect(result.error).toBeUndefined()
  4997. expect(result.value).toBe(6_000_000)
  4998. }, 90_000)
  4999. it('decodes a multi-megabyte binding reply without regex backtracking state', async () => {
  5000. // The child parses every host reply with `_decode_json_plain`. Its scalar
  5001. // regex matched strings with a `(?:[^"\\]|\\.)*` repetition, which makes
  5002. // CPython's backtracking engine retain state proportional to the string's
  5003. // WIDTH -- measured at ~146 MiB of engine state for a 1 MiB string and
  5004. // ~558 MiB for 4 MiB. A legitimate multi-megabyte reply therefore raised
  5005. // MemoryError inside `_pump_replies`; because that pump is the only settler
  5006. // of the call's future, the run stranded until the wall clock reported a
  5007. // `timeout` instead of returning the value the binding produced.
  5008. //
  5009. // Strings now scan chunk-to-chunk over a character class (no backtracking
  5010. // state). Measured on this exact 4 MiB reply: the pre-fix regex peaks at
  5011. // 557.8 MiB, past the default 512 MiB address space, while the scanner peaks
  5012. // at the 4.0 MiB result itself. Linux-only, like the other RLIMIT_AS repros:
  5013. // Darwin does not apply the limit, so the spike is merely allocated there.
  5014. const reply = 'A'.repeat(4 * 1024 * 1024)
  5015. const { runtime } = await setup({ maxWallMs: 60_000 })
  5016. const result = await runtime.run({
  5017. program: 'value = await tools.big({})\nreturn len(value)',
  5018. bindings: [{ global: 'tools', functions: { big: async () => reply } }],
  5019. })
  5020. expect(result.error).toBeUndefined()
  5021. expect(result.value).toBe(reply.length)
  5022. }, 90_000)
  5023. it('drops a late binding resolution before snapshotting it', async () => {
  5024. // `sendReply` checks `settled`, but only after the resolution has been walked
  5025. // and copied by `snapshotJsonValue`. Binding resolution carries no seam-level
  5026. // byte cap, so a binding that resolves a wide value AFTER the run already
  5027. // settled (here on `maxWallMs`) spent host heap building a frame that is then
  5028. // discarded. The check now runs before the snapshot.
  5029. //
  5030. // The binding resolves well after the 1s wall clock with a 2M-element array;
  5031. // the run must still report `timeout`, and the late value must not appear.
  5032. let resolvedLate = false
  5033. const { runtime } = await setup({ maxWallMs: 1_000 })
  5034. const result = await runtime.run({
  5035. program: 'return await tools.slow({})',
  5036. bindings: [{
  5037. global: 'tools',
  5038. functions: {
  5039. slow: async () => {
  5040. await new Promise(resolve => setTimeout(resolve, 2_500))
  5041. resolvedLate = true
  5042. return Array.from({ length: 2_000_000 }, () => 0)
  5043. },
  5044. },
  5045. }],
  5046. })
  5047. expect(result.error?.kind).toBe('timeout')
  5048. expect(result.value).toBeUndefined()
  5049. // Pin that the late path actually ran, so the assertion above is not vacuous.
  5050. await new Promise(resolve => setTimeout(resolve, 2_000))
  5051. expect(resolvedLate).toBe(true)
  5052. }, 90_000)
  5053. it('paces concurrent binding replies instead of queueing every frame at once', async () => {
  5054. // Binding resolution carries no seam-level byte cap. Before pacing, a program
  5055. // resolving several large values in one `asyncio.gather` round encoded them
  5056. // all in the same turn and queued every frame in fd 3's writable buffer,
  5057. // which exhausted the host heap and killed the whole process rather than
  5058. // failing the run. Replies are now encoded one at a time, waiting for
  5059. // `drain` when the pipe is full.
  5060. //
  5061. // Eight concurrent 4 MiB replies (32 MiB of frames) must all round-trip. The
  5062. // program sums the lengths, so the assertion proves every reply arrived and
  5063. // was matched to its own call -- pacing must not drop or misroute any. What
  5064. // this case cannot show is the peak itself, which lives in the stream's
  5065. // buffer: measured directly on a 64 KiB-highWaterMark pipe with this same
  5066. // 8x4 MiB shape, the unpaced writes buffered 32.0 MiB while the paced ones
  5067. // peaked at 0.0 MiB.
  5068. const chunk = 'A'.repeat(4 * 1024 * 1024)
  5069. const { runtime } = await setup({ maxWallMs: 60_000 })
  5070. const result = await runtime.run({
  5071. program: [
  5072. 'import asyncio',
  5073. 'parts = await asyncio.gather(*[tools.chunk({}) for _ in range(8)])',
  5074. 'return sum(len(p) for p in parts)',
  5075. ].join('\n'),
  5076. bindings: [{ global: 'tools', functions: { chunk: async () => chunk } }],
  5077. })
  5078. expect(result.error).toBeUndefined()
  5079. expect(result.value).toBe(8 * chunk.length)
  5080. }, 90_000)
  5081. it('drops queued binding replies when the child dies mid-drain, without hanging', async () => {
  5082. // drainReplies waits for `drain` when fd 3's buffer is full. If the child
  5083. // exits while a reply is queued, the pipe never emits `drain` again — the
  5084. // wait must also settle on `close`/`error`/destroyed, or `draining` stays
  5085. // true and the queue is pinned with the closure forever. The program fills
  5086. // the pipe with a wide binding reply and then exits without reading it, so
  5087. // the host is blocked mid-drain when the child dies; the run must still
  5088. // settle promptly (worker-exit from the close) rather than hanging on the
  5089. // drain wait.
  5090. const chunk = 'A'.repeat(4 * 1024 * 1024)
  5091. const { runtime } = await setup({ maxWallMs: 10_000 })
  5092. const result = await runtime.run({
  5093. program: [
  5094. 'import asyncio',
  5095. // Resolve a reply big enough to backpressure fd 3, then exit without
  5096. // reading it: the child's `close` lands while the host still waits for
  5097. // `drain`, exercising the destroyed-pipe branch of the reply drain.
