1
0

runtime.spec.ts 256 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131213221332134213521362137213821392140214121422143214421452146214721482149215021512152215321542155215621572158215921602161216221632164216521662167216821692170217121722173217421752176217721782179218021812182218321842185218621872188218921902191219221932194219521962197219821992200220122022203220422052206220722082209221022112212221322142215221622172218221922202221222222232224222522262227222822292230223122322233223422352236223722382239224022412242224322442245224622472248224922502251225222532254225522562257225822592260226122622263226422652266226722682269227022712272227322742275227622772278227922802281228222832284228522862287228822892290229122922293229422952296229722982299230023012302230323042305230623072308230923102311231223132314231523162317231823192320232123222323232423252326232723282329233023312332233323342335233623372338233923402341234223432344234523462347234823492350235123522353235423552356235723582359236023612362236323642365236623672368236923702371237223732374237523762377237823792380238123822383238423852386238723882389239023912392239323942395239623972398239924002401240224032404240524062407240824092410241124122413241424152416241724182419242024212422242324242425242624272428242924302431243224332434243524362437243824392440244124422443244424452446244724482449245024512452245324542455245624572458245924602461246224632464246524662467246824692470247124722473247424752476247724782479248024812482248324842485248624872488248924902491249224932494249524962497249824992500250125022503250425052506250725082509251025112512251325142515251625172518251925202521252225232524252525262527252825292530253125322533253425352536253725382539254025412542254325442545254625472548254925502551255225532554255525562557255825592560256125622563256425652566256725682569257025712572257325742575257625772578257925802581258225832584258525862587258825892590259125922593259425952596259725982599260026012602260326042605260626072608260926102611261226132614261526162617261826192620262126222623262426252626262726282629263026312632263326342635263626372638263926402641264226432644264526462647264826492650265126522653265426552656265726582659266026612662266326642665266626672668266926702671267226732674267526762677267826792680268126822683268426852686268726882689269026912692269326942695269626972698269927002701270227032704270527062707270827092710271127122713271427152716271727182719272027212722272327242725272627272728272927302731273227332734273527362737273827392740274127422743274427452746274727482749275027512752275327542755275627572758275927602761276227632764276527662767276827692770277127722773277427752776277727782779278027812782278327842785278627872788278927902791279227932794279527962797279827992800280128022803280428052806280728082809281028112812281328142815281628172818281928202821282228232824282528262827282828292830283128322833283428352836283728382839284028412842284328442845284628472848284928502851285228532854285528562857285828592860286128622863286428652866286728682869287028712872287328742875287628772878287928802881288228832884288528862887288828892890289128922893289428952896289728982899290029012902290329042905290629072908290929102911291229132914291529162917291829192920292129222923292429252926292729282929293029312932293329342935293629372938293929402941294229432944294529462947294829492950295129522953295429552956295729582959296029612962296329642965296629672968296929702971297229732974297529762977297829792980298129822983298429852986298729882989299029912992299329942995299629972998299930003001300230033004300530063007300830093010301130123013301430153016301730183019302030213022302330243025302630273028302930303031303230333034303530363037303830393040304130423043304430453046304730483049305030513052305330543055305630573058305930603061306230633064306530663067306830693070307130723073307430753076307730783079308030813082308330843085308630873088308930903091309230933094309530963097309830993100310131023103310431053106310731083109311031113112311331143115311631173118311931203121312231233124312531263127312831293130313131323133313431353136313731383139314031413142314331443145314631473148314931503151315231533154315531563157315831593160316131623163316431653166316731683169317031713172317331743175317631773178317931803181318231833184318531863187318831893190319131923193319431953196319731983199320032013202320332043205320632073208320932103211321232133214321532163217321832193220322132223223322432253226322732283229323032313232323332343235323632373238323932403241324232433244324532463247324832493250325132523253325432553256325732583259326032613262326332643265326632673268326932703271327232733274327532763277327832793280328132823283328432853286328732883289329032913292329332943295329632973298329933003301330233033304330533063307330833093310331133123313331433153316331733183319332033213322332333243325332633273328332933303331333233333334333533363337333833393340334133423343334433453346334733483349335033513352335333543355335633573358335933603361336233633364336533663367336833693370337133723373337433753376337733783379338033813382338333843385338633873388338933903391339233933394339533963397339833993400340134023403340434053406340734083409341034113412341334143415341634173418341934203421342234233424342534263427342834293430343134323433343434353436343734383439344034413442344334443445344634473448344934503451345234533454345534563457345834593460346134623463346434653466346734683469347034713472347334743475347634773478347934803481348234833484348534863487348834893490349134923493349434953496349734983499350035013502350335043505350635073508350935103511351235133514351535163517351835193520352135223523352435253526352735283529353035313532353335343535353635373538353935403541354235433544354535463547354835493550355135523553355435553556355735583559356035613562356335643565356635673568356935703571357235733574357535763577357835793580358135823583358435853586358735883589359035913592359335943595359635973598359936003601360236033604360536063607360836093610361136123613361436153616361736183619362036213622362336243625362636273628362936303631363236333634363536363637363836393640364136423643364436453646364736483649365036513652365336543655365636573658365936603661366236633664366536663667366836693670367136723673367436753676367736783679368036813682368336843685368636873688368936903691369236933694369536963697369836993700370137023703370437053706370737083709371037113712371337143715371637173718371937203721372237233724372537263727372837293730373137323733373437353736373737383739374037413742374337443745374637473748374937503751375237533754375537563757375837593760376137623763376437653766376737683769377037713772377337743775377637773778377937803781378237833784378537863787378837893790379137923793379437953796379737983799380038013802380338043805380638073808380938103811381238133814381538163817381838193820382138223823382438253826382738283829383038313832383338343835383638373838383938403841384238433844384538463847384838493850385138523853385438553856385738583859386038613862386338643865386638673868386938703871387238733874387538763877387838793880388138823883388438853886388738883889389038913892389338943895389638973898389939003901390239033904390539063907390839093910391139123913391439153916391739183919392039213922392339243925392639273928392939303931393239333934393539363937393839393940394139423943394439453946394739483949395039513952395339543955395639573958395939603961396239633964396539663967396839693970397139723973397439753976397739783979398039813982398339843985398639873988398939903991399239933994399539963997399839994000400140024003400440054006400740084009401040114012401340144015401640174018401940204021402240234024402540264027402840294030403140324033403440354036403740384039404040414042404340444045404640474048404940504051405240534054405540564057405840594060406140624063406440654066406740684069407040714072407340744075407640774078407940804081408240834084408540864087408840894090409140924093409440954096409740984099410041014102410341044105410641074108410941104111411241134114411541164117411841194120412141224123412441254126412741284129413041314132413341344135413641374138413941404141414241434144414541464147414841494150415141524153415441554156415741584159416041614162416341644165416641674168416941704171417241734174417541764177417841794180418141824183418441854186418741884189419041914192419341944195419641974198419942004201420242034204420542064207420842094210421142124213421442154216421742184219422042214222422342244225422642274228422942304231423242334234423542364237423842394240424142424243424442454246424742484249425042514252425342544255425642574258425942604261426242634264426542664267426842694270427142724273427442754276427742784279428042814282428342844285428642874288428942904291429242934294429542964297429842994300430143024303430443054306430743084309431043114312431343144315431643174318431943204321432243234324432543264327432843294330433143324333433443354336433743384339434043414342434343444345434643474348434943504351435243534354435543564357435843594360436143624363436443654366436743684369437043714372437343744375437643774378437943804381438243834384438543864387438843894390439143924393439443954396439743984399440044014402440344044405440644074408440944104411441244134414441544164417441844194420442144224423442444254426442744284429443044314432443344344435443644374438443944404441444244434444444544464447444844494450445144524453445444554456445744584459446044614462446344644465446644674468446944704471447244734474447544764477447844794480448144824483448444854486448744884489449044914492449344944495449644974498449945004501450245034504450545064507450845094510451145124513451445154516451745184519452045214522452345244525452645274528452945304531453245334534453545364537453845394540454145424543454445454546454745484549455045514552455345544555455645574558455945604561456245634564456545664567456845694570457145724573457445754576457745784579458045814582458345844585458645874588458945904591459245934594459545964597459845994600460146024603460446054606460746084609461046114612461346144615461646174618461946204621462246234624462546264627462846294630463146324633463446354636463746384639464046414642464346444645464646474648464946504651465246534654465546564657465846594660466146624663466446654666466746684669467046714672467346744675467646774678467946804681468246834684468546864687468846894690469146924693469446954696469746984699470047014702470347044705470647074708470947104711471247134714471547164717471847194720472147224723472447254726472747284729473047314732473347344735473647374738473947404741474247434744474547464747474847494750475147524753475447554756475747584759476047614762476347644765476647674768476947704771477247734774477547764777477847794780478147824783478447854786478747884789479047914792479347944795479647974798479948004801480248034804480548064807480848094810481148124813481448154816481748184819482048214822482348244825482648274828482948304831483248334834483548364837483848394840484148424843484448454846484748484849485048514852485348544855485648574858485948604861486248634864486548664867486848694870487148724873487448754876487748784879488048814882488348844885488648874888488948904891489248934894489548964897489848994900490149024903490449054906490749084909491049114912491349144915491649174918491949204921492249234924492549264927492849294930493149324933493449354936493749384939494049414942494349444945494649474948494949504951495249534954495549564957495849594960496149624963496449654966496749684969497049714972497349744975497649774978497949804981498249834984498549864987498849894990499149924993499449954996499749984999500050015002500350045005500650075008500950105011501250135014501550165017501850195020502150225023502450255026502750285029503050315032503350345035503650375038503950405041504250435044504550465047504850495050505150525053505450555056505750585059506050615062506350645065506650675068506950705071507250735074507550765077507850795080508150825083508450855086508750885089509050915092509350945095509650975098509951005101510251035104510551065107510851095110511151125113511451155116511751185119512051215122512351245125512651275128512951305131513251335134513551365137513851395140514151425143514451455146514751485149515051515152515351545155515651575158515951605161516251635164516551665167516851695170517151725173517451755176517751785179518051815182518351845185518651875188518951905191519251935194519551965197519851995200520152025203520452055206520752085209521052115212521352145215521652175218521952205221522252235224522552265227522852295230523152325233523452355236523752385239524052415242524352445245524652475248524952505251525252535254525552565257525852595260526152625263526452655266526752685269527052715272527352745275527652775278527952805281528252835284528552865287528852895290529152925293529452955296529752985299530053015302530353045305530653075308530953105311531253135314531553165317531853195320532153225323532453255326532753285329533053315332533353345335
  1. import { existsSync, readdirSync, realpathSync, rmSync, statSync, writeFileSync } from 'node:fs'
  2. import { mkdtemp, writeFile } from 'node:fs/promises'
  3. import { tmpdir } from 'node:os'
  4. import { basename, dirname, join } from 'node:path'
  5. import { describe, expect, it, vi } from 'vitest'
  6. import { Context } from '@deepseek-ai/cordis'
  7. import { PythonCodeRuntime, readProcessStart, resolvePythonBin } from '../src/index.ts'
  8. import { logTruncationMarker } from '../src/protocol.ts'
  9. import type { Config } from '../src/index.ts'
  10. // Absolute interpreter path for the shell wrappers: the runtime spawns the
  11. // child with env:{} (an empty environment by design), so a bare 'python3' in a
  12. // wrapper resolves against /bin/sh's compiled-in default PATH, which misses
  13. // interpreters only reachable through the caller's PATH (Nix, pyenv). Baking
  14. // the resolved absolute path mirrors what resolvePythonBin does for the product
  15. // spawn.
  16. const PYABS = resolvePythonBin('python3') ?? 'python3'
  17. import type { CodeBindingFunction, CodeJsonValue, CodeRunResult } from '@deepseek-ai/dsh-code-runtime'
  18. /**
  19. * Names one `py/` script whose `copyFileSync` must fail, for the partial-staging
  20. * case. A real disk-full or missing-asset failure mid-copy cannot be produced
  21. * from a test, and the leak only shows when `mkdtempSync` has already succeeded.
  22. */
  23. const { failNextCopyOf } = vi.hoisted(() => ({ failNextCopyOf: { value: undefined as string | undefined } }))
  24. vi.mock('node:fs', async (importOriginal) => {
  25. const actual = await importOriginal<typeof import('node:fs')>()
  26. return {
  27. ...actual,
  28. copyFileSync(source: string, destination: string): void {
  29. if (failNextCopyOf.value !== undefined && basename(source) === failNextCopyOf.value) {
  30. failNextCopyOf.value = undefined
  31. throw Object.assign(new Error('simulated ENOSPC on copy'), { code: 'ENOSPC' })
  32. }
  33. actual.copyFileSync(source, destination)
  34. },
  35. }
  36. })
  37. /**
  38. * Integration suite over REAL python3 subprocesses (no subprocess mocks — it is
  39. * cheap and local, per docs/testing.md's real-over-mock policy; the only mock is
  40. * `node:fs.copyFileSync` for the staging-failure cases). Each test builds a fresh
  41. * runtime so budgets can be tuned per case.
  42. */
  43. async function setup(config: Config = {}) {
  44. const ctx = new Context()
  45. const fiber = await ctx.plugin(PythonCodeRuntime, config)
  46. const runtime = ctx.codeRuntime as PythonCodeRuntime
  47. return { ctx, fiber, runtime }
  48. }
  49. /** Convenience: one namespace `tools` with the given functions. */
  50. function tools(functions: Record<string, CodeBindingFunction>) {
  51. return [{ global: 'tools', functions }]
  52. }
  53. describe('PythonCodeRuntime — seam descriptors and misuse', () => {
  54. it('registers the seam descriptors', async () => {
  55. const { runtime } = await setup()
  56. expect(runtime.language).toBe('python')
  57. expect(runtime.isolation).toBe('process')
  58. })
  59. it('rejects non-positive config as seam misuse', async () => {
  60. const ctx = new Context()
  61. await expect(ctx.plugin(PythonCodeRuntime, { cpuSeconds: 0 }))
  62. .rejects.toThrow(/cpuSeconds must be a positive number/)
  63. await expect(ctx.plugin(PythonCodeRuntime, { maxWallMs: -1 }))
  64. .rejects.toThrow(/maxWallMs must be a positive number/)
  65. })
  66. it('rejects a non-integer cpuSeconds at load (setrlimit needs an int)', async () => {
  67. const ctx = new Context()
  68. await expect(ctx.plugin(PythonCodeRuntime, { cpuSeconds: 1.5 }))
  69. .rejects.toThrow(/cpuSeconds must be a positive integer, got 1.5/)
  70. })
  71. it('rejects a non-integer byte budget at load (the child int()-truncates it)', async () => {
  72. // maxLogBytes/maxValueBytes cross to the child, which reads them through
  73. // int(...): a float would floor there while the host meters the fraction, so
  74. // the two sides would enforce different public config. Reject at load.
  75. const ctxLog = new Context()
  76. await expect(ctxLog.plugin(PythonCodeRuntime, { maxLogBytes: 3.5 }))
  77. .rejects.toThrow(/maxLogBytes must be a positive integer/)
  78. const ctxValue = new Context()
  79. await expect(ctxValue.plugin(PythonCodeRuntime, { maxValueBytes: 1024.5 }))
  80. .rejects.toThrow(/maxValueBytes must be a positive integer/)
  81. })
  82. it('rejects finite numeric config that cannot cross as an exact rlimit integer', async () => {
  83. // `Number.isFinite` and `Number.isInteger` both admit values that cannot
  84. // round-trip. `addressSpaceMb: 1e308` overflows to `Infinity` once multiplied
  85. // by 1 MiB, and `encodeJsonPlain` renders that as `null`, so the child gets no
  86. // limit at all; `cpuSeconds: 1e100` clears `Number.isInteger` while sitting
  87. // far past the safe range, so `setrlimit` receives a different number than was
  88. // configured. Both used to end every run in a bootstrap exception instead of
  89. // failing at load, where a self-contained configuration error belongs.