  5098. 'pending = asyncio.create_task(tools.chunk({}))',
  5099. 'await asyncio.sleep(0.05)',
  5100. 'return "done"',
  5101. ].join('\n'),
  5102. bindings: [{ global: 'tools', functions: { chunk: async () => chunk } }],
  5103. })
  5104. // The program returned, so the completion wins over the mid-flight reply;
  5105. // whatever the result, the run must settle (no hang on the drain wait).
  5106. expect(result.error).toBeUndefined()
  5107. expect(result.value).toBe('done')
  5108. }, 30_000)
  5109. it('caps the pending reply backlog when a child floods calls without reading its replies', async () => {
  5110. // drainReplies writes one reply at a time and waits for `drain` when fd 3's
  5111. // buffer is full. A child that never reads its replies (it only writes
  5112. // call frames, never draining the reply side) leaves the pipe full, so
  5113. // every call frame it keeps sending resolves a binding and adds a reply the
  5114. // drain cannot write: without a bound, the backlog grows until the wall
  5115. // clock, pinning each binding result in host memory. The cap settles the
  5116. // run as worker-exit instead, mirroring the frame cap's treatment of an
  5117. // oversized frame. The child floods 5000 sequential valid calls and never
  5118. // reads fd 3 (its reply pump is starved by the synchronous write loop and
  5119. // the blocking sleep); the pipe buffer absorbs ~1600 tiny replies, so the
  5120. // pending backlog crosses MAX_PENDING_REPLIES long before maxWallMs, and
  5121. // the run must settle worker-exit with the reply-queue message, not a
  5122. // wall-clock timeout.
  5123. const { runtime } = await setup({ maxWallMs: 30_000 })
  5124. const result = await runtime.run({
  5125. program: [
  5126. 'import os, time',
  5127. 'frame = b\'{"type":"call","id":%d,"global":"tools","name":"echo","args":{}}\\n\'',
  5128. 'for i in range(5000):',
  5129. ' view = memoryview(frame % i)',
  5130. ' while view:',
  5131. ' view = view[os.write(3, view):]',
  5132. // Keep the child alive without reading fd 3: the run must settle via
  5133. // the reply-backlog cap, not by the child finishing or exiting.
  5134. 'time.sleep(30)',
  5135. 'return "unreachable"',
  5136. ].join('\n'),
  5137. bindings: [{ global: 'tools', functions: { echo: async (args: unknown) => args as CodeJsonValue } }],
  5138. })
  5139. expect(result.error?.kind).toBe('worker-exit')
  5140. expect(result.error?.message).toContain('reply queue exceeded')
  5141. }, 30_000)
  5142. it('caps the outstanding binding-call backlog when a child floods calls against a binding that never settles', async () => {
  5143. // The reply backlog cap only counts RESOLVED calls (`pendingReplies` grows
  5144. // after the await), so a child flooding calls against a binding whose
  5145. // promise never settles would accumulate one async closure per frame until
  5146. // the wall clock without tripping it. The outstanding-call counter bounds
  5147. // the in-flight closures to MAX_PENDING_REPLIES and settles the run as
  5148. // worker-exit, mirroring the reply cap. The binding below never resolves,
  5149. // so no reply is ever produced; the flood of 5000 sequential calls must
  5150. // cross the in-flight bound long before maxWallMs.
  5151. const { runtime } = await setup({ maxWallMs: 30_000 })
  5152. const result = await runtime.run({
  5153. program: [
  5154. 'import os, time',
  5155. 'frame = b\'{"type":"call","id":%d,"global":"tools","name":"hang","args":{}}\\n\'',
  5156. 'for i in range(5000):',
  5157. ' view = memoryview(frame % i)',
  5158. ' while view:',
  5159. ' view = view[os.write(3, view):]',
  5160. 'time.sleep(30)',
  5161. 'return "unreachable"',
  5162. ].join('\n'),
  5163. bindings: [{ global: 'tools', functions: { hang: async () => await new Promise<never>(() => {}) } }],
  5164. })
  5165. expect(result.error?.kind).toBe('worker-exit')
  5166. expect(result.error?.message).toContain('call backlog exceeded')
  5167. }, 30_000)
  5168. it('runs a legitimate gather of more than 1024 concurrent binding calls', async () => {
  5169. // The in-flight call cap must not count a synchronous batch of instant
  5170. // calls: the async bodies' finallys run on the microtask queue, which
  5171. // drains only between 'data' events, so a per-frame check would trip on
  5172. // the 1025th frame of a single event even though every binding settled
  5173. // immediately — killing a valid large concurrent gather as worker-exit.
  5174. // The cap is checked at event boundaries (after the microtask queue
  5175. // drained), so this gather of 1025 instant calls completes.
  5176. const { runtime } = await setup({ maxWallMs: 30_000 })
  5177. const result = await runtime.run({
  5178. program: [
  5179. 'import asyncio',
  5180. 'return len(await asyncio.gather(*[tools.echo(i) for i in range(1025)]))',
  5181. ].join('\n'),
  5182. bindings: [{ global: 'tools', functions: { echo: async (args: unknown) => args as CodeJsonValue } }],
  5183. })
  5184. expect(result.error).toBeUndefined()
  5185. expect(result.value).toBe(1025)
  5186. }, 30_000)
  5187. it('completes normally when a program returns with binding calls still outstanding', async () => {
  5188. // The in-flight call cap refuses to admit NEW calls past the bound; it must
  5189. // not reclassify a `done` frame as worker-exit just because the program
  5190. // returned with calls it started but never awaited. The child schedules
  5191. // exactly 1024 slow bindings (still pending when the program returns), so
  5192. // the done frame arrives with the outstanding count AT the cap — the event
  5193. // must complete with its value, not settle as `call backlog exceeded`.
  5194. const { runtime } = await setup({ maxWallMs: 30_000 })
  5195. const result = await runtime.run({
  5196. program: [
  5197. 'import asyncio',
  5198. 'for i in range(1024):',
  5199. ' asyncio.create_task(tools.slow(i))',
  5200. 'await asyncio.sleep(0.2)',
  5201. 'return "done"',
  5202. ].join('\n'),
  5203. bindings: [{
  5204. global: 'tools',
  5205. functions: { slow: async () => { await new Promise((resolve) => { setTimeout(resolve, 5_000) }); return 1 } },
  5206. }],
  5207. })
  5208. expect(result.error).toBeUndefined()
  5209. expect(result.value).toBe('done')
  5210. }, 30_000)
  5211. it('settles a single-batch never-settling flood as worker-exit without further frames', async () => {
  5212. // The outstanding-call cap must take effect even when the whole flood fits
  5213. // in ONE data event: a per-event admission snapshot never re-checks once no
  5214. // further frames arrive, so a single 62 KiB write of 1025 compact calls
  5215. // against a never-settling binding would otherwise wait out the full wall
  5216. // clock instead of tripping the cap. The post-macrotask check runs after
  5217. // the batch's finallys (which never run for this binding) and settles the
  5218. // run as worker-exit long before maxWallMs.
  5219. const { runtime } = await setup({ maxWallMs: 30_000 })
  5220. const result = await runtime.run({
  5221. program: [
  5222. 'import os, time',
  5223. 'frame = b\'{"type":"call","id":%d,"global":"tools","name":"hang","args":{}}\\n\'',
  5224. 'payload = b"".join(frame % i for i in range(1025))',
  5225. 'view = memoryview(payload)',
  5226. 'while view:',
  5227. ' view = view[os.write(3, view):]',
  5228. 'time.sleep(30)',
  5229. 'return "unreachable"',
  5230. ].join('\n'),
  5231. bindings: [{ global: 'tools', functions: { hang: async () => await new Promise<never>(() => {}) } }],
  5232. })
  5233. expect(result.error?.kind).toBe('worker-exit')
  5234. expect(result.error?.message).toContain('call backlog exceeded')
  5235. }, 30_000)
  5236. it('runs a burst of 1300 instant calls whose frames split across pipe reads', async () => {
  5237. // Flowing mode can fire several 'data' events within one macrotask, before
  5238. // any microtask drains, so a per-event snapshot of the outstanding count
  5239. // could see the first chunk's in-flight calls in the second chunk's check
  5240. // and false-positive on a legitimate burst. The post-macrotask check always
  5241. // sees the true count (all finallys have run), so this burst of compact
  5242. // frames — sized so the pipe read splits it — completes with all results.
  5243. const { runtime } = await setup({ maxWallMs: 30_000 })
  5244. const result = await runtime.run({
  5245. program: [
  5246. 'import asyncio',
  5247. 'return len(await asyncio.gather(*[t.e(i) for i in range(1300)]))',
  5248. ].join('\n'),
  5249. bindings: [{ global: 't', functions: { e: async (args: unknown) => args as CodeJsonValue } }],
  5250. })
  5251. expect(result.error).toBeUndefined()
  5252. expect(result.value).toBe(1300)
  5253. }, 30_000)
  5254. it('settles as worker-exit when a done frame lands in the same batch as a call flood', async () => {
  5255. // A done frame processed in the SAME data event as more than 1024 call
  5256. // frames settles the run before the post-macrotask check runs (which no-ops
  5257. // once settled), so a child could finish "successfully" while leaving the
  5258. // outstanding closures behind — one sub-64 KiB write carries 1025 compact
  5259. // calls plus a done. The done handler re-checks the count before accepting
  5260. // the frame, so the run settles as worker-exit with the call-backlog
  5261. // message instead.
  5262. const { runtime } = await setup({ maxWallMs: 30_000 })
  5263. const result = await runtime.run({
  5264. program: [
  5265. 'import os, time',
  5266. 'frame = b\'{"type":"call","id":%d,"global":"tools","name":"hang","args":{}}\\n\'',
  5267. 'payload = b"".join(frame % i for i in range(1025)) + b\'{"type":"done","value":1}\\n\'',
  5268. 'view = memoryview(payload)',
  5269. 'while view:',
  5270. ' view = view[os.write(3, view):]',
  5271. 'time.sleep(30)',
  5272. 'return "unreachable"',
  5273. ].join('\n'),
  5274. bindings: [{ global: 'tools', functions: { hang: async () => await new Promise<never>(() => {}) } }],
  5275. })
  5276. expect(result.error?.kind).toBe('worker-exit')
  5277. expect(result.error?.message).toContain('call backlog exceeded')
  5278. }, 30_000)
  5279. it('rejects a completion whose dict keys fold to one JSON member', async () => {
  5280. // `_dump_string` folds a spelled-out surrogate pair into its astral code
  5281. // point, so `"\ud83d\ude00"` and `"\U0001f600"` are DIFFERENT Python keys
  5282. // that encode to the SAME JSON member — the host's JSON.parse would
  5283. // silently drop one of them, violating the lossless-JSON completion
  5284. // contract. The child's meter rejects the collision before encoding.
  5285. const { runtime } = await setup()
  5286. const result = await runtime.run({
  5287. program: 'return {"\\ud83d\\ude00": 1, "\\U0001f600": 2}',
  5288. bindings: [],
  5289. })
  5290. expect(result.error?.kind).toBe('invalid-output')
  5291. expect(result.error?.message).toContain('duplicate dict key')
  5292. }, 30_000)
  5293. it('rejects binding arguments whose dict keys fold to one JSON member', async () => {
  5294. // The same collision on the binding-argument path: the call is rejected as
  5295. // not lossless JSON, so the program's `await` raises and the program
  5296. // surfaces the rejection message.