  90. const ctx = new Context()
  91. await expect(ctx.plugin(PythonCodeRuntime, { addressSpaceMb: 1e308 }))
  92. .rejects.toThrow(/addressSpaceMb must be at most \d+ .*exact integer/)
  93. await expect(ctx.plugin(PythonCodeRuntime, { cpuSeconds: 1e100 }))
  94. .rejects.toThrow(/cpuSeconds must be at most \d+ .*exact integers/)
  95. // The boundary values still load: the bound rejects what cannot be encoded,
  96. // not everything large.
  97. const okMb = await ctx.plugin(PythonCodeRuntime, { addressSpaceMb: Math.floor(Number.MAX_SAFE_INTEGER / (1024 * 1024)) })
  98. await okMb.dispose()
  99. const okCpu = await ctx.plugin(PythonCodeRuntime, { cpuSeconds: Number.MAX_SAFE_INTEGER - 1 })
  100. await okCpu.dispose()
  101. })
  102. it('rejects an output cap whose payload could not cross the frame ceiling', async () => {
  103. // The caps budget a payload that must arrive inside ONE fd-3 frame, and the
  104. // 64 MiB frame parse cap is fixed. A larger cap is unsatisfiable rather
  105. // than generous: a completion the cap admits arrives as an over-ceiling
  106. // frame and fails the run as `worker-exit`, inverting the `output-limit`
  107. // the cap describes. Both budgets are metered in already-escaped serialized
  108. // bytes, so a payload occupies at most `cap + envelope` on the wire; the
  109. // bound is `parse-cap - envelope`, not `(ceiling - envelope) / 6` (that
  110. // divided in escape expansion the charge already counts). The receive path
  111. // rejects raw frames past the 64 MiB parse cap (the run settles as a
  112. // worker-exit), so a budget above it would admit a config whose honest
  113. // child frames the host then rejects.
  114. const admissible = 64 * 1024 * 1024 - 64
  115. const ctx = new Context()
  116. await expect(ctx.plugin(PythonCodeRuntime, { maxLogBytes: admissible + 1 }))
  117. .rejects.toThrow(/maxLogBytes must not exceed 67108800/)
  118. await expect(ctx.plugin(PythonCodeRuntime, { maxValueBytes: admissible + 1 }))
  119. .rejects.toThrow(/maxValueBytes must not exceed 67108800/)
  120. // The boundary value itself loads: the bound is the largest cap a frame can
  121. // still carry, not one below it. It needs an address space large enough to
  122. // clear the separate maxValueBytes/addressSpaceMb worst-case gate (the cap
  123. // times the 12x Unicode expansion must fit), so this pairs it with a 4 GiB
  124. // addressSpaceMb — the two load-time bounds are independent.
  125. const boundary = await ctx.plugin(PythonCodeRuntime, { maxValueBytes: admissible, addressSpaceMb: 4096 })
  126. await boundary.dispose()
  127. })
  128. it('rejects a pythonBin that spawn() would throw on, at load', async () => {
  129. // Both values pass the string schema and both make `spawn` throw
  130. // SYNCHRONOUSLY from inside run() — ERR_INVALID_ARG_VALUE for the empty
  131. // path, ERR_INVALID_ARG_TYPE for the NUL — so run() would REJECT instead of
  132. // resolving the worker-exit the seam promises for a child that cannot
  133. // start. Both are self-contained configuration errors, so they fail here.
  134. const ctx = new Context()
  135. await expect(ctx.plugin(PythonCodeRuntime, { pythonBin: '' }))
  136. .rejects.toThrow(/pythonBin must be a non-empty path without NUL bytes/)
  137. await expect(ctx.plugin(PythonCodeRuntime, { pythonBin: 'py\u0000thon3' }))
  138. .rejects.toThrow(/pythonBin must be a non-empty path without NUL bytes/)
  139. })
  140. it('skips relative PATH entries when resolving a basename pythonBin', async () => {
  141. // resolvePythonBin must return an absolute path: a RELATIVE PATH entry
  142. // ('.' here) would otherwise resolve the basename against the host CWD.
  143. // This run's CWD holds no executable named python3, so both the relative
  144. // skip and the accessSync-miss fall through to the absolute entry — the
  145. // case pins the contract (absolute candidate wins over a relative PATH
  146. // prefix), not a worker-exit distinction, which would need an executable
  147. // named python3 in the test CWD.
  148. const cp = await import('node:child_process')
  149. const nodePath = await import('node:path')
  150. const pythonDir = nodePath.dirname(cp.execFileSync('which', ['python3'], { encoding: 'utf8' }).trim())
  151. vi.stubEnv('PATH', `.:${pythonDir}`)
  152. try {
  153. const { runtime, fiber } = await setup({ pythonBin: 'python3', maxWallMs: 30_000 })
  154. const result = await runtime.run({ program: 'return 1', bindings: [] })
  155. expect(result.error).toBeUndefined()
  156. expect(result.value).toBe(1)
  157. await fiber.dispose()
  158. } finally {
  159. vi.unstubAllEnvs()
  160. }
  161. }, 45_000)
  162. it('ignores a forged second boot-ack without re-sending the run frame', async () => {
  163. // The run frame is sent once, from the first boot-ack; a program that
  164. // forges an extra boot-ack frame on fd 3 must not re-enter the gate (a
  165. // second run frame would confuse the child's frame reader). The honest
  166. // child sends exactly one ack; the forged one exercises the re-entry
  167. // guard.
  168. const { runtime } = await setup()
  169. const result = await runtime.run({
  170. program: [
  171. 'import os',
  172. // One forged boot-ack after the program starts; the run already went
  173. // out on the real ack.
  174. "os.write(3, b'{\"type\":\"boot-ack\"}\\n')",
  175. 'return "done"',
  176. ].join('\n'),
  177. bindings: [],
  178. })
  179. expect(result.error).toBeUndefined()
  180. expect(result.value).toBe('done')
  181. }, 15_000)
  182. it('skips a PATH entry that is an executable DIRECTORY named like the interpreter', async () => {
  183. // accessSync(X_OK) succeeds on directories, so without the isFile guard a
  184. // PATH entry like a `python3` directory would be chosen over a later real
  185. // interpreter. The stub PATH puts such a directory first and asserts the
  186. // real interpreter is used.
  187. const cp = await import('node:child_process')
  188. const nodePath = await import('node:path')
  189. const { mkdtempSync, mkdirSync } = await import('node:fs')
  190. const { tmpdir } = await import('node:os')
  191. const realPythonDir = nodePath.dirname(cp.execFileSync('which', ['python3'], { encoding: 'utf8' }).trim())
  192. const fakeDir = mkdtempSync(nodePath.join(tmpdir(), 'dsh-fake-bin-'))
  193. mkdirSync(nodePath.join(fakeDir, 'python3')) // A directory named python3, executable by default.
  194. vi.stubEnv('PATH', `${fakeDir}:${realPythonDir}`)
  195. try {
  196. const { runtime, fiber } = await setup({ pythonBin: 'python3', maxWallMs: 30_000 })
  197. const result = await runtime.run({ program: 'return 1', bindings: [] })
  198. expect(result.error).toBeUndefined()
  199. expect(result.value).toBe(1)
  200. await fiber.dispose()
  201. } finally {
  202. vi.unstubAllEnvs()
  203. }
  204. }, 45_000)
  205. it('rejects a timer budget setTimeout would silently clamp to 1 ms', async () => {
  206. // Node stores a setTimeout delay as a signed 32-bit value and substitutes
  207. // 1 ms for anything larger, inverting the knob's meaning: a huge maxWallMs
  208. // would time every run out at once, and a huge graceMs would SIGKILL one
  209. // millisecond after SIGTERM. Both must fail at load instead.
  210. const ctx = new Context()
  211. await expect(ctx.plugin(PythonCodeRuntime, { maxWallMs: 2_147_483_648 }))
  212. .rejects.toThrow(/maxWallMs must not exceed 2147483647/)
  213. // graceMs is bounded by the close deadline's added margin, not by the raw
  214. // timer maximum, because that sum is what gets armed.
  215. await expect(ctx.plugin(PythonCodeRuntime, { graceMs: 2_147_481_648 }))
  216. .rejects.toThrow(/graceMs must not exceed 2147481647/)
  217. // The exact maxima still load.
  218. await expect(ctx.plugin(PythonCodeRuntime, { maxWallMs: 2_147_483_647, graceMs: 2_147_481_647 }))
  219. .resolves.toBeDefined()
  220. })
  221. it('rejects loading this Unix-only backend on Windows', async () => {
  222. // The bootstrap needs the POSIX `resource` module, a positional fd 3, and
  223. // negative-PID process-group signals — none on Windows. The constructor
  224. // must throw at load rather than register ctx.codeRuntime and defer the
  225. // failure to the first run.
  226. const original = process.platform
  227. Object.defineProperty(process, 'platform', { value: 'win32', configurable: true })
  228. try {
  229. const ctx = new Context()
  230. await expect(ctx.plugin(PythonCodeRuntime, {})).rejects.toThrow(/requires a Unix platform/)
  231. } finally {
  232. Object.defineProperty(process, 'platform', { value: original, configurable: true })
  233. }
  234. })
  235. it('rejects a binding global that is not a Python identifier or is reserved', async () => {
  236. const { runtime } = await setup()
  237. await expect(runtime.run({
  238. program: 'return 1',
  239. bindings: [{ global: '1bad', functions: {} }],
  240. })).rejects.toThrow(/is not a usable Python identifier/)
  241. await expect(runtime.run({
  242. program: 'return 1',
  243. bindings: [{ global: 'class', functions: {} }],
  244. })).rejects.toThrow(/is not a usable Python identifier/)
  245. })
  246. it('rejects duplicate binding namespaces', async () => {
  247. const { runtime } = await setup()
  248. await expect(runtime.run({
  249. program: 'return 1',
  250. bindings: [
  251. { global: 'tools', functions: {} },
  252. { global: 'tools', functions: {} },
  253. ],
  254. })).rejects.toThrow(/duplicate binding global/)
  255. })
  256. it('rejects run() after disposal, and unregisters ctx.codeRuntime', async () => {
  257. const { ctx, fiber, runtime } = await setup()
  258. await fiber.dispose()
  259. await expect(runtime.run({ program: 'return 1', bindings: [] }))
  260. .rejects.toThrow(/after disposal/)
  261. expect(ctx.get('codeRuntime')).toBeUndefined()
  262. })
  263. it('short-circuits when the request signal is already aborted', async () => {
  264. const { runtime } = await setup()
  265. const signal = AbortSignal.abort('already-cancelled')
  266. const result = await runtime.run({ program: 'return 1', bindings: [], signal })
  267. expect(result.error?.kind).toBe('abort')
  268. expect(result.error?.message).toContain('already-cancelled')
  269. expect(result.logs).toEqual([])
  270. })
  271. it('short-circuits on an already-aborted signal whose reason cannot be converted', async () => {
  272. // The pre-flight arm converted the reason with a bare `String()`, so a
  273. // hostile reason threw out of `run()` — the seam promises to reject only for
  274. // misuse, and a caller's cancellation token is not misuse.
  275. const { runtime } = await setup()
  276. const signal = AbortSignal.abort({
  277. [Symbol.toPrimitive]() { throw new Error('reason blew up') },
  278. })
  279. const result = await runtime.run({ program: 'return 1', bindings: [], signal })
  280. expect(result.error?.kind).toBe('abort')
  281. expect(result.error?.message).toBe('<unrenderable rejection value>')
  282. expect(result.logs).toEqual([])
  283. })
  284. it('runs the interpreter from materialized scripts outside the package, and removes them per run', async () => {
  285. // The interpreter is an EXTERNAL process, so it can only open paths the OS
  286. // resolves. Inside the single-file Python-SDK executable the packaged `py/`
  287. // directory lives in pkg's virtual filesystem, which Node reads through its
  288. // patched `fs` but `python3` cannot see, so spawning from that path fails
  289. // with ENOENT. The scripts are therefore copied to a real directory first.
  290. //
  291. // The path is read from the child's own `__main__` module, so it proves
  292. // where the interpreter actually loaded the entry script — asserting on a
  293. // host-side constant would only restate the source. The program namespace
  294. // seeds `__name__` but no `__file__`, hence the module lookup.
  295. // `protocol.py` must land in the SAME directory, since `bootstrap.py` puts
  296. // its own directory on `sys.path` to import it; the run completing at all
  297. // already exercises that import.
  298. const { runtime } = await setup()
  299. const entryOf = async (): Promise<string> => {
  300. const result = await runtime.run({ program: 'import sys\nreturn sys.modules["__main__"].__file__', bindings: [] })
  301. expect(result.error).toBeUndefined()
  302. return result.value as string
  303. }
  304. const entry = await entryOf()
  305. expect(entry.endsWith('/bootstrap.py')).toBe(true)
  306. const dir = dirname(entry)
  307. expect(dir.startsWith(realpathSync(tmpdir()))).toBe(true)
  308. expect(dir).not.toContain('/packages/')
  309. // Staging is per RUN and removed at settlement, so by the time `run()`
  310. // resolved the directory is already gone — nothing survives to be rewritten
  311. // by a later run. `protocol.py` had to be beside the entry script for the run
  312. // to complete at all, since `bootstrap.py` imports it off `sys.path`.
  313. expect(existsSync(dir)).toBe(false)
  314. // A second run stages its own copy rather than reusing the first.
  315. expect(dirname(await entryOf())).not.toBe(dir)
  316. })
  317. it('contains a program that rewrites its own bootstrap to the run that did it', async () => {
  318. // The child runs as the same UID as the host, so `0o700` does not stop model
  319. // code from rewriting the scripts it was started from —
  320. // `sys.modules['__main__'].__file__` names them. While all runs shared one
  321. // staged copy, a program that overwrote `bootstrap.py` broke the NEXT run
  322. // (measured: it settled as `worker-exit`), and substituted code would have
  323. // run before the resource limits were applied.
  324. const { runtime } = await setup({ maxWallMs: 10_000 })
  325. const sabotage = await runtime.run({
  326. program: [
  327. 'import sys',
  328. 'path = sys.modules["__main__"].__file__',
  329. 'open(path, "w").write("raise SystemExit(1)\\n")',
  330. 'return path',
  331. ].join('\n'),
  332. bindings: [],
  333. })
  334. expect(sabotage.error).toBeUndefined()
  335. // The damage stayed inside the run that caused it.
  336. const after = await runtime.run({ program: 'return 1 + 1', bindings: [] })
  337. expect(after.error).toBeUndefined()
  338. expect(after.value).toBe(2)
  339. }, 20_000)
  340. it('leaves no subprocess or scripts behind when disposal races the first run', async () => {
  341. // Staging runs SYNCHRONOUSLY so no async boundary opens between `run()` and
  342. // the point where `execute` registers the run in `live` and installs the
  343. // abort listener. With an `await` there, a disposal landing in that window
  344. // saw an empty `live`, returned, removed the script directory, and let the
  345. // continuation spawn a subprocess after the fiber was gone.
  346. //
  347. // `dispose()` is called in the same synchronous turn as `run()`, with no
  348. // `await` between them, so it lands exactly in that window.
  349. //
  350. // The leak assertion compares before and after rather than requiring an
  351. // empty tmpdir: other tests in this file build runtimes they never dispose,
  352. // so only the directories this test adds are its own evidence.
  353. const staged = (): string[] =>
  354. readdirSync(realpathSync(tmpdir())).filter(name => name.startsWith('dsh-code-runtime-python-'))
  355. const before = new Set(staged())
  356. const { fiber, runtime } = await setup({ maxWallMs: 8_000 })
  357. const pending = runtime.run({ program: 'import time\nwhile True: time.sleep(0.1)', bindings: [] })
  358. const disposed = fiber.dispose()
  359. const result = await pending
  360. await disposed
  361. // Whatever the run reports, it must be terminal and must not be a success.
  362. expect(result.value).toBeUndefined()
  363. expect(['abort', 'worker-exit', 'timeout']).toContain(result.error?.kind)
  364. // Disposal is to quiescence, so this run's directory is gone once it
  365. // resolves, and nothing recreated it afterwards.
  366. expect(staged().filter(name => !before.has(name))).toEqual([])
  367. }, 15_000)
  368. it('settles as abort when the signal fires in the same turn as the first run', async () => {
  369. // Same window, the other listener. `addEventListener('abort')` does not
  370. // replay an event that already fired, so an abort landing before the
  371. // listener was installed used to be missed entirely and the program ran to
  372. // success or the wall ceiling instead of resolving as `abort`. Synchronous
  373. // staging keeps the pre-flight check and the listener in one turn, leaving
  374. // no gap for the signal to slip through.
  375. const { runtime } = await setup({ maxWallMs: 4_000, graceMs: 200 })
  376. const controller = new AbortController()
  377. const pending = runtime.run({
  378. program: 'import time\nwhile True: time.sleep(0.1)',
  379. bindings: [],
  380. signal: controller.signal,
  381. })
  382. controller.abort('same-turn-abort')
  383. const result = await pending
  384. expect(result.error?.kind).toBe('abort')
  385. expect(result.error?.message).toContain('same-turn-abort')
  386. }, 15_000)
  387. it('reports a staging failure as worker-exit instead of rejecting run()', async () => {
  388. // Staging touches the filesystem, so it can fail for reasons that are not
  389. // the caller's doing: a full or read-only temp filesystem, or a deployment
  390. // that failed to ship the packaged scripts. Those are SUBSTRATE failures,
  391. // the same class as a child that cannot start, and the seam reserves
  392. // rejection for misuse — so `run()` must resolve, not throw.
  393. //
  394. // `TMPDIR` is the honest lever: `mkdtempSync` builds its path from
  395. // `os.tmpdir()`, so pointing it at a path that is not a directory makes the
  396. // real call fail without stubbing the module under test.
  397. const previous = process.env.TMPDIR
  398. const notADirectory = join(await mkdtemp(join(tmpdir(), 'dsh-staging-')), 'file')
  399. await writeFile(notADirectory, '')
  400. process.env.TMPDIR = notADirectory
  401. try {
  402. const { runtime } = await setup()
  403. const result = await runtime.run({ program: 'return 1', bindings: [] })
  404. expect(result.error?.kind).toBe('worker-exit')
  405. expect(result.error?.message).toContain('failed to stage the python bootstrap')
  406. expect(result.logs).toEqual([])
  407. } finally {
  408. if (previous === undefined) delete process.env.TMPDIR
  409. else process.env.TMPDIR = previous
  410. }
  411. })
  412. it('leaves no staging directory behind when a script copy fails', async () => {
  413. // `mkdtempSync` succeeding and a later `copyFileSync` failing is its own
  414. // case: the directory exists but is only partially populated. Recording it
  415. // before the copies would leak it, because `run` retries staging on the next
  416. // call and overwrites the single recorded path — teardown could then remove
  417. // only the newest attempt. Staging must clean up its own partial directory.
  418. //
  419. // Only `copyFileSync` is stubbed, and only for the second script, so
  420. // `mkdtempSync` really runs and the directory under assertion is real.
  421. const staged = (): string[] =>
  422. readdirSync(realpathSync(tmpdir())).filter(name => name.startsWith('dsh-code-runtime-python-'))
  423. const before = new Set(staged())
  424. failNextCopyOf.value = 'protocol.py'
  425. try {
  426. const { runtime } = await setup()
  427. const result = await runtime.run({ program: 'return 1', bindings: [] })
  428. expect(result.error?.kind).toBe('worker-exit')
  429. expect(result.error?.message).toContain('failed to stage the python bootstrap')
  430. // The partial directory is gone, so nothing accumulates across retries.
  431. expect(staged().filter(name => !before.has(name))).toEqual([])
  432. } finally {
  433. failNextCopyOf.value = undefined
  434. }
  435. }, 15_000)
  436. })
  437. describe('PythonCodeRuntime — process identity', () => {
  438. it('reads a live process start time and distinguishes it from an absent pid', () => {
  439. // The teardown guard signals `-child.pid` with a RAW `process.kill`, which
  440. // (unlike `child.kill()`) has no handle check, so it would reach a recycled
  441. // pgid during the window between the leader being reaped and `close` firing.
  442. // A pid alone cannot separate the original from its replacement -- both
  443. // answer `kill(pid, 0)` -- so the guard compares START TIME, and this pins
  444. // that the reading is stable for one process and absent for a pid that
  445. // cannot be read.
  446. const own = readProcessStart(process.pid)
  447. if (process.platform === 'linux') {
  448. // Same process, two reads: the identity must be stable, or the guard would
  449. // refuse to signal its own live group.
  450. expect(own).toBeDefined()
  451. expect(readProcessStart(process.pid)).toBe(own)
  452. // Pid 0 is never a readable /proc entry, so the guard degrades to
  453. // undefined rather than throwing on a teardown path. This is also the
  454. // reading a REAPED leader produces -- its /proc entry is gone while the
  455. // group it led can still hold survivors -- so `undefined` must NOT be
  456. // treated as an identity mismatch. Reading it as one refused the SIGKILL
  457. // that the same-group survivor tests depend on, which is why they went red
  458. // on Linux while passing on Darwin (where the reader always returns
  459. // undefined and the guard is inert).
  460. expect(readProcessStart(0)).toBeUndefined()
  461. } else {
  462. // Darwin has no /proc: the reader reports undefined, and `killGroup`
  463. // signals the pgid without the identity re-check instead of paying a `ps`
  464. // fork per signal.
  465. expect(own).toBeUndefined()
  466. }
  467. })
  468. })
  469. describe('PythonCodeRuntime — inherited resource limits', () => {
  470. it('runs under an inherited hard limit tighter than addressSpaceMb', async () => {
  471. // An unprivileged process may lower a hard rlimit but never raise it. Under
  472. // a harness started with `ulimit -v` below `addressSpaceBytes`, requesting
  473. // the configured cap made `setrlimit` raise `ValueError` and every run
  474. // returned a bootstrap exception — even though the inherited limit is
  475. // STRONGER than the one asked for. The bootstrap clamps to the inherited
  476. // hard limit instead, so the run proceeds under the stricter bound.
  477. //
  478. // `pythonBin` is the honest lever: a wrapper that lowers RLIMIT_AS and then
  479. // execs the real interpreter reproduces the inherited-limit condition
  480. // without touching this test process's own limits.
  481. const dir = await mkdtemp(join(tmpdir(), 'dsh-rlimit-'))
  482. const wrapper = join(dir, 'python3-capped')
  483. // 256 MiB, half the 512 MiB addressSpaceMb default, so the requested cap is
  484. // unambiguously above the inherited ceiling.
  485. await writeFile(wrapper, `#!/bin/sh\nulimit -v 262144\nexec "${PYABS}" "$@"\n`, { mode: 0o755 })
  486. const { runtime } = await setup({ pythonBin: wrapper })
  487. const result = await runtime.run({
  488. program: 'import resource\nreturn resource.getrlimit(resource.RLIMIT_AS)[1]',
  489. bindings: [],
  490. })
  491. expect(result.error).toBeUndefined()
  492. // The applied hard limit is the inherited one, not the configured 512 MiB.
  493. expect(result.value).toBe(256 * 1024 * 1024)
  494. }, 15_000)
  495. it('rejects at boot when an inherited RLIMIT_AS is too tight for the output budgets', async () => {
  496. // The host gate validates the output budgets against the CONFIGURED
  497. // addressSpaceMb, but a launch environment can inherit a STRICTER RLIMIT_AS
  498. // (a `ulimit -v` wrapper below addressSpaceMb), which the bootstrap clamps the
  499. // effective limit down to — leaving the budgets sized for a ceiling the child
  500. // never gets, so a near-budget output would OOM mid-run as an opaque
  501. // worker-exit. The bootstrap re-checks both budgets against the EFFECTIVE
  502. // clamped limit and fails loud at boot instead. A 128 MiB inherited limit
  503. // leaves 64 MiB budgetable (~5 MiB admissible under the 12x multiple), under
  504. // which a 32 MiB maxLogBytes — admitted by the 512 MiB configured default — is
  505. // rejected. The rejection surfaces as an 'exception' (bootstrap's
  506. // setrlimit-phase failure class), not a mid-run OOM. The repro is Linux-only
  507. // (macOS ignores `ulimit -v`); there the run proceeds.
  508. const dir = await mkdtemp(join(tmpdir(), 'dsh-rlimit-'))
  509. const wrapper = join(dir, 'python3-tight')
  510. await writeFile(wrapper, `#!/bin/sh\nulimit -v 131072\nexec "${PYABS}" "$@"\n`, { mode: 0o755 })
  511. const { runtime } = await setup({ pythonBin: wrapper, maxLogBytes: 32 * 1024 * 1024, addressSpaceMb: 512 })
  512. const result = await runtime.run({ program: 'return 1', bindings: [] })
  513. if (process.platform === 'darwin') {
  514. expect(result.error).toBeUndefined()
  515. } else {
  516. // The re-check raises inside bootstrap's resource-limit block, which
  517. // reports every setrlimit-phase failure as kind 'exception'; the message
  518. // discriminates this config rejection from a generic setrlimit error.
  519. expect(result.error?.kind).toBe('exception')
  520. expect(result.error?.message).toContain('too large for the inherited RLIMIT_AS')
  521. }
  522. }, 15_000)
  523. it('applies the configured limits when nothing tighter is inherited', async () => {
  524. // The clamp must not weaken the normal path: with an infinite inherited hard
  525. // limit there is nothing to clamp against, and RLIM_INFINITY compares as -1,
  526. // so treating it as a numeric bound would collapse every limit to -1.
  527. const { runtime } = await setup({ cpuSeconds: 42, addressSpaceMb: 400 })
  528. const result = await runtime.run({
  529. // `getrlimit` returns a tuple, which the lossless-JSON completion check
  530. // rejects; the pair is listed explicitly rather than converted.
  531. program: 'import resource\ncpu = resource.getrlimit(resource.RLIMIT_CPU)\nreturn [cpu[0], cpu[1], resource.getrlimit(resource.RLIMIT_AS)[1]]',
  532. bindings: [],
  533. })
  534. expect(result.error).toBeUndefined()
  535. // Soft at cpuSeconds, hard at +1 (the SIGKILL backstop), address space at
  536. // the configured megabytes — exactly what the unclamped path applied.
  537. expect(result.value).toEqual([42, 43, 400 * 1024 * 1024])
  538. }, 15_000)
  539. it('preserves an inherited soft limit stricter than the configured cap', async () => {
  540. // Clamping reads BOTH inherited bounds, not just the hard one. A deployment
  541. // that inherited a soft rlimit below the configured cap must keep that
  542. // stricter soft: returning the configured value would RAISE the effective
  543. // soft limit, loosening containment. The wrapper lowers only the SOFT CPU
  544. // limit (`ulimit -S -t`) and leaves the hard limit unlimited, so the
  545. // requested soft (`cpuSeconds`) sits above the inherited soft — the case that
  546. // exposed the bug. RLIMIT_CPU is used because macOS ignores `ulimit -v`
  547. // (RLIMIT_AS), which is exactly why the backend skips address space there.
  548. const dir = await mkdtemp(join(tmpdir(), 'dsh-rlimit-soft-'))
  549. const wrapper = join(dir, 'python3-soft-capped')
  550. // Soft CPU 5 s, well below the configured 30 s, hard left unlimited.
  551. await writeFile(wrapper, `#!/bin/sh\nulimit -S -t 5\nexec "${PYABS}" "$@"\n`, { mode: 0o755 })
  552. const { runtime } = await setup({ pythonBin: wrapper, cpuSeconds: 30 })
  553. const result = await runtime.run({
  554. program: 'import resource\nreturn resource.getrlimit(resource.RLIMIT_CPU)[0]',
  555. bindings: [],
  556. })
  557. expect(result.error).toBeUndefined()
  558. // The applied SOFT limit is the inherited 5 s, not the configured 30 s.
  559. expect(result.value).toBe(5)
  560. }, 15_000)
  561. it('reports a CPU overrun under a dual-limit ulimit as a timeout, not a worker-exit', async () => {
  562. // `ulimit -t N` sets BOTH the soft and hard CPU limit to N. The kernel
  563. // checks the hard limit and SIGKILLs a busy loop directly, so with
  564. // soft == hard the SIGXCPU signal is never delivered — and the host
  565. // classifies a CPU overrun ONLY on `signal === 'SIGXCPU'`, so the overrun
  566. // would be misreported as a `worker-exit` instead of a timeout. `_clamped`
  567. // now lowers a clamped soft==hard result by one unit (when hard >= 2), so
  568. // the SIGXCPU signal fires at the softer limit and the run reports a
  569. // timeout. This uses `ulimit -t 2` (hard == 2, so the soft is lowered to 1)
  570. // and leaves SIGXCPU unhandled, so the kernel terminates the busy loop at
  571. // 1 s with SIGXCPU and the host classifies it as a timeout.
  572. const dir = await mkdtemp(join(tmpdir(), 'dsh-rlimit-dual-'))
  573. const wrapper = join(dir, 'python3-dual-capped')
  574. // Both soft and hard CPU 2 s; configured cpuSeconds 30 s.
  575. await writeFile(wrapper, `#!/bin/sh\nulimit -t 2\nexec "${PYABS}" "$@"\n`, { mode: 0o755 })
  576. const { runtime } = await setup({ pythonBin: wrapper, cpuSeconds: 30, maxWallMs: 12_000 })
  577. const result = await runtime.run({
  578. program: [
  579. 'while True:',
  580. ' pass',
  581. 'return "unreachable"',
  582. ].join('\n'),
  583. bindings: [],
  584. })
  585. expect(result.error?.kind).toBe('timeout')
  586. expect(result.error?.message).toContain('CPU time exhausted')
  587. }, 15_000)
  588. it('reports a timeout when a program masks SIGXCPU and returns past the soft limit', async () => {
  589. // A program can mask SIGXCPU (pthread_sigmask SIG_BLOCK), burn past the
  590. // soft CPU limit, and return during the soft-to-hard gap. The settlement
  591. // recheck (`die_if_cpu_exhausted`) must UNBLOCK the signal before
  592. // re-delivering it, or the SIGXCPU stays pending and the child exits
  593. // normally with a success result. With the unblock, the re-delivered
  594. // SIGXCPU (default disposition) terminates the child and the host
  595. // classifies the run as a timeout. Fail-before: without the unblock the
  596. // run reports `value: "escaped"` and no error. The masking is guarded by
  597. // hasattr so the case is a no-op on platforms without pthread_sigmask.
  598. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 12_000 })
  599. const result = await runtime.run({
  600. program: [
  601. 'import signal, time',
  602. 'if hasattr(signal, "pthread_sigmask"):',
  603. ' signal.pthread_sigmask(signal.SIG_BLOCK, {signal.SIGXCPU})',
  604. 'end = time.process_time() + 1.05',
  605. 'while time.process_time() < end:',
  606. ' pass',
  607. 'return "escaped"',
  608. ].join('\n'),
  609. bindings: [],
  610. })
  611. expect(result.error?.kind).toBe('timeout')
  612. expect(result.value).toBeUndefined()
  613. }, 20_000)
  614. it('reports a timeout when the interpreter was started with SIGXCPU ignored (inherited state)', async () => {
  615. // The child inherits the host's SIGXCPU disposition: a wrapper that
  616. // ignores SIGXCPU before exec'ing python3 hands the child a soft
  617. // RLIMIT_CPU that cannot stop it. The bootstrap resets SIGXCPU to SIG_DFL
  618. // before model code runs, so a busy loop still ends as a timeout rather
  619. // than running to the hard limit and being misclassified as worker-exit.
  620. const wrapper = join(tmpdir(), `dsh-xcpu-ignore-${process.pid}.sh`)
  621. writeFileSync(wrapper, `#!/bin/sh\ntrap "" XCPU\nexec "${PYABS}" "$@"\n`, { mode: 0o755 })
  622. try {
  623. const { runtime } = await setup({ maxWallMs: 30_000, cpuSeconds: 1, pythonBin: wrapper })
  624. const result = await runtime.run({
  625. program: ['while True: pass'].join('\n'),
  626. bindings: [],
  627. })
  628. expect(result.error?.kind).toBe('timeout')
  629. } finally {
  630. rmSync(wrapper, { force: true })
  631. }
  632. }, 20_000)
  633. it('reports a timeout when a program traps AND masks SIGXCPU and returns past the soft limit', async () => {
  634. // The mask-only case exercises the unblock; the trap+mask combination is
  635. // the harder one: a program that installed a custom handler AND masked the
  636. // signal has that PENDING handler run the moment the signal is unblocked
  637. // (CPython delivers it at the next eval-breaker checkpoint in model code),
  638. // and the handler re-masks — so the settlement recheck must restore the
  639. // default disposition BEFORE unblocking. With SIG_DFL restored first, the
  640. // pending signal kills the process inside the kernel with no bytecode
  641. // window; without it, the handler re-blocks and the child exits normally
  642. // with a success value. Fail-before: the run reports `value: "escaped"`.