  5297. const { runtime } = await setup()
  5298. const result = await runtime.run({
  5299. program: [
  5300. 'try:',
  5301. ' await tools.echo({"\\ud83d\\ude00": 1, "\\U0001f600": 2})',
  5302. ' return "no-error"',
  5303. 'except Exception as e:',
  5304. ' return str(e)',
  5305. ].join('\n'),
  5306. bindings: [{ global: 'tools', functions: { echo: async (args: unknown) => args as CodeJsonValue } }],
  5307. })
  5308. expect(result.error).toBeUndefined()
  5309. expect(result.value).toContain('duplicate dict key')
  5310. }, 30_000)
  5311. it('compacts the reply queue mid-drain without dropping pending frames', async () => {
  5312. // A reply larger than the writable high-water mark makes the FIRST write
  5313. // return false, suspending the drain loop; the frames queued behind it
  5314. // push the drain's consumed head past MAX_PENDING_REPLIES, so the resumed
  5315. // drain compacts the queue mid-run. The child reads fd 3 itself (blocking
  5316. // the asyncio pump, so its reads cannot race the host's pushes) and sends
  5317. // a second wave of calls AFTER reading part of the first wave's replies —
  5318. // those replies are still pending when the drain's head crosses the
  5319. // compaction bound, so a compaction that dropped pending frames would
  5320. // leave the child's reply count short and the read loop spinning to the
  5321. // wall clock. No fixed sleep: the child's reads pace at the drain's
  5322. // delivery rate (each write blocks until the child reads), and the host
  5323. // finishes pushing all of a wave within milliseconds — orders of magnitude
  5324. // before the head crosses the bound — so the queue is always full at the
  5325. // splice. Newlines are counted per chunk (each reply carries exactly one),
  5326. // never by re-scanning the accumulated total, which would be O(n²).
  5327. const { runtime } = await setup({ maxWallMs: 60_000 })
  5328. const result = await runtime.run({
  5329. program: [
  5330. 'import os',
  5331. 'frame = b\'{"type":"call","id":%d,"global":"tools","name":"big","args":{}}\\n\'',
  5332. 'for i in range(1024):',
  5333. ' view = memoryview(frame % i)',
  5334. ' while view:',
  5335. ' view = view[os.write(3, view):]',
  5336. 'seen = 0',
  5337. 'while seen < 500:',
  5338. ' chunk = os.read(3, 65536)',
  5339. ' if not chunk:',
  5340. ' break',
  5341. ' seen += chunk.count(b"\\n")',
  5342. 'for i in range(500):',
  5343. ' view = memoryview(frame % (1024 + i))',
  5344. ' while view:',
  5345. ' view = view[os.write(3, view):]',
  5346. 'while seen < 1524:',
  5347. ' chunk = os.read(3, 65536)',
  5348. ' if not chunk:',
  5349. ' break',
  5350. ' seen += chunk.count(b"\\n")',
  5351. 'return "done"',
  5352. ].join('\n'),
  5353. bindings: [{ global: 'tools', functions: { big: async () => 'x'.repeat(65 * 1024) } }],
  5354. })
  5355. expect(result.error).toBeUndefined()
  5356. expect(result.value).toBe('done')
  5357. }, 60_000)
  5358. it('bounds a flood of zero-byte log lines through the per-entry separator charge', async () => {
  5359. // Blank print() lines carry zero content bytes; without the +1 separator
  5360. // charge they would bypass maxLogBytes entirely and grow the retained
  5361. // array without bound. Each empty entry costs one byte, so a 64-byte
  5362. // budget retains at most 64 entries before the marker.
  5363. const { runtime } = await setup({ maxLogBytes: 64, maxWallMs: 10_000 })
  5364. const result = await runtime.run({
  5365. program: [
  5366. 'for _ in range(10000):',
  5367. ' print()',
  5368. 'return "done"',
  5369. ].join('\n'),
  5370. bindings: [],
  5371. })
  5372. expect(result.error).toBeUndefined()
  5373. expect(result.logs.length).toBeLessThanOrEqual(65)
  5374. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  5375. })
  5376. it('reassembles multibyte UTF-8 split across stray-output pipe chunks', async () => {
  5377. // A single os.write far past the 64 KiB pipe buffer forces multiple
  5378. // 'data' chunks; when the boundary lands inside a multibyte sequence,
  5379. // per-chunk decoding would corrupt it into replacement characters. Raw bytes
  5380. // are buffered and only decoded once a complete line (or the whole tail at
  5381. // flush) is assembled, so the split sequence is whole by the time it is
  5382. // decoded. The payload spans every valid multibyte lead class so
  5383. // accrueStrayCost's per-lead continuation ranges are all exercised: U+0900
  5384. // (E0 A4 80, the range-restricted E0 lead), U+4F60 and U+597D (E4/E5, plain
  5385. // 3-byte), U+1F600 (F0, the range-restricted F0 lead), and U+10FFFF (F4 8F
  5386. // BF BF, the range-restricted F4 lead).
  5387. const { runtime } = await setup({ maxLogBytes: 1024 * 1024 })
  5388. const result = await runtime.run({
  5389. program: [
  5390. 'import os',
  5391. // os.write is one syscall and returns a partial count on a full
  5392. // pipe, so loop until the whole payload (odd prefix -> a chunk
  5393. // boundary lands inside a multibyte sequence) is out.
  5394. String.raw`payload = b"a" * 65535 + "\u0900\u4f60\u597d\U0001f600\U0010ffff".encode("utf-8")`,
  5395. 'view = memoryview(payload)',
  5396. 'while view:',
  5397. ' view = view[os.write(1, view):]',
  5398. 'return "done"',
  5399. ].join('\n'),
  5400. bindings: [],
  5401. })
  5402. expect(result.error).toBeUndefined()
  5403. const text = result.logs.join('')
  5404. expect(text).toContain('\u0900\u4f60\u597d\u{1f600}\u{10ffff}')
  5405. expect(text).not.toContain('\ufffd')
  5406. })
  5407. it('flushes a stray-output byte sequence left incomplete when the pipe ends', async () => {
  5408. // The child writes the first two bytes of a 3-byte UTF-8 character to fd 1
  5409. // and exits, so the pipe closes with the sequence unfinished in the raw
  5410. // residual. The 'end' flush decodes the residual with `toString('utf8')`,
  5411. // which renders the stranded bytes as U+FFFD instead of dropping them.
  5412. const { runtime } = await setup({ maxLogBytes: 1024 * 1024 })
  5413. const result = await runtime.run({
  5414. program: [
  5415. 'import os',
  5416. // b"\xe4\xbd" is the leading two bytes of U+4F60; no continuation byte
  5417. // follows before exit.