  643. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 12_000 })
  644. const result = await runtime.run({
  645. program: [
  646. 'import signal, time',
  647. 'if hasattr(signal, "pthread_sigmask"):',
  648. ' def h(signum, frame):',
  649. ' signal.pthread_sigmask(signal.SIG_BLOCK, {signal.SIGXCPU})',
  650. ' signal.signal(signal.SIGXCPU, h)',
  651. ' signal.pthread_sigmask(signal.SIG_BLOCK, {signal.SIGXCPU})',
  652. ' end = time.process_time() + 1.05',
  653. ' while time.process_time() < end:',
  654. ' pass',
  655. 'return "escaped"',
  656. ].join('\n'),
  657. bindings: [],
  658. })
  659. expect(result.error?.kind).toBe('timeout')
  660. expect(result.value).toBeUndefined()
  661. }, 20_000)
  662. it('rechecks CPU at settlement against the effective inherited soft limit', async () => {
  663. // The settlement-time CPU recheck must compare against the EFFECTIVE soft
  664. // limit (`_clamped` may have lowered it to a stricter inherited value), not
  665. // the configured `cpuSeconds`. A program that traps SIGXCPU, burns past the
  666. // inherited soft, and returns inside the soft-to-hard gap would otherwise be
  667. // compared to the configured value and falsely reported successful, bypassing
  668. // the inherited limit. The wrapper sets a 1 s soft CPU limit; the program
  669. // traps SIGXCPU and busy-loops past it, then returns — the recheck must
  670. // re-deliver SIGXCPU so the host classifies the run as a timeout.
  671. const dir = await mkdtemp(join(tmpdir(), 'dsh-cpu-recheck-'))
  672. const wrapper = join(dir, 'python3-cpu-capped')
  673. await writeFile(wrapper, `#!/bin/sh\nulimit -S -t 1\nexec "${PYABS}" "$@"\n`, { mode: 0o755 })
  674. const { runtime } = await setup({ pythonBin: wrapper, cpuSeconds: 30, maxWallMs: 12_000 })
  675. const result = await runtime.run({
  676. program: [
  677. 'import signal, time',
  678. // Trap SIGXCPU so the soft limit does not terminate the program; burn
  679. // CPU well past the inherited 1 s soft, then return normally.
  680. 'signal.signal(signal.SIGXCPU, lambda *a: None)',
  681. 'end = time.process_time() + 2.5',
  682. 'while time.process_time() < end:',
  683. ' pass',
  684. 'return "returned"',
  685. ].join('\n'),
  686. bindings: [],
  687. })
  688. // The recheck compares spent CPU against the effective 1 s soft, not 30 s, so
  689. // the run is a timeout rather than a false success.
  690. expect(result.error?.kind).toBe('timeout')
  691. }, 20_000)
  692. })
  693. describe('PythonCodeRuntime — programs and bindings', () => {
  694. it('runs a top-level script, captures print output, and returns `result`', async () => {
  695. const { runtime } = await setup()
  696. const result = await runtime.run({
  697. program: [
  698. 'x = 40',
  699. 'y = 2',
  700. 'print("hello", x + y)',
  701. 'return {"answer": x + y}',
  702. ].join('\n'),
  703. bindings: [],
  704. })
  705. expect(result.error).toBeUndefined()
  706. expect(result.value).toEqual({ answer: 42 })
  707. // `print` in Python emits: text, ' ', text, '\n'. Concat the captured
  708. // fragments and assert the model-visible message survives.
  709. expect(result.logs.join('')).toContain('hello 42')
  710. // 15s: this is usually the suite's first real subprocess — a cold python3
  711. // start (interpreter + asyncio import) on a loaded CI runner can exceed
  712. // the 5s default alone; later tests reuse the warm page cache.
  713. }, 15_000)
  714. it('bridges binding calls both ways and rejects the program-side call on a host rejection', async () => {
  715. const { runtime } = await setup()
  716. const calls: unknown[] = []
  717. const result = await runtime.run({
  718. program: [
  719. 'first = await tools.echo({"n": 1})',
  720. 'caught = ""',
  721. 'try:',
  722. ' await tools.fail({})',
  723. 'except RuntimeError as e:',
  724. ' caught = str(e)',
  725. 'return {"first": first, "caught": caught}',
  726. ].join('\n'),
  727. bindings: tools({
  728. echo: async (args) => { calls.push(args); return { echoed: args as CodeJsonValue } },
  729. fail: async () => { throw new Error('nope') },
  730. }),
  731. })
  732. expect(result.error).toBeUndefined()
  733. expect(result.value).toEqual({ first: { echoed: { n: 1 } }, caught: 'nope' })
  734. expect(calls).toEqual([{ n: 1 }])
  735. })
  736. it('keeps decoding binding replies when _decode_json_plain is rebound', async () => {
  737. // read_frame_async resolves _decode_json_plain at call time; a program that
  738. // rebinds __main__._decode_json_plain would otherwise kill the reply pump
  739. // (a broken decode strands every pending Future to the wall clock). The
  740. // decode primitives are def-time captures on the channel methods, so a
  741. // rebind cannot break reply delivery.
  742. const { runtime } = await setup()
  743. const result = await runtime.run({
  744. program: [
  745. 'import __main__',
  746. '__main__._decode_json_plain = None',
  747. 'first = await tools.echo({"n": 1})',
  748. 'return first',
  749. ].join('\n'),
  750. bindings: tools({
  751. echo: async args => ({ echoed: args as CodeJsonValue }),
  752. }),
  753. })
  754. expect(result.error).toBeUndefined()
  755. expect(result.value).toEqual({ echoed: { n: 1 } })
  756. }, 15_000)
  757. it('keeps dispatch working when _lossless_json_violation, asyncio, and send_sync are rebound', async () => {
  758. // dispatch binds _lossless_json_violation, asyncio.get_event_loop, and the
  759. // channel's send method into _run locals before the program runs, so a
  760. // rebind of __main__._lossless_json_violation/__main__.asyncio/
  761. // __main__.ProtocolChannel.send_sync cannot turn a legitimate binding call
  762. // into an exception or a wall-clock timeout.
  763. const { runtime } = await setup()
  764. const result = await runtime.run({
  765. program: [
  766. 'import __main__',
  767. 'def boom(*a, **k):',
  768. ' raise RuntimeError("hijacked")',
  769. '__main__._lossless_json_violation = boom',
  770. '__main__.asyncio = boom',
  771. '__main__.ProtocolChannel.send_sync = boom',
  772. '__main__._encode_json_plain = boom',
  773. '__main__.ProtocolChannel.write_encoded = boom',
  774. 'first = await tools.echo({"n": 1})',
  775. 'return first',
  776. ].join('\n'),
  777. bindings: tools({
  778. echo: async args => ({ echoed: args as CodeJsonValue }),
  779. }),
  780. })
  781. expect(result.error).toBeUndefined()
  782. expect(result.value).toEqual({ echoed: { n: 1 } })
  783. }, 15_000)
  784. it('keeps the reply pump reading when the read_frame_async class attribute is rebound', async () => {
  785. // _pump_replies' frame reader is a bound method captured by _run before the
  786. // program runs and passed in as an explicit argument, so a program rebinding
  787. // `__main__.ProtocolChannel.read_frame_async` cannot redirect the pump (a
  788. // body-local `channel.read_frame_async` lookup would resolve the rebound
  789. // class attribute, since the pump starts after the program's top-level
  790. // statements).
  791. const { runtime } = await setup()
  792. const result = await runtime.run({
  793. program: [
  794. 'import __main__',
  795. 'async def boom(*a, **k):',
  796. ' raise RuntimeError("hijacked reader")',
  797. '__main__.ProtocolChannel.read_frame_async = boom',
  798. 'first = await tools.echo({"n": 1})',
  799. 'return first',
  800. ].join('\n'),
  801. bindings: tools({
  802. echo: async args => ({ echoed: args as CodeJsonValue }),
  803. }),
  804. })
  805. expect(result.error).toBeUndefined()
  806. expect(result.value).toEqual({ echoed: { n: 1 } })
  807. }, 15_000)
  808. it('keeps the rejection contract when _BindingRejection is rebound', async () => {
  809. // `dispatch`'s except clause resolves `_BindingRejection` at call time; a
  810. // program that rebinds `__main__._BindingRejection = ValueError` would
  811. // otherwise let the internal marker type leak into model code (the program
  812. // would catch a `ValueError` for a host rejection). The class is now bound
  813. // into `_run` locals before the program runs, so a host rejection still
  814. // surfaces as the declared `RuntimeError`.
  815. const { runtime } = await setup()
  816. const result = await runtime.run({
  817. program: [
  818. 'import __main__',
  819. '__main__._BindingRejection = ValueError',
  820. 'caught = ""',
  821. 'try:',
  822. ' await tools.fail({})',
  823. 'except RuntimeError as e:',
  824. ' caught = e.args[0] if e.args else ""',
  825. 'except Exception as e:',
  826. ' caught = "WRONG TYPE: " + type(e).__name__',
  827. 'return caught',
  828. ].join('\n'),
  829. bindings: tools({
  830. fail: async () => { throw new Error('nope') },
  831. }),
  832. })
  833. expect(result.error).toBeUndefined()
  834. expect(result.value).toBe('nope')
  835. }, 15_000)
  836. it('still answers the call when the rejection value cannot be converted to a string', async () => {
  837. // `messageOf` calls `String(error)`, which runs the value's own conversion,
  838. // and this call site is a DETACHED async reply callback. A rejection whose
  839. // `Symbol.toPrimitive` throws therefore escaped as an unhandled rejection:
  840. // the reply frame was never written, the program stayed blocked on `await`,
  841. // and the run degraded to a `maxWallMs` timeout (observed) — a host with no
  842. // `unhandledRejection` listener would exit instead. The rejection must reach
  843. // the program as an ordinary error carrying a fixed placeholder.
  844. const { runtime } = await setup({ maxWallMs: 8_000 })
  845. const result = await runtime.run({
  846. program: [
  847. 'try:',
  848. ' await tools.hostile({})',
  849. 'except RuntimeError as e:',
  850. ' return "rejected: " + str(e)',
  851. 'return "no rejection"',
  852. ].join('\n'),
  853. bindings: tools({
  854. hostile: async () => {
  855. throw { [Symbol.toPrimitive]() { throw new Error('toPrimitive blew up') } }
  856. },
  857. }),
  858. })
  859. expect(result.error).toBeUndefined()
  860. expect(result.value).toBe('rejected: <unrenderable rejection value>')
  861. }, 15_000)
  862. it('still answers the call when an Error carries a cyclic value in place of its message', async () => {
  863. // `Error.message` is typed `string` but is a plain writable property, so a
  864. // rejection can carry any value there. Returning it verbatim handed a
  865. // non-string to `sendReply`, breaching `encodeJsonPlain`'s JSON-plain
  866. // precondition: a cyclic object grew the encoder stack until the host threw
  867. // RangeError from the detached reply callback, so no reply frame was written
  868. // and the run degraded to a `maxWallMs` timeout (observed). The conversion
  869. // must contain it — `String()` on a cycle throws inside the guard and lands
  870. // on the placeholder, so the program sees an ordinary error.
  871. const { runtime } = await setup({ maxWallMs: 8_000 })
  872. const result = await runtime.run({
  873. program: [
  874. 'try:',
  875. ' await tools.hostile({})',
  876. 'except RuntimeError as e:',
  877. ' return "rejected: " + str(e)',
  878. 'return "no rejection"',
  879. ].join('\n'),
  880. bindings: tools({
  881. hostile: async () => {
  882. const cyclic: { self?: unknown; [Symbol.toPrimitive]: () => string } = {
  883. // A cycle alone is inert for `String()`; the throwing conversion is
  884. // what proves the guard runs rather than the encoder.
  885. [Symbol.toPrimitive]: () => { throw new Error('cyclic message') },
  886. }
  887. cyclic.self = cyclic
  888. const error = new Error('placeholder')
  889. // Writable per spec, so no cast is needed to install a non-string.
  890. ;(error as unknown as { message: unknown }).message = cyclic
  891. throw error
  892. },
  893. }),
  894. })
  895. expect(result.error).toBeUndefined()
  896. expect(result.value).toBe('rejected: <unrenderable rejection value>')
  897. }, 15_000)
  898. it('renders an Error whose message is a value with no JSON form', async () => {
  899. // The non-cyclic arm. A number would not discriminate: `scalarJson` renders
  900. // it as digits and the child `str()`s the field back, so it survives the
  901. // wire either way. `undefined` is the value that separates the two orders —
  902. // `scalarJson` emits a bare `undefined` token, so the reply line is not JSON
  903. // at all, the child's parse drops the frame, and the program stays blocked
  904. // on `await` until the wall ceiling (observed). Converting first sends the
  905. // string "undefined", which the program receives as an ordinary rejection.
  906. const { runtime } = await setup({ maxWallMs: 8_000 })
  907. const result = await runtime.run({
  908. program: [
  909. 'try:',
  910. ' await tools.absent({})',
  911. 'except RuntimeError as e:',
  912. ' return "rejected: " + str(e)',
  913. 'return "no rejection"',
  914. ].join('\n'),
  915. bindings: tools({
  916. absent: async () => {
  917. const error = new Error('placeholder')
  918. ;(error as unknown as { message: unknown }).message = undefined
  919. throw error
  920. },
  921. }),
  922. })
  923. expect(result.error).toBeUndefined()
  924. expect(result.value).toBe('rejected: undefined')
  925. }, 15_000)
  926. it('runs a program with no await', async () => {
  927. const { runtime } = await setup()
  928. const result = await runtime.run({
  929. program: 'return 2 + 2',
  930. bindings: [],
  931. })
  932. expect(result.error).toBeUndefined()
  933. expect(result.value).toBe(4)
  934. })
  935. it('returns JSON null whether the program returns None or falls off the end', async () => {
  936. // Python has no `undefined`: an async body that returns None and one that
  937. // never returns both yield None, so both complete as an exact JSON null.
  938. // (The worker/TS backend can tell `return undefined` from `return null`;
  939. // Python cannot, and reporting null for both is the honest rendering.)
  940. const { runtime } = await setup()
  941. const explicit = await runtime.run({ program: 'return None', bindings: [] })
  942. expect(explicit.error).toBeUndefined()
  943. expect(explicit.value).toBeNull()
  944. const noReturn = await runtime.run({ program: 'x = 1', bindings: [] })
  945. expect(noReturn.error).toBeUndefined()
  946. expect(noReturn.value).toBeNull()
  947. })
  948. it('settles with no value on a forged valueless done frame', async () => {
  949. // The child always sends a value now (return None → JSON null), so a done
  950. // frame with no value key can only be forged; the host settles it as a
  951. // value-less completion rather than crashing on the absent field.
  952. const { runtime } = await setup()
  953. const result = await runtime.run({
  954. program: [
  955. 'import os',
  956. 'os.write(3, b\'{"type":"done"}\\n\')',
  957. 'import time',
  958. 'time.sleep(5)',
  959. ].join('\n'),
  960. bindings: [],
  961. })
  962. expect(result.error).toBeUndefined()
  963. expect(result.value).toBeUndefined()
  964. })
  965. it('coalesces print arguments into one log line, not per-write fragments', async () => {
  966. // print("a","b") calls write() per arg/sep/newline; the stream must emit
  967. // one logical line "a b" so Code Mode's join(newline) does not insert
  968. // spurious blank lines. Two prints → exactly two entries, no empties.
  969. const { runtime } = await setup()
  970. const result = await runtime.run({
  971. program: ['print("a", "b")', 'print("c")', 'return None'].join('\n'),
  972. bindings: [],
  973. })
  974. expect(result.error).toBeUndefined()
  975. expect(result.logs).toEqual(['a b', 'c'])
  976. })
  977. it('flushes a print with no trailing newline', async () => {
  978. const { runtime } = await setup()
  979. const result = await runtime.run({
  980. program: ['print("partial", end="")', 'return None'].join('\n'),
  981. bindings: [],
  982. })
  983. expect(result.error).toBeUndefined()
  984. expect(result.logs).toEqual(['partial'])
  985. })
  986. it('aggregates a large newline-free native write into one log entry, not one per pipe chunk', async () => {
  987. // A single `os.write` larger than one pipe read arrives as several Node
  988. // `data` chunks. `logs` entries are joined with `\n` downstream, so pushing
  989. // one entry per transport chunk would insert model-visible newlines at
  990. // arbitrary pipe boundaries inside one native write. Stray capture holds a
  991. // per-stream residual and admits only on a real `\n`, so a 200 KiB blast
  992. // with no newline reads back as exactly one entry with no interior breaks.
  993. const { runtime } = await setup({ maxLogBytes: 300_000 })
  994. const size = 200_000
  995. const result = await runtime.run({
  996. program: ['import os', `os.write(1, b"A" * ${size})`, 'return None'].join('\n'),
  997. bindings: [],
  998. })
  999. expect(result.error).toBeUndefined()
  1000. expect(result.logs).toEqual(['A'.repeat(size)])
  1001. })
  1002. it('splits native output on its own newlines, one entry per line', async () => {
  1003. // The complement of the aggregation case: real newlines in a native write
  1004. // still delimit entries, matching the child's line-granular `log` frames.
  1005. const { runtime } = await setup()
  1006. const result = await runtime.run({
  1007. program: ['import os', 'os.write(1, b"one\\ntwo\\nthree")', 'return None'].join('\n'),
  1008. bindings: [],
  1009. })
  1010. expect(result.error).toBeUndefined()
  1011. expect(result.logs).toEqual(['one', 'two', 'three'])
  1012. })
  1013. it('bounds a newline-free native flood by the ledger instead of buffering it whole', async () => {
  1014. // A newline-free write far larger than maxLogBytes must not accumulate in
  1015. // the host-side residual: when the pending residual would cross the budget
  1016. // it is admitted (and truncated) immediately, and once the ledger has
  1017. // truncated, later chunks stop buffering entirely. The run still completes
  1018. // and the captured output ends at the truncation marker rather than
  1019. // retaining the whole flood.
  1020. const { runtime } = await setup({ maxLogBytes: 4096 })
  1021. const result = await runtime.run({
  1022. program: ['import os', 'os.write(1, b"A" * 2_000_000)', 'return None'].join('\n'),
  1023. bindings: [],
  1024. })
  1025. expect(result.error).toBeUndefined()
  1026. expect(result.logs.at(-1)).toBe(logTruncationMarker(4096))
  1027. // The retained output is bounded by the budget, not the 2 MB flood.
  1028. expect(result.logs.join('').length).toBeLessThan(4096)
  1029. })
  1030. it('bounds a newline-free single-character Python write drip by the fragment cap, not OOM', async () => {
  1031. // The child-side `_LogStream` buffers one fragment per `write` (so
  1032. // `print("x", end="")` does not concatenate quadratically). A newline-free
  1033. // drip of one character per call past a large `maxLogBytes` would otherwise
  1034. // accumulate one list slot (and one str object) per call — 25 M calls =
  1035. // ~25 M slots, which OOMs the host on its own accounting before the byte
  1036. // budget is reached. The stream seals the fragment list past
  1037. // `_PENDING_MAX_CHUNKS` into one joined block (character count unchanged),
  1038. // bounding the live fragment count exactly as the host-side `captureStray`
  1039. // seal does. This drives well past the cap and asserts the run still
  1040. // completes with a truncation marker rather than a MemoryError.
  1041. const { runtime } = await setup({ maxLogBytes: 4096 })
  1042. const result = await runtime.run({
  1043. program: [
  1044. 'import sys',
  1045. 'for _ in range(200_000):',
  1046. ' sys.stdout.write("x")',
  1047. 'return None',
  1048. ].join('\n'),
  1049. bindings: [],
  1050. })
  1051. expect(result.error).toBeUndefined()
  1052. expect(result.logs.at(-1)).toBe(logTruncationMarker(4096))
  1053. })
  1054. it('bounds a control-char-dense native residual by serialized cost, not raw length', async () => {
  1055. // A newline-free NUL flood passes the cheap `length + 3` lower bound at a
  1056. // raw length well under the budget, but each NUL serializes to `\u0000` (6
  1057. // bytes), so the true JSON cost is ~6x. The ledger must charge that
  1058. // serialized cost — and `jsonStringCostUpTo` must measure it WITHOUT
  1059. // allocating the escaped copy, so a near-budget line under a large
  1060. // maxLogBytes cannot momentarily allocate a multi-gigabyte `JSON.stringify`
  1061. // result. Under a small budget the residual is truncated once the serialized
  1062. // cost crosses it.
  1063. const { runtime } = await setup({ maxLogBytes: 4096 })
  1064. const result = await runtime.run({
  1065. program: ['import os', 'os.write(1, b"\\x00" * 4000)', 'return None'].join('\n'),
  1066. bindings: [],
  1067. })
  1068. expect(result.error).toBeUndefined()
  1069. expect(result.logs.at(-1)).toBe(logTruncationMarker(4096))
  1070. })
  1071. it('rejects an output budget that could breach addressSpaceMb during encode at load', async () => {
  1072. // The child builds, charges, and encodes a `maxLogBytes` log entry or a
  1073. // `maxValueBytes` completion value under RLIMIT_AS, and both trigger on
  1074. // character count against a serialized-byte budget — an astral character is
  1075. // one character but ~4 bytes stored and ~4 encoded, and THREE such copies are
  1076. // live at the peak (the caller's write argument, the slice/join handed to
  1077. // push, and the encode copy), so a budget approaching the address space lets a
  1078. // legitimate near-budget output breach it and die as worker-exit. The
  1079. // incompatible pair is rejected at load: each budget times the worst-case
  1080. // multiple (12) must fit the address space LEFT after the fixed interpreter
  1081. // baseline. Against a 256 MiB address space that leaves 192 MiB budgetable
  1082. // (~16 MiB admissible), so a 50 MB cap is far over; the default caps against
  1083. // 512 MiB are not. Both budgets are gated symmetrically — the value case sets
  1084. // a default-fitting maxLogBytes so the maxValueBytes check is what fires.
  1085. const ctxLog = new Context()
  1086. await expect(ctxLog.plugin(PythonCodeRuntime, { maxLogBytes: 50_000_000, addressSpaceMb: 256 }))
  1087. .rejects.toThrow(/maxLogBytes times the 12x worst-case Unicode expansion must fit/)
  1088. const ctxValue = new Context()
  1089. await expect(ctxValue.plugin(PythonCodeRuntime, { maxValueBytes: 50_000_000, addressSpaceMb: 256 }))
  1090. .rejects.toThrow(/maxValueBytes times the 12x worst-case Unicode expansion must fit/)
  1091. // Discriminates 12 from 8: a 48 MiB maxLogBytes against a 512 MiB address
  1092. // space leaves 448 MiB budgetable. 48*8 = 384 MiB fits (the old 8x multiple
  1093. // wrongly ADMITTED this), but 48*12 = 576 MiB does not. The ~12x peak this
  1094. // guards is the NEWLINE path's single near-budget write — the caller's own
  1095. // string, the line slice, and the encode copy live at once. The settlement
  1096. // flush is no longer the binding case: `flush_line` drops the pending chunks
  1097. // before its push, so it holds two copies, not three.
  1098. const ctxTwelve = new Context()
  1099. await expect(ctxTwelve.plugin(PythonCodeRuntime, { maxLogBytes: 48 * 1024 * 1024, addressSpaceMb: 512 }))
  1100. .rejects.toThrow(/maxLogBytes times the 12x worst-case Unicode expansion must fit/)
  1101. // An addressSpaceMb at or below the interpreter baseline leaves nothing
  1102. // budgetable, so no budget value can pass. It is rejected on its own terms:
  1103. // the budget loop would otherwise report "a limit of -1" (or -2796203 at
  1104. // 32 MiB) while naming maxLogBytes, sending the operator to the wrong knob.
  1105. const ctxBaseline = new Context()
  1106. await expect(ctxBaseline.plugin(PythonCodeRuntime, { addressSpaceMb: 64 }))
  1107. .rejects.toThrow(/addressSpaceMb must exceed the 67108864-byte interpreter baseline/)
  1108. const ctxBelow = new Context()
  1109. await expect(ctxBelow.plugin(PythonCodeRuntime, { addressSpaceMb: 32 }))
  1110. .rejects.toThrow(/addressSpaceMb must exceed the 67108864-byte interpreter baseline/)
  1111. // The default caps against the default 512 MiB address space load.
  1112. const ok = new Context()
  1113. const fiber = await ok.plugin(PythonCodeRuntime, { maxLogBytes: 65536, maxValueBytes: 32768, addressSpaceMb: 512 })
  1114. await fiber.dispose()
  1115. })
  1116. it('bounds an illegal-UTF-8 native residual by its U+FFFD-decoded cost', async () => {
  1117. // Every 0xFF byte is illegal in any UTF-8 sequence, so `toString('utf8')`
  1118. // renders each as U+FFFD (3 serialized bytes). `accrueStrayCost` must charge
  1119. // that 3, not the raw 1: otherwise the newline-free residual grows to a full
  1120. // budget's worth of RAW bytes before flushing — a ~3x undercount that near a
  1121. // large maxLogBytes retains hundreds of MiB then expands toward a ~1 GiB peak
  1122. // in flushStray's concat + toString. Paced single-byte writes (each its own
  1123. // `data` chunk, like the sealing case) expose the sub-chunk accrual: charged
  1124. // at 3 the residual crosses a 3072-byte budget after ~1024 bytes and flushes;
  1125. // charged at 1 it would need ~3072 bytes, so the peak residual triples. The
  1126. // largest merged buffer is the discriminator.
  1127. const realConcat = Buffer.concat.bind(Buffer)
  1128. let maxConcat = 0
  1129. Buffer.concat = (list: readonly Uint8Array[], total?: number): Buffer<ArrayBuffer> => {
  1130. const merged = realConcat(list, total)
  1131. if (merged.length > maxConcat) maxConcat = merged.length
  1132. return merged
  1133. }
  1134. let result: CodeRunResult
  1135. try {
  1136. const { runtime } = await setup({ maxLogBytes: 3072, maxWallMs: 30_000 })
  1137. result = await runtime.run({
  1138. program: [
  1139. 'import os',
  1140. 'for _ in range(6000):',
  1141. ' os.write(1, b"\\xff")',
  1142. ' os.sched_yield()',
  1143. 'return None',
  1144. ].join('\n'),
  1145. bindings: [],
  1146. })
  1147. } finally {
  1148. Buffer.concat = realConcat
  1149. }
  1150. expect(result.error).toBeUndefined()
  1151. expect(result.logs.at(-1)).toBe(logTruncationMarker(3072))
  1152. // Charged at 3, the residual flushes around 1024 raw bytes; the largest
  1153. // merged buffer stays well under 2048. A raw-byte undercount would let it
  1154. // reach ~3072 before flushing, so 2048 discriminates.
  1155. expect(maxConcat).toBeLessThan(2048)
  1156. })
  1157. it('charges a structurally-valid but illegal UTF-8 sequence its U+FFFD-decoded cost', async () => {
  1158. // A CESU-8 lone surrogate `ED A0 80` is structurally well-formed (a 3-byte
  1159. // lead plus two 0x80–0xBF continuations) but ILLEGAL: `toString('utf8')`
  1160. // renders each of the three bytes as its own U+FFFD (serialized cost 9), not
  1161. // one width-3 character. The newline-free flush trigger weighs the residual
  1162. // through `accrueStrayCost`, which must validate each lead's
  1163. // first-continuation range (ED excludes A0–BF) and charge the true 9 — else a
  1164. // CESU flood undercounts 3x and the residual grows toward a full budget's raw
  1165. // bytes before flushing, the same peak-memory vector as the 0xFF case. The
  1166. // bytes are written one at a time (each its own `data` chunk, no pipe
  1167. // coalescing) and `Buffer.concat` is wrapped to measure the peak residual.
  1168. const realConcat = Buffer.concat.bind(Buffer)
  1169. let maxConcat = 0
  1170. Buffer.concat = (list: readonly Uint8Array[], total?: number): Buffer<ArrayBuffer> => {
  1171. const merged = realConcat(list, total)
  1172. if (merged.length > maxConcat) maxConcat = merged.length
  1173. return merged
  1174. }
  1175. let result: CodeRunResult
  1176. try {
  1177. const { runtime } = await setup({ maxLogBytes: 3072, maxWallMs: 30_000 })
  1178. result = await runtime.run({
  1179. program: [
  1180. 'import os',
  1181. 'seq = (0xed, 0xa0, 0x80)',
  1182. 'for _ in range(2000):',
  1183. ' for b in seq:',
  1184. ' os.write(1, bytes((b,)))',
  1185. ' os.sched_yield()',
  1186. 'return None',
  1187. ].join('\n'),
  1188. bindings: [],
  1189. })
  1190. } finally {
  1191. Buffer.concat = realConcat
  1192. }
  1193. expect(result.error).toBeUndefined()
  1194. expect(result.logs.at(-1)).toBe(logTruncationMarker(3072))
  1195. // Each 3-byte sequence costs 9 (three U+FFFD), so single-byte-paced the
  1196. // residual crosses the 3072 budget after ~342 raw bytes and flushes; the
  1197. // largest merged buffer stays well under 2048. Charging the structural width
  1198. // 3 would need ~1024 raw bytes, tripling the peak past 2048.
  1199. expect(maxConcat).toBeLessThan(2048)
  1200. })
  1201. it('charges a lone surrogate its full six escaped bytes, not three', async () => {
  1202. // A forged `log` frame carrying `\ud800` escapes materializes lone
  1203. // surrogates after JSON.parse. `Buffer.byteLength` of U+FFFD is 3, but
  1204. // ES2019 well-formed `JSON.stringify` emits `\ud800` at 6 bytes, so charging
  1205. // the raw width would admit ~2x the configured budget of serialized bytes
  1206. // (the same family as the NUL-flood undercount, at 2x rather than 6x). The
  1207. // cost walker charges surrogates the full 6, so a flood truncates at budget.
  1208. // Forged on fd 3 because Python stdout will not emit lone surrogates.
  1209. const { runtime } = await setup({ maxLogBytes: 4096 })
  1210. const result = await runtime.run({
  1211. program: [
  1212. 'import os',
  1213. // 1000 \ud800 escapes: charged at the buggy raw width 1000 * 3 = 3000
  1214. // bytes fits under 4096 (wrongly admitted), but the correct serialized
  1215. // width 1000 * 6 = 6000 bytes is over budget — so the ledger must
  1216. // truncate. The count sits in the 683..1365 window where the two
  1217. // chargings disagree, making the test discriminate.
  1218. String.raw`frame = b'{"type":"log","text":"' + b'\\ud800' * 1000 + b'"}\n'`,
  1219. 'os.write(3, frame)',
  1220. 'return None',
  1221. ].join('\n'),
  1222. bindings: [],
  1223. })
  1224. expect(result.error).toBeUndefined()
  1225. expect(result.logs.at(-1)).toBe(logTruncationMarker(4096))
  1226. })
  1227. it('drops a second stray line in the same chunk once the first truncated the ledger', async () => {
  1228. // One `os.write` carrying two newline-terminated lines where the first
  1229. // exhausts maxLogBytes: the first line's admit truncates and marks the
  1230. // ledger, and the second line's admit — reached in the same `data` callback
  1231. // — must be the post-truncation no-op. Proves that branch is exercised, so
  1232. // it carries no v8-ignore. Kept to 108 bytes (< the smallest PIPE_BUF, 512 on
  1233. // macOS) so the whole payload lands in ONE atomic write and one `data`
  1234. // callback — the two newlines cannot split across callbacks and leave the
  1235. // branch un-exercised, which would be a hard-to-attribute per-file coverage
  1236. // flake. The first line's 100 bytes already exceed the 64-byte budget, so it truncates.