  5418. String.raw`os.write(1, b"\xe4\xbd")`,
  5419. 'return "done"',
  5420. ].join('\n'),
  5421. bindings: [],
  5422. })
  5423. expect(result.error).toBeUndefined()
  5424. expect(result.logs.join('')).toContain('�')
  5425. })
  5426. it('rejects reserved words of EITHER backend language as binding globals', async () => {
  5427. // The seam's portable contract: `lambda` (Python keyword, legal JS name)
  5428. // and `typeof` (JS keyword, legal Python name) are both refused, so a
  5429. // namespace list valid on one backend is valid on every backend.
  5430. const { runtime } = await setup()
  5431. for (const global of ['lambda', 'typeof']) {
  5432. await expect(runtime.run({
  5433. program: 'return 1',
  5434. bindings: [{ global, functions: {} }],
  5435. })).rejects.toThrow(/is not a usable Python identifier/)
  5436. }
  5437. })
  5438. it('captures stray stdout bytes the child writes bypassing sys.stdout', async () => {
  5439. // Model code that writes to fd 1 via os.write() bypasses the Python-side
  5440. // LogBuffer, so the host's stray-byte capture on child.stdout is what
  5441. // records it.
  5442. const { runtime } = await setup()
  5443. const result = await runtime.run({
  5444. program: [
  5445. 'import os',
  5446. 'os.write(1, b"stray stdout\\n")',
  5447. 'os.write(2, b"stray stderr\\n")',
  5448. 'return "done"',
  5449. ].join('\n'),
  5450. bindings: [],
  5451. })
  5452. expect(result.error).toBeUndefined()
  5453. expect(result.logs.join('')).toContain('stray stdout')
  5454. expect(result.logs.join('')).toContain('stray stderr')
  5455. })
  5456. it('flushes bytes written through sys.__stdout__/sys.__stderr__ before the done frame', async () => {
  5457. // The bootstrap only replaces sys.stdout/sys.stderr with the _LogStream;
  5458. // sys.__stdout__/sys.__stderr__ are the original block-buffered wrappers
  5459. // over fd 1/2. A program that writes through them without an explicit flush
  5460. // would lose those bytes when the host SIGTERMs the child right after the
  5461. // done frame (the default SIGTERM disposition terminates without
  5462. // interpreter finalization). The settlement flush now drains the original
  5463. // std streams before sending the done frame, so the bytes land in the
  5464. // kernel pipe buffer and the host's stray capture records them.
  5465. const { runtime } = await setup()
  5466. const result = await runtime.run({
  5467. program: [
  5468. 'import sys',
  5469. // `-u` makes the streams write-through; re-enable block buffering so
  5470. // the bytes sit in the wrapper until the SETTLEMENT drain flushes them
  5471. // — the drain path, not the -u immediate write, is what this case pins.
  5472. 'if hasattr(sys.__stdout__, "reconfigure"):',
  5473. ' sys.__stdout__.reconfigure(write_through=False)',
  5474. ' sys.__stderr__.reconfigure(write_through=False)',
  5475. 'sys.__stdout__.write("orig stdout\\n")',
  5476. 'sys.__stderr__.write("orig stderr\\n")',
  5477. 'return "done"',
  5478. ].join('\n'),
  5479. bindings: [],
  5480. })
  5481. expect(result.error).toBeUndefined()
  5482. expect(result.value).toBe('done')
  5483. expect(result.logs.join('')).toContain('orig stdout')
  5484. expect(result.logs.join('')).toContain('orig stderr')
  5485. }, 15_000)
  5486. it('escalates to SIGKILL when the program traps SIGTERM and ignores the grace period', async () => {
  5487. // A program that traps SIGTERM should still die: the kill() escalation
  5488. // fires SIGKILL after graceMs. The full run reports either timeout (wall)
  5489. // or worker-exit depending on which finish reason wins the race.
  5490. const { runtime } = await setup({ maxWallMs: 400, graceMs: 200 })
  5491. const result = await runtime.run({
  5492. program: [
  5493. 'import signal, time',
  5494. 'signal.signal(signal.SIGTERM, lambda *a: None)',
  5495. 'while True: time.sleep(1)',
  5496. ].join('\n'),
  5497. bindings: [],
  5498. })
  5499. expect(['timeout', 'worker-exit']).toContain(result.error?.kind)
  5500. }, 6000)
  5501. it('bounds the fd-3 receive buffer against a newline-free flood', async () => {
  5502. // A program looping os.write(3, ...) with no newline would grow the host
  5503. // accumulator unbounded (the child's RLIMIT_AS does not cover the host
  5504. // string). The frame cap is a fixed 64 MiB memory-safety invariant —
  5505. // deliberately NOT derived from maxValueBytes, because legitimate binding
  5506. // call frames may be large. We flood slightly past it in 8 MiB writes so
  5507. // the test terminates promptly once the guard trips.
  5508. const ceiling = 64 * 1024 * 1024
  5509. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  5510. const start = Date.now()
  5511. const result = await runtime.run({
  5512. program: [
  5513. 'import os',
  5514. `for _ in range(${Math.ceil((ceiling * 1.1) / (8 * 1024 * 1024))}):`,
  5515. ' os.write(3, b"A" * (8 * 1024 * 1024))',
  5516. 'return "never"',
  5517. ].join('\n'),
  5518. bindings: [],
  5519. })
  5520. const elapsed = Date.now() - start
  5521. expect(result.error?.kind).toBe('worker-exit')
  5522. expect(result.error?.message).toContain(`protocol frame exceeded ${ceiling} bytes`)
  5523. // The breach ends the run before the wall ceiling (the run did not idle
  5524. // out); absolute pipe throughput varies too much under parallel suites
  5525. // for a tight bound.
  5526. expect(elapsed).toBeLessThan(30_000)
  5527. }, 45_000)
  5528. it('fails a forged oversized done value host-side as output-limit', async () => {
  5529. // The Python-side _done_with_value check is bypassable by writing a done
  5530. // frame straight to fd 3. The host re-enforces maxValueBytes; the seam
  5531. // forbids substituting a truncated value, so the run FAILS as output-limit
  5532. // instead of returning a lie.