  1237. const { runtime } = await setup({ maxLogBytes: 64 })
  1238. const result = await runtime.run({
  1239. program: ['import os', 'os.write(1, b"A" * 100 + b"\\nSECOND\\n")', 'return None'].join('\n'),
  1240. bindings: [],
  1241. })
  1242. expect(result.error).toBeUndefined()
  1243. expect(result.logs.at(-1)).toBe(logTruncationMarker(64))
  1244. expect(result.logs.join('\n')).not.toContain('SECOND')
  1245. })
  1246. it('charges a broken multibyte sequence its U+FFFD bytes, split across pipe chunks', async () => {
  1247. // A 3-byte lead (0xE4) whose continuation never arrives — the next byte is a
  1248. // fresh ASCII 'A' — must be costed as U+FFFD (3) for the orphaned lead, not
  1249. // folded into a phantom character. Driven byte-by-byte so the lead and the
  1250. // breaking byte land in separate `data` chunks, exercising accrueStrayCost's
  1251. // cross-chunk broken-sequence branch. The run completes and the bytes are
  1252. // captured (rendered U+FFFD by toString), proving the walk resynchronizes.
  1253. const { runtime } = await setup({ maxLogBytes: 1024 })
  1254. const result = await runtime.run({
  1255. program: [
  1256. 'import os',
  1257. 'os.write(1, b"\\xe4")',
  1258. 'os.sched_yield()',
  1259. 'os.write(1, b"A\\n")',
  1260. 'return None',
  1261. ].join('\n'),
  1262. bindings: [],
  1263. })
  1264. expect(result.error).toBeUndefined()
  1265. expect(result.logs.join('')).toContain('A')
  1266. expect(result.logs.join('')).toContain('�')
  1267. })
  1268. it('charges the exact serialized cost of short-escape and quote/backslash characters', async () => {
  1269. // Exercises every branch of jsonStringCostUpTo's per-character cost: a tab
  1270. // and other C0 controls with short JSON forms (\t etc., 2 bytes), a quote
  1271. // and backslash (2 bytes each), a `\uXXXX` control (6 bytes), a multibyte
  1272. // BMP character (raw UTF-8 width), and plain ASCII. Under a budget large
  1273. // enough to admit it, the line survives verbatim — proving the cost walker
  1274. // does not over- or under-charge and the string round-trips unescaped.
  1275. const { runtime } = await setup({ maxLogBytes: 4096 })
  1276. const result = await runtime.run({
  1277. program: ['import os', String.raw`os.write(1, "\ta\"b\\c\x01é\n".encode("utf-8"))`, 'return None'].join('\n'),
  1278. bindings: [],
  1279. })
  1280. expect(result.error).toBeUndefined()
  1281. expect(result.logs).toEqual(['\ta"b\\c\x01é'])
  1282. })
  1283. it('fails a completion dict with a non-string key as invalid-output (no key coercion)', async () => {
  1284. // json.dumps would coerce {1: "a", "1": "b"} to a single "1" key, silently
  1285. // dropping data. The shape validator rejects it before encoding.
  1286. const { runtime } = await setup()
  1287. const result = await runtime.run({
  1288. program: 'return {1: "first", "1": "second"}',
  1289. bindings: [],
  1290. })
  1291. expect(result.value).toBeUndefined()
  1292. expect(result.error?.kind).toBe('invalid-output')
  1293. expect(result.error?.message).toContain('non-string dict key')
  1294. })
  1295. it('rejects a binding argument with a non-string dict key before dispatch', async () => {
  1296. const { runtime } = await setup()
  1297. let called = false
  1298. const result = await runtime.run({
  1299. program: [
  1300. 'caught = ""',
  1301. 'try:',
  1302. ' await tools.sink({1: "x"})',
  1303. 'except RuntimeError as e:',
  1304. ' caught = str(e)',
  1305. 'return caught',
  1306. ].join('\n'),
  1307. bindings: tools({ sink: async () => { called = true; return null } }),
  1308. })
  1309. expect(result.error).toBeUndefined()
  1310. expect(result.value).toContain('lossless JSON')
  1311. expect(called).toBe(false)
  1312. })
  1313. it('fails a non-JSON completion value as invalid-output (no repr substitution)', async () => {
  1314. // A set is not lossless JSON. The old draft substituted repr(); the seam
  1315. // now requires refusing the run instead.
  1316. const { runtime } = await setup()
  1317. const result = await runtime.run({
  1318. program: 'return {1, 2, 3}',
  1319. bindings: [],
  1320. })
  1321. expect(result.value).toBeUndefined()
  1322. expect(result.error?.kind).toBe('invalid-output')
  1323. expect(result.error?.message).toContain('lossless JSON')
  1324. expect(result.error?.message).toContain('set')
  1325. })
  1326. it('fails a negative-zero completion value as invalid-output (sign bit is lossy over JSON)', async () => {
  1327. // JSON serialization turns -0.0 into 0 (or JS -0), silently changing the
  1328. // sign bit; the canonical lossless-JSON boundary rejects it, so the
  1329. // Python side must too — as a completion and as a binding argument.
  1330. const { runtime } = await setup()
  1331. const completion = await runtime.run({
  1332. program: 'return -0.0',
  1333. bindings: [],
  1334. })
  1335. expect(completion.error?.kind).toBe('invalid-output')
  1336. expect(completion.error?.message).toContain('negative zero')
  1337. const argument = await runtime.run({
  1338. program: [
  1339. 'try:',
  1340. ' await tools.echo(-0.0)',
  1341. ' return "accepted"',
  1342. 'except RuntimeError as e:',
  1343. ' return str(e)',
  1344. ].join('\n'),
  1345. bindings: tools({ echo: async args => args as never }),
  1346. })
  1347. expect(argument.error).toBeUndefined()
  1348. expect(argument.value).toContain('negative zero')
  1349. })
  1350. it('fails a NaN completion value as invalid-output (allow_nan=False)', async () => {
  1351. // json.dumps would happily emit NaN by default, but NaN is not JSON; the
  1352. // bootstrap passes allow_nan=False so it fails as invalid-output.
  1353. const { runtime } = await setup()
  1354. const result = await runtime.run({
  1355. program: 'return float("nan")',
  1356. bindings: [],
  1357. })
  1358. expect(result.error?.kind).toBe('invalid-output')
  1359. })
  1360. it('fails an over-budget completion value as output-limit (child-side check)', async () => {
  1361. const { runtime } = await setup({ maxValueBytes: 64 })
  1362. const result = await runtime.run({
  1363. program: 'return "V" * 5000',
  1364. bindings: [],
  1365. })
  1366. expect(result.value).toBeUndefined()
  1367. expect(result.error?.kind).toBe('output-limit')
  1368. expect(result.error?.message).toContain('exceeded 64 bytes')
  1369. })
  1370. it('meters a control-heavy completion value without materializing its escaped form', async () => {
  1371. // The child's lower bound admits a string by CHARACTER count, then the meter
  1372. // charged what `_dump_string(current).encode()` returned -- building the
  1373. // escaped copy plus its encode. Each NUL escapes to six bytes, so metering a
  1374. // value the budget then REJECTS allocated ~6x the original twice over:
  1375. // measured at 228.9 MiB of peak for a 20M-NUL string, against 19.1 MiB for
  1376. // the counting path that returns the identical 120,000,002 bytes. Past
  1377. // RLIMIT_AS the meter died as `exception: MemoryError`, inverting the
  1378. // `output-limit` this seam promises for an over-budget value.
  1379. //
  1380. // 8M NULs is 8,000,002 raw but 48,000,002 escaped: over the 16 MiB budget
  1381. // only when charged the escaped cost, so this also pins that the cheap
  1382. // character bound alone does not decide the verdict.
  1383. const { runtime } = await setup({ maxValueBytes: 16 * 1024 * 1024, maxWallMs: 60_000 })
  1384. const result = await runtime.run({
  1385. program: 'return "\\x00" * 8_000_000',
  1386. bindings: [],
  1387. })
  1388. expect(result.value).toBeUndefined()
  1389. expect(result.error?.kind).toBe('output-limit')
  1390. }, 90_000)
  1391. it('rejects a wide completion as output-limit before materializing its traversal state', async () => {
  1392. // `[0] * 2000000` sits far above maxValueBytes but well below the frame
  1393. // ceiling. The folded checker must reject it via the pre-enqueue bound —
  1394. // BEFORE pushing two million elements onto the walk — so a small
  1395. // addressSpaceMb does not turn the check itself into an RLIMIT_AS death.
  1396. const { runtime } = await setup({ maxValueBytes: 64, addressSpaceMb: 256, maxWallMs: 15_000 })
  1397. const result = await runtime.run({
  1398. program: 'return [0] * 2000000',
  1399. bindings: [],
  1400. })
  1401. expect(result.value).toBeUndefined()
  1402. expect(result.error?.kind).toBe('output-limit')
  1403. expect(result.error?.message).toContain('exceeded 64 bytes')
  1404. }, 20_000)
  1405. it('rejects a wide dict as output-limit without materializing its items list', async () => {
  1406. // Same pre-enqueue bound on the dict branch: `len(current)` replaces
  1407. // `list(current.items())`, which allocated one tuple per member before the
  1408. // bound could reject the value. Two million entries under a 64-byte cap
  1409. // fits the 256 MiB address space as a dict but not as a dict PLUS a
  1410. // two-million-tuple list.
  1411. const { runtime } = await setup({ maxValueBytes: 64, addressSpaceMb: 256, maxWallMs: 15_000 })
  1412. const result = await runtime.run({
  1413. program: 'return {str(i): 0 for i in range(2000000)}',
  1414. bindings: [],
  1415. })
  1416. expect(result.value).toBeUndefined()
  1417. expect(result.error?.kind).toBe('output-limit')
  1418. expect(result.error?.message).toContain('exceeded 64 bytes')
  1419. }, 20_000)
  1420. it('meters a float completion in the host\'s number spelling', async () => {
  1421. // CPython's repr disagrees with the host's String(number): `1.0` is three
  1422. // bytes here and one there, `1e-07` pads the exponent the host writes as
  1423. // `1e-7`. Both sides meter the SAME budget, so the child must count the
  1424. // bytes the host will receive — otherwise a boundary-sized value is
  1425. // falsely reported as output-limit.
  1426. const { runtime } = await setup({ maxValueBytes: 1 })
  1427. const integral = await runtime.run({ program: 'return 1.0', bindings: [] })
  1428. expect(integral.error).toBeUndefined()
  1429. expect(integral.value).toBe(1)
  1430. const exponent = await setup({ maxValueBytes: 4 })
  1431. const small = await exponent.runtime.run({ program: 'return 1e-7', bindings: [] })
  1432. expect(small.error).toBeUndefined()
  1433. expect(small.value).toBe(1e-7)
  1434. // The spelling is a meter input, not a licence to overshoot: `1.5` is three
  1435. // bytes on both sides and still fails a two-byte budget.
  1436. const tight = await setup({ maxValueBytes: 2 })
  1437. const over = await tight.runtime.run({ program: 'return 1.5', bindings: [] })
  1438. expect(over.error?.kind).toBe('output-limit')
  1439. })
  1440. it('carries floats across the wire in the host\'s number spelling', async () => {
  1441. // The child ENCODES with the same speller it meters with, so the frame the
  1442. // host parses must reproduce every double exactly — including the branches
  1443. // where CPython and ECMAScript disagree (integral floats, sub-1e-6
  1444. // exponents, >= 1e21, and beyond-safe-range integral doubles whose exact
  1445. // digits differ from the shortest round-trip form).
  1446. const { runtime } = await setup()
  1447. const result = await runtime.run({
  1448. 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]',
  1449. bindings: [],
  1450. })
  1451. expect(result.error).toBeUndefined()
  1452. 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])
  1453. })
  1454. it('rejects a forged non-lossless done value host-side as invalid-output', async () => {
  1455. // A forged done frame bypasses the child's _check_done_value. JSON.parse
  1456. // turns 1e400 into Infinity; validateChildFrame no longer scans done.value,
  1457. // so the host's own checkDoneValue must catch the non-lossless number.
  1458. const { runtime } = await setup()
  1459. const result = await runtime.run({
  1460. program: [
  1461. 'import os',
  1462. String.raw`os.write(3, b'{"type":"done","value":1e400}' + b'\n')`,
  1463. 'import time',
  1464. 'time.sleep(5)',
  1465. ].join('\n'),
  1466. bindings: [],
  1467. })
  1468. expect(result.value).toBeUndefined()
  1469. expect(result.error?.kind).toBe('invalid-output')
  1470. expect(result.error?.message).toContain('non-lossless number')
  1471. })
  1472. it('reports a syntax error as an exception without settling with a value', async () => {
  1473. const { runtime } = await setup()
  1474. const result = await runtime.run({
  1475. program: '$$invalid python$$',
  1476. bindings: [],
  1477. })
  1478. expect(result.error?.kind).toBe('exception')
  1479. expect(result.error?.message).toContain('SyntaxError')
  1480. // The parse-time diagnostic must carry the same source label as compile and
  1481. // runtime tracebacks (ast.parse passes filename="<model>"); a stale
  1482. // "<unknown>" label would leak an inconsistent origin to the model.
  1483. expect(result.error?.message).toContain('File \"<model>\"')
  1484. expect(result.value).toBeUndefined()
  1485. })
  1486. it('reports a runtime raise as an exception with the traceback', async () => {
  1487. const { runtime } = await setup()
  1488. const result = await runtime.run({
  1489. program: 'raise ValueError("intentional")',
  1490. bindings: [],
  1491. })
  1492. expect(result.error?.kind).toBe('exception')
  1493. expect(result.error?.message).toContain('ValueError')
  1494. expect(result.error?.message).toContain('intentional')
  1495. })
  1496. it('bounds a deep exception cause chain instead of burning the wall budget formatting it', async () => {
  1497. // A chain thousands of links deep would make the rendering walk and
  1498. // format() linear in its length, consuming maxWallMs. Rendering is capped
  1499. // at 100 links with a marker; the run reports the exception well within
  1500. // budget rather than timing out.
  1501. const { runtime } = await setup({ maxValueBytes: 1024 * 1024, maxWallMs: 20_000 })
  1502. const start = Date.now()
  1503. const result = await runtime.run({
  1504. program: [
  1505. 'err = None',
  1506. 'for i in range(3000):',
  1507. ' try:',
  1508. ' raise ValueError(i) from err',
  1509. ' except ValueError as e:',
  1510. ' err = e',
  1511. 'raise err',
  1512. ].join('\n'),
  1513. bindings: [],
  1514. })
  1515. expect(result.error?.kind).toBe('exception')
  1516. expect(result.error?.message).toContain('exception chain truncated at 100 links')
  1517. expect(Date.now() - start).toBeLessThan(15_000)
  1518. }, 25_000)
  1519. it('bounds an over-cap chain without assigning to the live exception', async () => {
  1520. // The cap used to be applied by severing the over-cap link ON the live
  1521. // exception. An exception class overriding __setattr__ to raise turned that
  1522. // assignment into model code running inside the bootstrap's failure
  1523. // handler; the throw skipped the `done` send that sits after the handler,
  1524. // so the host blocked on fd 3 and reported a maxWallMs timeout instead of
  1525. // the model's own exception. Cutting the chain on the TracebackException
  1526. // COPY touches no model hook, so the marker still appears and the run
  1527. // reports `exception`.
  1528. const { runtime } = await setup({ maxValueBytes: 1024 * 1024, maxWallMs: 15_000 })
  1529. const start = Date.now()
  1530. const result = await runtime.run({
  1531. program: [
  1532. 'class Sealed(Exception):',
  1533. ' def __setattr__(self, name, value):',
  1534. ' raise RuntimeError("live mutation refused")',
  1535. 'err = None',
  1536. 'for i in range(150):',
  1537. ' try:',
  1538. ' raise Sealed(i) from err',
  1539. ' except Sealed as e:',
  1540. ' err = e',
  1541. 'raise err',
  1542. ].join('\n'),
  1543. bindings: [],
  1544. })
  1545. expect(result.error?.kind).toBe('exception')
  1546. expect(result.error?.message).toContain('Sealed')
  1547. expect(result.error?.message).toContain('exception chain truncated at 100 links')
  1548. // The sever attempt is what used to leak: its message must not appear, and
  1549. // the run must settle well inside the wall budget rather than timing out.
  1550. expect(result.error?.message).not.toContain('live mutation refused')
  1551. expect(Date.now() - start).toBeLessThan(10_000)
  1552. }, 20_000)
  1553. it('still sends done when rendering the diagnostic itself raises', async () => {
  1554. // format() reaches the exception's own __str__, so a model class whose
  1555. // __str__ raises can throw from inside the failure handler. CPython's
  1556. // _safe_string absorbs a raising __str__ during formatting, but the
  1557. // fallback must hold for any throw on that path (a raising __repr__ of an
  1558. // argument, a MemoryError under RLIMIT_AS), so the assertion is the
  1559. // invariant that matters: a `done` frame carrying `exception`, never a
  1560. // timeout, and never the failing renderer's own message.
  1561. const { runtime } = await setup({ maxWallMs: 10_000 })
  1562. const result = await runtime.run({
  1563. program: [
  1564. 'class Unprintable(Exception):',
  1565. ' def __str__(self):',
  1566. ' raise RuntimeError("str refused")',
  1567. ' def __repr__(self):',
  1568. ' raise RuntimeError("repr refused")',
  1569. 'raise Unprintable()',
  1570. ].join('\n'),
  1571. bindings: [],
  1572. })
  1573. expect(result.error?.kind).toBe('exception')
  1574. expect(result.error?.message).toContain('Unprintable')
  1575. expect(result.error?.message).not.toContain('str refused')
  1576. expect(result.error?.message).not.toContain('repr refused')
  1577. }, 15_000)
  1578. it('sends done with an inert diagnostic when the whole rendering path raises', async () => {
  1579. // Drive the fallback itself. `TracebackException.format` reads the
  1580. // exception class's `__module__` to decide whether to qualify the name, and
  1581. // a metaclass property can raise there — a throw INSIDE the formatter,
  1582. // reached with no rebinding of anything the bootstrap owns. Without the
  1583. // wrapper it escapes the handler, the `done` send never runs, and the host
  1584. // times out at maxWallMs.
  1585. const { runtime } = await setup({ maxWallMs: 10_000 })
  1586. const result = await runtime.run({
  1587. program: [
  1588. 'class Meta(type):',
  1589. ' @property',
  1590. ' def __module__(cls):',
  1591. ' raise RuntimeError("renderer refused")',
  1592. 'class Hostile(ValueError, metaclass=Meta):',
  1593. ' pass',
  1594. 'raise Hostile("original failure")',
  1595. ].join('\n'),
  1596. bindings: [],
  1597. })
  1598. expect(result.error?.kind).toBe('exception')
  1599. // The inert fallback names the class and a fixed literal; it must not carry
  1600. // the renderer's message, and must not have become a timeout. `__name__` is
  1601. // still a plain str here, so the class name survives.
  1602. expect(result.error?.message).toBe('Hostile: <diagnostic rendering failed>')
  1603. }, 15_000)
  1604. it('falls back to a placeholder class name when __name__ itself raises', async () => {
  1605. // The fallback reads type(exc).__name__, which a metaclass property can
  1606. // hijack. It must neither run that override's failure into the handler nor
  1607. // format a non-str __name__ into the message. The hostile `__module__` is
  1608. // what drives execution into the fallback in the first place.
  1609. const { runtime } = await setup({ maxWallMs: 10_000 })
  1610. const result = await runtime.run({
  1611. program: [
  1612. 'class Meta(type):',
  1613. ' @property',
  1614. ' def __module__(cls):',
  1615. ' raise RuntimeError("renderer refused")',
  1616. ' @property',
  1617. ' def __name__(cls):',
  1618. ' raise RuntimeError("name refused")',
  1619. 'class Nameless(Exception, metaclass=Meta):',
  1620. ' pass',
  1621. 'raise Nameless()',
  1622. ].join('\n'),
  1623. bindings: [],
  1624. })
  1625. expect(result.error?.kind).toBe('exception')
  1626. expect(result.error?.message).toBe('<unknown>: <diagnostic rendering failed>')
  1627. }, 15_000)
  1628. it('reports the real exception when the program rebinds every name the failure path uses', async () => {
  1629. // The bootstrap IS __main__, so `import __main__; __main__._X = ...` reaches
  1630. // any module global a call-time lookup would read. The failure path is the
  1631. // worst place for that: the reporter, the byte cap, the traceback formatter,
  1632. // the settlement flush and the `done` send all run AFTER the `except` block,
  1633. // so a replacement that raises skips the send, leaves the host blocked on
  1634. // fd 3, and the run reports a maxWallMs timeout instead of the model's own
  1635. // exception. Rebind all of them at once; the run must still carry the real
  1636. // ValueError.
  1637. const { runtime } = await setup({ maxWallMs: 10_000 })
  1638. const result = await runtime.run({
  1639. program: [
  1640. 'import __main__',
  1641. 'def boom(*a, **k):',
  1642. ' raise RuntimeError("hijacked")',
  1643. '__main__._SAFE_MODEL_TRACEBACK = boom',
  1644. '__main__._cap_message = boom',
  1645. '__main__._model_traceback = boom',
  1646. '__main__._UNRENDERABLE_DIAGNOSTIC = boom',
  1647. '__main__._LogStream.flush_line = boom',
  1648. // `send_done` writes via LOCALLY-BOUND `_encode_json_plain` +
  1649. // `ProtocolChannel.write_encoded`; rebinding these at call time must not
  1650. // redirect the done frame (a late lookup would be `boom` -> worker-exit).
  1651. '__main__.ProtocolChannel.send_sync = boom',
  1652. '__main__.ProtocolChannel.write_encoded = boom',
  1653. '__main__._encode_json_plain = boom',
  1654. 'raise ValueError("real failure")',
  1655. ].join('\n'),
  1656. bindings: [],
  1657. })
  1658. expect(result.error?.kind).toBe('exception')
  1659. expect(result.error?.message).toContain('ValueError: real failure')
  1660. expect(result.error?.message).not.toContain('hijacked')
  1661. }, 15_000)
  1662. it('still delivers a done frame when a transitive encode name is rebound', async () => {
  1663. // `send_done` binds `_encode_json_plain` and `ProtocolChannel.write_encoded`
  1664. // into locals, but those callables' BODIES still resolve transitive module
  1665. // globals at call time: `_encode_json_plain` reaches `_dump_scalar`/`_dump_string`/
  1666. // `json.dumps`, and `write_encoded` reaches `os.write`. This bootstrap is
  1667. // `__main__`, so rebinding `__main__._dump_scalar` to a raising function makes
  1668. // the error-frame encode throw AFTER the `except` block. `send_done` catches
  1669. // that and writes a fixed literal done frame (kind `exception`) with the
  1670. // LOCALLY-BOUND `_os_write`/`_memoryview`/`_FALLBACK_DONE_FRAME` captured
  1671. // before the program runs, so the host still gets a verdict — the run must be an
  1672. // `exception`, never a `worker-exit`. The real message is lost (the literal
  1673. // carries a fixed `<unrenderable>` text), which is acceptable: the verdict
  1674. // outranks the diagnostic detail.
  1675. const { runtime } = await setup({ maxWallMs: 10_000 })
  1676. const result = await runtime.run({
  1677. program: [
  1678. 'import __main__',
  1679. 'def boom(*a, **k):',
  1680. ' raise RuntimeError("hijacked")',
  1681. '__main__._dump_scalar = boom',
  1682. '__main__.os = boom',
  1683. // The fallback must also survive a rebind of its own primitives.
  1684. '__main__._os_write = boom',
  1685. '__main__._memoryview = boom',
  1686. '__main__._FALLBACK_DONE_FRAME = boom',
  1687. 'raise ValueError("real failure")',
  1688. ].join('\n'),
  1689. bindings: [],
  1690. })
  1691. expect(result.error?.kind).toBe('exception')
  1692. expect(result.error?.kind).not.toBe('worker-exit')
  1693. }, 15_000)
  1694. it('still reports a model exception when the program rebinds BaseException', async () => {
  1695. // `_run`'s outer try/except catches the program's failure and builds a
  1696. // `done` frame. The clause previously used the module-global `BaseException`,
  1697. // which the program (running as `__main__`) can rebind: `__main__.BaseException
  1698. // = RuntimeError` makes the `except BaseException` resolve to `RuntimeError`,
  1699. // so a subsequent `ValueError` does not match and escapes `_run` with no
  1700. // `done` frame — misreporting the run as a `worker-exit`. The exception class
  1701. // is now bound into a `_run` LOCAL before the program runs, so the rebind
  1702. // cannot change which class the clause catches; the run must still report an
  1703. // `exception`, not a `worker-exit`.
  1704. const { runtime } = await setup({ maxWallMs: 10_000 })
  1705. const result = await runtime.run({
  1706. program: [
  1707. 'import __main__',
  1708. '__main__.BaseException = RuntimeError',
  1709. 'raise ValueError("real failure")',
  1710. ].join('\n'),
  1711. bindings: [],
  1712. })
  1713. expect(result.error?.kind).toBe('exception')
  1714. expect(result.error?.kind).not.toBe('worker-exit')
  1715. }, 15_000)
  1716. it('still reports the exception when BaseException and the traceback reporter are rebound together', async () => {
  1717. // The two rebind families compose: `__main__.BaseException = ValueError`
  1718. // must not change which class the `_run` catch resolves (it is a pre-program
  1719. // local), and a rebound reporter (`_SAFE_MODEL_TRACEBACK`/`_cap_message`/
  1720. // `_model_traceback`/`_UNRENDERABLE_DIAGNOSTIC`) must not break the done
  1721. // frame — `safe_model_traceback` holds its primitives as import-time closure
  1722. // cells. A `KeyError` (not a `ValueError` subclass) escapes a catch that
  1723. // resolves to the rebound class, so without the local binding the run would
  1724. // misreport as `worker-exit`; with it, the run reports the exception and the
  1725. // fallback reporter still produces the fixed literal.
  1726. const { runtime } = await setup({ maxWallMs: 10_000 })
  1727. const result = await runtime.run({
  1728. program: [
  1729. 'import __main__',
  1730. 'def boom(*a, **k):',
  1731. ' raise RuntimeError("hijacked")',
  1732. '__main__.BaseException = ValueError',
  1733. '__main__._SAFE_MODEL_TRACEBACK = boom',
  1734. '__main__._cap_message = boom',
  1735. '__main__._model_traceback = boom',
  1736. '__main__._UNRENDERABLE_DIAGNOSTIC = boom',
  1737. 'raise KeyError("real failure")',
  1738. ].join('\n'),
  1739. bindings: [],
  1740. })
  1741. expect(result.error?.kind).toBe('exception')
  1742. expect(result.error?.kind).not.toBe('worker-exit')
  1743. }, 15_000)
  1744. it('rejects an fd-3 frame whose raw length exceeds the parse cap before joining it', async () => {
  1745. // The 64 MiB frame parse cap bounds the RAW frame bytes, not the decoded
  1746. // structure; a compact wide frame near that ceiling could decode to far
  1747. // more host memory. The unframed-buffer counter is checked against
  1748. // FRAME_PARSE_CAP_BYTES BEFORE the Buffer.concat join, so an oversized
  1749. // frame is dropped at one copy of its wire bytes instead of being fully
  1750. // joined (a second copy) and only then discarded in the line loop — the
  1751. // peak-memory doubling the pre-join check exists to prevent. Fail-before:
  1752. // without the check the frame is joined whole and parsed (its log text
  1753. // admitted, truncating the ledger), and the run completes normally.
  1754. const { runtime } = await setup({ maxWallMs: 60_000 })
  1755. const result = await runtime.run({
  1756. program: [
  1757. 'import os',
  1758. // One frame just past the 64 MiB parse cap.
  1759. 'os.write(3, b"{\\"type\\":\\"log\\",\\"text\\":\\"" + b"a" * (65 * 1024 * 1024) + b"\\"}\\n")',
  1760. 'return "done"',
  1761. ].join('\n'),
  1762. bindings: [],
  1763. })
  1764. expect(result.error?.kind).toBe('worker-exit')
  1765. expect(result.error?.message).toContain('protocol frame exceeded')
  1766. }, 90_000)
  1767. it('caps an oversized rejection diagnostic so an invalid completion stays invalid-output', async () => {
  1768. // _done_with_value caps its rejection diagnostic through _cap_message: a
  1769. // hostile class name (huge type(value).__name__) would otherwise push the
  1770. // done frame past the host's 64 MiB parse cap, misreporting an
  1771. // invalid-output run as a worker-exit. The diagnostic is capped to the
  1772. // value budget, so the frame always crosses the parser.
  1773. const { runtime } = await setup({ maxWallMs: 60_000 })
  1774. const result = await runtime.run({
  1775. program: [
  1776. 'return type("N" * (70 * 1024 * 1024), (), {})()',
  1777. ].join('\n'),
  1778. bindings: [],
  1779. })
  1780. expect(result.error?.kind).toBe('invalid-output')
  1781. expect(result.error?.kind).not.toBe('worker-exit')
  1782. }, 90_000)
  1783. it('appends a flushed unterminated line to the next entry without a fake newline', async () => {
  1784. // An explicit flush of an unterminated line (print(..., end='', flush=True))
  1785. // used to push a full log frame, so the following print() landed in a
  1786. // SECOND entry and logs.join('\n') rendered 'a\nb' for what the program
  1787. // printed as one line. The flush frame now carries `open: true` and the
  1788. // host appends the next frame to the same entry.
  1789. const { runtime } = await setup()
  1790. const result = await runtime.run({
  1791. program: [
  1792. "print('a', end='', flush=True)",
  1793. "print('b')",
  1794. 'return "done"',
  1795. ].join('\n'),
  1796. bindings: [],
  1797. })
  1798. expect(result.error).toBeUndefined()
  1799. expect(result.logs).toEqual(['ab'])
  1800. }, 15_000)
  1801. it('keeps a flushed unterminated line when the run ends with it still open', async () => {
  1802. // The settlement flush pushes the residual with `open: true`; finish()
  1803. // admits it so a program that commits a partial line and returns does not
  1804. // lose it from logs.
  1805. const { runtime } = await setup()
  1806. const result = await runtime.run({
  1807. program: [
  1808. "print('committed', end='', flush=True)",
  1809. 'return "done"',
  1810. ].join('\n'),
  1811. bindings: [],
  1812. })
  1813. expect(result.error).toBeUndefined()
  1814. expect(result.logs).toEqual(['committed'])
  1815. }, 15_000)
  1816. it('skips the hold for a zero-content open continuation', async () => {
  1817. // An empty open continuation bills 0 and is NOT pushed into the held
  1818. // fragment array (an empty fragment contributes nothing to the merged
  1819. // entry, and holding it would let a forged empty-open flood grow host
  1820. // memory without touching the ledger).
  1821. const { runtime } = await setup({ maxLogBytes: 64 })
  1822. const result = await runtime.run({
  1823. program: [
  1824. 'import os',
  1825. "os.write(3, b'{\"type\":\"log\",\"text\":\"x\",\"open\":true}\\n')",
  1826. "os.write(3, b'{\"type\":\"log\",\"text\":\"\",\"open\":true}\\n')",
  1827. "os.write(3, b'{\"type\":\"log\",\"text\":\"y\"}\\n')",
  1828. 'return "done"',
  1829. ].join('\n'),
  1830. bindings: [],
  1831. })
  1832. expect(result.error).toBeUndefined()
  1833. expect(result.logs).toEqual(['xy'])
  1834. }, 15_000)
  1835. it('bounds a forged open-frame flood against the log budget', async () => {
  1836. // The open hold must be bounded by the ledger: without the exact-cost check
  1837. // a forged open flood would grow the held fragment without touching
  1838. // logBudget — unbounded host retention under a small budget. The flood now
  1839. // truncates to the marker like any over-budget log traffic.