  5533. const maxValueBytes = 64
  5534. const { runtime } = await setup({ maxValueBytes })
  5535. const result = await runtime.run({
  5536. program: [
  5537. 'import os, json',
  5538. 'big = "B" * 5000',
  5539. 'os.write(3, json.dumps({"type":"done","value":big}).encode() + b"\\n")',
  5540. // The real done never sends; the forged one settles the run.
  5541. 'import time',
  5542. 'time.sleep(5)',
  5543. ].join('\n'),
  5544. bindings: [],
  5545. })
  5546. expect(result.value).toBeUndefined()
  5547. expect(result.error?.kind).toBe('output-limit')
  5548. expect(result.error?.message).toContain('exceeded 64 bytes')
  5549. }, 8000)
  5550. it('drops a forged oversized log frame on its code-unit lower bound, before escaping it', async () => {
  5551. // A forged `log` frame carrying a control-heavy string: NULs escape
  5552. // several-fold (one NUL -> six bytes `\u0000`). The raw frame stays under
  5553. // the host's 64 MiB parse cap (4 MiB of `\u0000` text = 24 MiB raw) while
  5554. // the escaped form would be ~24 MiB. Charging it required building that
  5555. // escaped copy first, so a 32-byte maxLogBytes could still force a large
  5556. // host allocation. The cheap `length + 3` lower bound truncates it instead.
  5557. // The host's own heap is what is under test, so keep the child's address
  5558. // space generous enough to BUILD the frame.
  5559. const { runtime } = await setup({ maxLogBytes: 128, addressSpaceMb: 1024, maxWallMs: 60_000 })
  5560. const before = process.memoryUsage().heapUsed
  5561. const result = await runtime.run({
  5562. program: [
  5563. 'import os',
  5564. // Written as a raw frame so the child's own ledger never sees it.
  5565. 'os.write(3, b\'{"type":"log","text":"\' + b"\\\\u0000" * (4 * 1024 * 1024) + b\'"}\\n\')',
  5566. 'return "settled"',
  5567. ].join('\n'),
  5568. bindings: [],
  5569. })
  5570. expect(result.error).toBeUndefined()
  5571. expect(result.value).toBe('settled')
  5572. // The frame was dropped as one truncation marker, not retained.
  5573. expect(result.logs).toEqual([logTruncationMarker(128)])
  5574. // The escaped copy (~144 MiB) was never materialized.
  5575. expect(process.memoryUsage().heapUsed - before).toBeLessThan(256 * 1024 * 1024)
  5576. }, 90_000)
  5577. it('charges a forged log frame its escaped cost once past the code-unit lower bound', async () => {
  5578. // The cheap lower bound only rejects what cannot possibly fit; a SHORT
  5579. // control-heavy frame clears it and must still be charged what it costs on
  5580. // the wire. Eleven NULs are 14 against the 64-byte ledger's cheap bound
  5581. // (six bytes each, two quotes, one separator), so the full charge truncates.
  5582. const { runtime } = await setup({ maxLogBytes: 64 })
  5583. const result = await runtime.run({
  5584. program: [
  5585. 'import os',
  5586. 'os.write(3, b\'{"type":"log","text":"\' + b"\\\\u0000" * 11 + b\'"}\\n\')',
  5587. 'return "settled"',
  5588. ].join('\n'),
  5589. bindings: [],
  5590. })
  5591. expect(result.error).toBeUndefined()
  5592. expect(result.value).toBe('settled')
  5593. expect(result.logs).toEqual([logTruncationMarker(64)])
  5594. }, 8000)
  5595. it('caps a forged done error.message from its code-unit prefix, never encoding the whole message', async () => {
  5596. // `Buffer.from(message)` on a message near the frame ceiling allocates a
  5597. // full UTF-8 copy before maxValueBytes applies. Only the first
  5598. // maxValueBytes code units can fit the cap, so only that prefix is encoded
  5599. // — at most 3x the cap in bytes. The message here is 48 MiB of ASCII: its
  5600. // full encode would be another 48 MiB in the host.
  5601. const maxValueBytes = 64
  5602. const { runtime } = await setup({ maxValueBytes, addressSpaceMb: 1024, maxWallMs: 60_000 })
  5603. const before = process.memoryUsage().heapUsed
  5604. const result = await runtime.run({
  5605. program: [
  5606. 'import os',
  5607. 'os.write(3, b\'{"type":"done","error":{"kind":"exception","message":"\' + b"E" * (48 * 1024 * 1024) + b\'"}}\\n\')',
  5608. 'import time',
  5609. 'time.sleep(30)',
  5610. ].join('\n'),
  5611. bindings: [],
  5612. })
  5613. expect(result.error?.kind).toBe('exception')
  5614. const message = result.error?.message ?? ''
  5615. // The marker's 15 bytes come OUT of the 64-byte cap, so 49 E's precede it
  5616. // and the whole string is exactly 64 bytes — not 64 plus the marker.
  5617. expect(message).toBe(`${'E'.repeat(maxValueBytes - 15)}… [truncated]`)
  5618. expect(Buffer.byteLength(message, 'utf8')).toBe(maxValueBytes)
  5619. // JSON.parse already holds the 48 MiB string; the cap must not add a
  5620. // second full-length copy on top of it.
  5621. expect(process.memoryUsage().heapUsed - before).toBeLessThan(256 * 1024 * 1024)
  5622. }, 90_000)
  5623. it('keeps a capped diagnostic within maxValueBytes, marker included', async () => {
  5624. // The marker is part of the emitted diagnostic, so its bytes are reserved
  5625. // from the cap rather than appended past it — the host meters this same
  5626. // field downstream. Checked on BOTH producers: the child's own _cap_message
  5627. // (a raised exception) and the host's capMessage (a forged done frame).
  5628. const maxValueBytes = 40
  5629. const { runtime } = await setup({ maxValueBytes })
  5630. const raised = await runtime.run({
  5631. program: 'raise ValueError("R" * 100000)',
  5632. bindings: [],
  5633. })
  5634. expect(raised.error?.kind).toBe('exception')
  5635. const raisedMessage = raised.error?.message ?? ''
  5636. expect(raisedMessage.endsWith('… [truncated]')).toBe(true)
  5637. expect(Buffer.byteLength(raisedMessage, 'utf8')).toBeLessThanOrEqual(maxValueBytes)
  5638. const forged = await runtime.run({
  5639. program: [
  5640. 'import os, json',
  5641. 'msg = "F" * 100000',
  5642. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":msg}}).encode() + b"\\n")',
  5643. 'import time',
  5644. 'time.sleep(5)',
  5645. ].join('\n'),
  5646. bindings: [],
  5647. })
  5648. expect(forged.error?.kind).toBe('exception')
  5649. const forgedMessage = forged.error?.message ?? ''
  5650. expect(forgedMessage.endsWith('… [truncated]')).toBe(true)
  5651. expect(Buffer.byteLength(forgedMessage, 'utf8')).toBe(maxValueBytes)
  5652. }, 15_000)
  5653. it('emits the marker alone when the cap is smaller than the marker itself', async () => {
  5654. // With maxValueBytes below the marker's own 15 bytes there is no room for
  5655. // message text; the marker still goes out, so the truncation stays reported
  5656. // instead of the diagnostic silently becoming empty. Both producers agree.
  5657. const { runtime } = await setup({ maxValueBytes: 4 })
  5658. const raised = await runtime.run({ program: 'raise ValueError("R" * 500)', bindings: [] })
  5659. expect(raised.error?.kind).toBe('exception')
  5660. expect(raised.error?.message).toBe('… [truncated]')
  5661. const forged = await runtime.run({
  5662. program: [
  5663. 'import os, json',
  5664. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":"F" * 500}}).encode() + b"\\n")',
  5665. 'import time',
  5666. 'time.sleep(5)',
  5667. ].join('\n'),
  5668. bindings: [],
  5669. })
  5670. expect(forged.error?.kind).toBe('exception')
  5671. expect(forged.error?.message).toBe('… [truncated]')
  5672. }, 15_000)
  5673. it('caps a forged done error.message without splitting a surrogate pair', async () => {
  5674. // At exactly maxValueBytes code units the prefix can end on a high
  5675. // surrogate whose low half sits just outside it. `Buffer.from` encodes that
  5676. // orphan as U+FFFD — the same corruption a mid-sequence byte cut causes —
  5677. // and those three replacement bytes sit past the marker-reserved budget, so
  5678. // the byte trim-back drops them.
  5679. const maxValueBytes = 32
  5680. const { runtime } = await setup({ maxValueBytes })
  5681. const result = await runtime.run({
  5682. program: [
  5683. 'import os, json',
  5684. // 32 ASCII chars then astral characters: code unit 32 is the first
  5685. // character's high surrogate (Python spells it as one code point, so
  5686. // json.dumps emits the raw 4 bytes the host reads back as a pair).
  5687. 'msg = "A" * 32 + "\\U0001f600" * 4',
  5688. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":msg}}).encode() + b"\\n")',
  5689. 'import time',
  5690. 'time.sleep(5)',
  5691. ].join('\n'),
  5692. bindings: [],
  5693. })
  5694. expect(result.error?.kind).toBe('exception')
  5695. // 17 A's fill the marker-reserved budget; no orphaned half, no U+FFFD.
  5696. expect(result.error?.message).toBe(`${'A'.repeat(17)}… [truncated]`)
  5697. expect(Buffer.byteLength(result.error?.message ?? '', 'utf8')).toBe(maxValueBytes)
  5698. }, 8000)
  5699. it('returns a diagnostic under a third of the cap untouched, skipping the encode', async () => {
  5700. // Under maxValueBytes/3 code units a message cannot overflow the cap
  5701. // whatever it holds (3 bytes is the per-code-unit maximum), so the fast
  5702. // path returns it without encoding anything. Non-ASCII proves the bound is
  5703. // the code-unit count, not a byte assumption: 6 characters at 3 bytes each
  5704. // is 18 bytes, inside the 64-byte cap.
  5705. const { runtime } = await setup({ maxValueBytes: 64 })
  5706. const result = await runtime.run({
  5707. program: [
  5708. 'import os, json',
  5709. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":"中文中文中文"}}).encode() + b"\\n")',
  5710. 'import time',
  5711. 'time.sleep(5)',
  5712. ].join('\n'),
  5713. bindings: [],
  5714. })
  5715. expect(result.error?.kind).toBe('exception')
  5716. expect(result.error?.message).toBe('中文中文中文')
  5717. }, 8000)
  5718. it('re-caps a forged done error.message host-side on a UTF-8 boundary', async () => {
  5719. // A forged done frame can carry an arbitrarily long error message; the
  5720. // host caps it to maxValueBytes and appends the shared marker. The
  5721. // message is emoji-dense and the cap is chosen so the marker-reserved
  5722. // 51-byte cut lands INSIDE a 4-byte sequence (one ASCII byte then 4-byte
  5723. // runs, so only a cut at 1 + 4k is aligned) — the cap must trim back to a
  5724. // code-point boundary rather than decode a replacement character, which
  5725. // would also exceed the cap.
  5726. const maxValueBytes = 66
  5727. const { runtime } = await setup({ maxValueBytes })
  5728. const result = await runtime.run({
  5729. program: [
  5730. 'import os, json',
  5731. 'msg = "E" + "\\U0001f600" * 2000',
  5732. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":msg}}).encode() + b"\\n")',
  5733. 'import time',
  5734. 'time.sleep(5)',
  5735. ].join('\n'),
  5736. bindings: [],
  5737. })
  5738. expect(result.error?.kind).toBe('exception')
  5739. const message = result.error?.message ?? ''
  5740. expect(message.endsWith('… [truncated]')).toBe(true)
  5741. const marker = '… [truncated]'
  5742. const body = message.slice(0, message.length - marker.length)
  5743. // The WHOLE message, marker included, honors the cap.
  5744. expect(Buffer.byteLength(message, 'utf8')).toBeLessThanOrEqual(maxValueBytes)
  5745. // 'E' plus 12 emoji is 49 bytes: the trim-back walked the 51-byte budget
  5746. // down past two continuation bytes rather than splitting the 13th.
  5747. expect(body).toBe(`E${'\u{1f600}'.repeat(12)}`)
  5748. // The cut landed on a code-point boundary — no replacement character.
  5749. expect(body).not.toContain('\ufffd')
  5750. }, 8000)
  5751. it('bounds a single oversized newline-terminated line on fd 3', async () => {
  5752. // The same cap applies to one giant framed line. Write EXACTLY the
  5753. // cap with no newline — at the limit, not past it, so nothing trips —
  5754. // then a small newline tail, which is the chunk that crosses.
  5755. const ceiling = 64 * 1024 * 1024
  5756. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  5757. const result = await runtime.run({
  5758. program: [
  5759. 'import os',
  5760. 'chunk = b"A" * (8 * 1024 * 1024)',
  5761. `for _ in range(${ceiling / (8 * 1024 * 1024)}):`,
  5762. ' os.write(3, chunk)',
  5763. 'os.write(3, b"AAAA\\n")',
  5764. 'return "never"',
  5765. ].join('\n'),
  5766. bindings: [],
  5767. })
  5768. expect(result.error?.kind).toBe('worker-exit')
  5769. expect(result.error?.message).toContain(`protocol frame exceeded ${ceiling} bytes`)
  5770. }, 90_000)
  5771. it('rejects an over-cap newline-free fd-3 buffer without first joining it into one line', async () => {
  5772. // The cap has to be enforced on the byte COUNTER before Buffer.concat,
  5773. // not on the joined line afterwards: the join is a second copy of
  5774. // everything held, so a program could force roughly twice the advertised
  5775. // 64 MiB of host memory before anything rejected it.
  5776. //
  5777. // This program writes past the cap with no newline: the counter crosses on
  5778. // the 9th 8 MiB write (72 MiB) while the buffer is still a single unframed
  5779. // line, so the pre-join check rejects it without concat-ing a second copy.
  5780. // Checking the joined line instead would have produced a 72 MiB FIRST LINE
  5781. // that the per-line bound then dropped only after the doubling had happened.
  5782. const ceiling = 64 * 1024 * 1024
  5783. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  5784. const result = await runtime.run({
  5785. program: [
  5786. 'import os',
  5787. 'chunk = b"A" * (8 * 1024 * 1024)',
  5788. `for _ in range(${ceiling / (8 * 1024 * 1024) + 1}):`,
  5789. ' os.write(3, chunk)',
  5790. 'return "never"',
  5791. ].join('\n'),
  5792. bindings: [],
  5793. })
  5794. expect(result.value).toBeUndefined()
  5795. expect(result.error?.kind).toBe('worker-exit')
  5796. expect(result.error?.message).toContain(`protocol frame exceeded ${ceiling} bytes`)
  5797. }, 120_000)
  5798. it('keeps two within-cap frames whose combined buffer crosses the cap', async () => {
  5799. // The unframed byte counter charges the WHOLE buffer, which legitimately
  5800. // holds several frames each within FRAME_PARSE_CAP_BYTES. A first frame of
  5801. // exactly the cap followed by a second frame crosses the counter without
  5802. // either frame exceeding the cap; the first-frame check (not the counter)
  5803. // must let them through, or a legitimate near-cap frame plus a trailing
  5804. // frame would be misreported as a worker-exit.
  5805. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  5806. const result = await runtime.run({
  5807. program: [
  5808. 'import os',
  5809. 'chunk = b"A" * (8 * 1024 * 1024)',
  5810. // Exactly the cap, no newline — at the limit, so nothing trips.
  5811. 'for _ in range(8):',
  5812. ' os.write(3, chunk)',
  5813. // A newline, then a small legitimate log frame.
  5814. 'os.write(3, b"\\n{\\"type\\":\\"log\\",\\"text\\":\\"after-cap-frames\\"}\\n")',
  5815. 'return "done"',
  5816. ].join('\n'),
  5817. bindings: [],
  5818. })
  5819. expect(result.error).toBeUndefined()
  5820. expect(result.value).toBe('done')
  5821. expect(result.logs).toContain('after-cap-frames')
  5822. }, 120_000)
  5823. it('rejects an oversized first frame that lands on the sealing threshold with a newline', async () => {
  5824. // The fragment-count seal runs only on newline-free chunks (the ELSE half
  5825. // of the newline branch), so a chunk that carries the first newline always
  5826. // reaches the join and its first-frame check; sealing it into a block
  5827. // would empty pendingChunks, leave sawNewline false, and skip that check.
  5828. // Whether the pipe delivers exactly 1024 chunks is timing-dependent, but
  5829. // the oversized first frame (63.9 MiB of A's + 12289 more before the
  5830. // newline) exceeds FRAME_PARSE_CAP_BYTES no matter how it arrives — the
  5831. // case pins the worker-exit settlement, not a pre/post copy-count
  5832. // distinction (both orders reject an over-cap frame).
  5833. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  5834. const result = await runtime.run({
  5835. program: [
  5836. 'import os',
  5837. // 4 KiB writes are <= PIPE_BUF, so each os.write is atomic and the
  5838. // host sees one chunk per write; 16384 of them accumulate 64 MiB of
  5839. // newline-free bytes (16 fragment-count seals of 1024 chunks).
  5840. 'chunk = b"A" * 4096',
  5841. 'for _ in range(16384):',
  5842. ' os.write(3, chunk)',
  5843. // 12289 more A's push the first frame past 64 MiB; drain-loop so the
  5844. // write cannot truncate, then a newline and a small legitimate frame.
  5845. "data = b'A' * 12289 + b'\\n' + b'{\"type\":\"log\",\"text\":\"after-seal\"}\\n'",
  5846. 'view = memoryview(data)',
  5847. 'while view:',
  5848. ' view = view[os.write(3, view):]',
  5849. 'return "done"',
  5850. ].join('\n'),
  5851. bindings: [],
  5852. })
  5853. expect(result.error?.kind).toBe('worker-exit')
  5854. expect(result.error?.message).toContain('protocol frame exceeded')
  5855. }, 120_000)
  5856. })