  1840. const { runtime } = await setup({ maxLogBytes: 64 })
  1841. const result = await runtime.run({
  1842. program: [
  1843. 'import os',
  1844. // 2000 forged open frames, each under the frame parse cap.
  1845. 'for _ in range(2000):',
  1846. " os.write(3, b'{\"type\":\"log\",\"text\":\"a\",\"open\":true}\\n')",
  1847. 'return "done"',
  1848. ].join('\n'),
  1849. bindings: [],
  1850. })
  1851. expect(result.error).toBeUndefined()
  1852. expect(result.logs).toEqual(['a'.repeat(60), logTruncationMarker(64)])
  1853. }, 15_000)
  1854. it('commits a flushed open prefix before the truncation marker', async () => {
  1855. // A flushed unterminated line is billed and committed; when a later
  1856. // over-budget write truncates, the committed prefix must appear BEFORE the
  1857. // marker — the ledger charged for it, so it cannot vanish.
  1858. const { runtime } = await setup({ maxLogBytes: 64 })
  1859. const result = await runtime.run({
  1860. program: [
  1861. "print('committed', end='', flush=True)",
  1862. "print('x' * 100)",
  1863. 'return "done"',
  1864. ].join('\n'),
  1865. bindings: [],
  1866. })
  1867. expect(result.error).toBeUndefined()
  1868. expect(result.logs).toEqual(['committed', logTruncationMarker(64)])
  1869. }, 15_000)
  1870. it('drops a forged fd-3 frame with illegal UTF-8 instead of accepting a mangled value', async () => {
  1871. // toString('utf8') would replace the illegal 0xFF with U+FFFD, so a forged
  1872. // done frame could land a corrupted completion value; the fatal decode
  1873. // throws and the frame is dropped. The program's real return still settles
  1874. // the run with the honest value.
  1875. const { runtime } = await setup()
  1876. const result = await runtime.run({
  1877. program: [
  1878. 'import os',
  1879. "os.write(3, b'{\"type\":\"done\",\"value\":\"bad' + bytes([0xFF]) + b'\"}\\n')",
  1880. 'return "ok"',
  1881. ].join('\n'),
  1882. bindings: [],
  1883. })
  1884. expect(result.error).toBeUndefined()
  1885. expect(result.value).toBe('ok')
  1886. }, 15_000)
  1887. it('no-ops a closing frame once an open flood already truncated the ledger', async () => {
  1888. // The closing-frame branch's post-truncation arm: an open flood exhausts
  1889. // the ledger (logsTruncated set, marker pushed), then a closing frame
  1890. // arrives — it must be a no-op, not append content past the marker.
  1891. const { runtime } = await setup({ maxLogBytes: 64 })
  1892. const result = await runtime.run({
  1893. program: [
  1894. 'import os',
  1895. 'for _ in range(2000):',
  1896. " os.write(3, b'{\"type\":\"log\",\"text\":\"a\",\"open\":true}\\n')",
  1897. "os.write(3, b'{\"type\":\"log\",\"text\":\"b\"}\\n')",
  1898. 'return "done"',
  1899. ].join('\n'),
  1900. bindings: [],
  1901. })
  1902. expect(result.error).toBeUndefined()
  1903. expect(result.logs).toEqual(['a'.repeat(60), logTruncationMarker(64)])
  1904. }, 15_000)
  1905. it('bills a merged open entry once, not per fragment', async () => {
  1906. // A merged entry's wire cost is billed ONCE, split across its fragments
  1907. // (first fragment pays quotes+separator, continuations pay only content).
  1908. // Under maxLogBytes: 64, 16 single-character flushes merge to one 16-char
  1909. // entry (2 quotes + 16 content + 1 separator = 19), which fits; per-
  1910. // fragment billing (each charged quotes+separator, ~4 bytes) would truncate
  1911. // at 16 x 4 = 64.
  1912. const { runtime } = await setup({ maxLogBytes: 64 })
  1913. const result = await runtime.run({
  1914. program: [
  1915. 'for _ in range(16):',
  1916. " print('x', end='', flush=True)",
  1917. "print('')",
  1918. 'return "done"',
  1919. ].join('\n'),
  1920. bindings: [],
  1921. })
  1922. expect(result.error).toBeUndefined()
  1923. expect(result.logs).toEqual(['x'.repeat(16)])
  1924. }, 15_000)
  1925. it('admits a compliant merged entry whose closing frame fits the remaining budget', async () => {
  1926. // The review's arithmetic check: print('a'*30, flush); print('b'*25) under
  1927. // maxLogBytes: 64 has a merged wire cost of 2 quotes + 55 content + 1
  1928. // separator = 58 <= 63, so it MUST be admitted as one entry. The earlier
  1929. // cap math (logBudget - openCost) made the closing frame's walk see a
  1930. // negative cap and truncate a compliant entry.
  1931. const { runtime } = await setup({ maxLogBytes: 64 })
  1932. const result = await runtime.run({
  1933. program: [
  1934. "print('a' * 30, end='', flush=True)",
  1935. "print('b' * 25)",
  1936. 'return "done"',
  1937. ].join('\n'),
  1938. bindings: [],
  1939. })
  1940. expect(result.error).toBeUndefined()
  1941. expect(result.logs).toEqual(['a'.repeat(30) + 'b'.repeat(25)])
  1942. }, 15_000)
  1943. it('rejects an open frame that would overflow the ledger by one byte', async () => {
  1944. // The review's arithmetic check: an open frame whose full JSON cost is 63
  1945. // (maxLogBytes: 64 -> ledger 63) must be rejected by the first-fragment
  1946. // cap logBudget - 1 (62), not admitted with a bill of 64 that pushes the
  1947. // ledger negative. The frame is FORGED on fd 3 so the child ledger cannot
  1948. // truncate first: a reverted cap of logBudget (63) would admit the frame,
  1949. // hold it, and flush it at settlement, so the marker assertion fails.
  1950. const { runtime } = await setup({ maxLogBytes: 64 })
  1951. const result = await runtime.run({
  1952. program: [
  1953. 'import os',
  1954. "os.write(3, ('{\"type\":\"log\",\"text\":\"' + 'x' * 61 + '\",\"open\":true}\\n').encode())",
  1955. 'return "done"',
  1956. ].join('\n'),
  1957. bindings: [],
  1958. })
  1959. expect(result.error).toBeUndefined()
  1960. expect(result.logs).toEqual([logTruncationMarker(64)])
  1961. }, 15_000)
  1962. it('bills the closing frame as the merged tail under an exact-fit budget', async () => {
  1963. // The child's split billing: a 30-char open + a 30-char closing frame cost
  1964. // 2 + 60 + 1 = 63 = ledger 63 exactly; the closing frame must be billed as
  1965. // the merged tail (content only), not as a fresh entry (which would
  1966. // double-charge the quotes+separator and truncate an exact-fit entry).
  1967. const { runtime } = await setup({ maxLogBytes: 64 })
  1968. const result = await runtime.run({
  1969. program: [
  1970. "print('a' * 30, end='', flush=True)",
  1971. "print('b' * 30)",
  1972. 'return "done"',
  1973. ].join('\n'),
  1974. bindings: [],
  1975. })
  1976. expect(result.error).toBeUndefined()
  1977. expect(result.logs).toEqual(['a'.repeat(30) + 'b'.repeat(30)])
  1978. }, 15_000)
  1979. it('does not over-reject an exact-fit closing line while an open entry accumulates', async () => {
  1980. // The write-path pre-check's cheap bound used +3 (quotes + separator) even
  1981. // while an open entry was accumulating, so an exact-fit merged TAIL was
  1982. // truncated. The recipe below goes through the SCAN pre-check (the
  1983. // newline-terminated write arrives with an empty pending buffer, so the
  1984. // buffered-chunks branch is skipped): 'a'*29 flush bills 32 (ledger 31
  1985. // left), then one write of 'b'*30 + newline merges 30 more chars whose
  1986. // cheap bound is 30, not 33 — the +3 form saw 30 + 3 = 33 > 31, sliced to
  1987. // a budget prefix, and pushed past the ledger, emitting the marker for a
  1988. // line that fits (merged cost 2 + 59 + 1 = 62 <= 63).
  1989. const { runtime } = await setup({ maxLogBytes: 64 })
  1990. const result = await runtime.run({
  1991. program: [
  1992. 'import sys',
  1993. "sys.stdout.write('a' * 29)",
  1994. 'sys.stdout.flush()',
  1995. "sys.stdout.write('b' * 30 + chr(10))",
  1996. 'return "done"',
  1997. ].join('\n'),
  1998. bindings: [],
  1999. })
  2000. expect(result.error).toBeUndefined()
  2001. expect(result.logs).toEqual(['a'.repeat(29) + 'b'.repeat(30)])
  2002. }, 15_000)
  2003. it('rejects a new open entry once the ledger has only two bytes left', async () => {
  2004. // The jsonStringCostUpTo sub-2-byte guard: forged open frames drive the
  2005. // host ledger down to 1 byte, then a new open entry's first-fragment cap
  2006. // (logBudget - 1 = 0) trips the guard and truncates.
  2007. const { runtime } = await setup({ maxLogBytes: 64 })
  2008. const result = await runtime.run({
  2009. program: [
  2010. 'import os',
  2011. "os.write(3, ('{\"type\":\"log\",\"text\":\"' + 'a' * 28 + '\",\"open\":true}\\n').encode())",
  2012. "os.write(3, ('{\"type\":\"log\",\"text\":\"' + 'a' * 31 + '\"}\\n').encode())",
  2013. "os.write(3, b'{\"type\":\"log\",\"text\":\"x\",\"open\":true}\\n')",
  2014. 'return "done"',
  2015. ].join('\n'),
  2016. bindings: [],
  2017. })
  2018. expect(result.error).toBeUndefined()
  2019. expect(result.logs).toEqual(['a'.repeat(59), logTruncationMarker(64)])
  2020. }, 15_000)
  2021. it('truncates when the closing frame of a merged entry overflows the budget', async () => {
  2022. // The merged entry's billed-once cost: an open fragment that nearly
  2023. // exhausts the budget, then a closing frame whose content no longer fits —
  2024. // the closing frame's exact-cost walk trips and the marker replaces the
  2025. // entry, exactly like any other over-budget log traffic.
  2026. const { runtime } = await setup({ maxLogBytes: 64 })
  2027. const result = await runtime.run({
  2028. program: [
  2029. "print('x' * 40, end='', flush=True)",
  2030. "print('y' * 40)",
  2031. 'return "done"',
  2032. ].join('\n'),
  2033. bindings: [],
  2034. })
  2035. expect(result.error).toBeUndefined()
  2036. expect(result.logs).toEqual(['x'.repeat(40), logTruncationMarker(64)])
  2037. }, 15_000)
  2038. it('keeps a float completion exact when the program mutates the decimal context', async () => {
  2039. // The float encoder's Decimal(repr(value)).normalize() used the process
  2040. // GLOBAL decimal context: a legitimate program setting
  2041. // `getcontext().prec = 2` silently rounded the completion value's digits,
  2042. // and `traps[Inexact] = True` made the encode raise, misclassifying a
  2043. // successful run as an exception. A fixed module-level Context(prec=28)
  2044. // makes the spelling decision context-independent.
  2045. const { runtime } = await setup()
  2046. const result = await runtime.run({
  2047. program: [
  2048. 'from decimal import getcontext',
  2049. 'getcontext().prec = 2',
  2050. 'getcontext().traps[__import__("decimal").Inexact] = True',
  2051. 'return 1.2345678901234567',
  2052. ].join('\n'),
  2053. bindings: [],
  2054. })
  2055. expect(result.error).toBeUndefined()
  2056. expect(result.value).toBe(1.2345678901234567)
  2057. }, 15_000)
  2058. it('bounds an over-cap exception-group nesting on the copy', async () => {
  2059. // Exception groups link through `exceptions`, not the cause/context
  2060. // dunders, so the cap has to count that edge too — otherwise a deeply
  2061. // nested group walks past the bound the marker claims to enforce.
  2062. const { runtime } = await setup({ maxValueBytes: 1024 * 1024, maxWallMs: 15_000 })
  2063. const result = await runtime.run({
  2064. program: [
  2065. 'import sys',
  2066. // ExceptionGroup is a 3.11+ builtin; on 3.10 the NameError is the
  2067. // failure mode being probed, so skip to keep the assertion meaningful.
  2068. 'if sys.version_info < (3, 11):',
  2069. ' raise ValueError("skip-old <model>")',
  2070. 'group = ValueError("leaf")',
  2071. 'for i in range(150):',
  2072. ' group = ExceptionGroup(f"g{i}", [group])',
  2073. 'raise group',
  2074. ].join('\n'),
  2075. bindings: [],
  2076. })
  2077. expect(result.error?.kind).toBe('exception')
  2078. // The version guard skips on Python < 3.11 (ExceptionGroup is a 3.11+
  2079. // builtin) with a distinct message; the truncation assertion applies on
  2080. // 3.11+ where the group nesting is what is being probed.
  2081. expect(result.error?.message).toMatch(/exception chain truncated at 100 links|skip-old/)
  2082. }, 20_000)
  2083. it('filters every bootstrap frame from the traceback of an uncaught binding rejection', async () => {
  2084. // A rejection re-raised by the bootstrap's dispatch adds bootstrap frames
  2085. // AFTER the model's own; only <model> frames may reach model-visible,
  2086. // durable output — a bootstrap.py path would leak host absolutes and make
  2087. // transcripts machine-dependent.
  2088. const { runtime } = await setup()
  2089. const result = await runtime.run({
  2090. program: 'await tools.boom({})',
  2091. bindings: tools({ boom: async () => { throw new Error('exploded') } }),
  2092. })
  2093. expect(result.error?.kind).toBe('exception')
  2094. expect(result.error?.message).toContain('exploded')
  2095. expect(result.error?.message).toContain('<model>')
  2096. expect(result.error?.message).not.toContain('bootstrap.py')
  2097. })
  2098. it('renders a non-Error thrown value from a host binding as its String form', async () => {
  2099. const { runtime } = await setup()
  2100. const result = await runtime.run({
  2101. program: [
  2102. 'caught = ""',
  2103. 'try:',
  2104. ' await tools.failRaw({})',
  2105. 'except RuntimeError as e:',
  2106. ' caught = str(e)',
  2107. 'return caught',
  2108. ].join('\n'),
  2109. bindings: tools({
  2110. failRaw: async () => { throw 'raw-nope' },
  2111. }),
  2112. })
  2113. expect(result.error).toBeUndefined()
  2114. expect(result.value).toContain('raw-nope')
  2115. })
  2116. it('reassembles a frame split across writes behind a completed one', async () => {
  2117. // One os.write carrying "<frame>\n<partial...>" leaves a non-empty
  2118. // residual after the newline loop; the tail must survive until its own
  2119. // newline arrives and then parse as a normal frame.
  2120. const { runtime } = await setup()
  2121. const result = await runtime.run({
  2122. program: [
  2123. 'import os, json',
  2124. 'head = json.dumps({"type":"log","text":"first"}).encode()',
  2125. 'tail = json.dumps({"type":"log","text":"second"}).encode()',
  2126. 'import time',
  2127. 'os.write(3, head + b"\\n" + tail[:5])',
  2128. 'time.sleep(0.2)',
  2129. 'os.write(3, tail[5:] + b"\\n")',
  2130. 'return "ok"',
  2131. ].join('\n'),
  2132. bindings: [],
  2133. })
  2134. expect(result.error).toBeUndefined()
  2135. expect(result.value).toBe('ok')
  2136. expect(result.logs).toContain('first')
  2137. expect(result.logs).toContain('second')
  2138. })
  2139. it('raises the declared errorClass with the member name on rejection', async () => {
  2140. // Code Mode declares { name: ToolCallError, memberNameProperty: toolName };
  2141. // a host rejection must surface as that class, carrying the failed tool.
  2142. const { runtime } = await setup()
  2143. const result = await runtime.run({
  2144. program: [
  2145. 'caught = ""',
  2146. 'try:',
  2147. ' await tools.fail({})',
  2148. 'except ToolCallError as e:',
  2149. ' caught = f"{type(e).__name__}:{e.toolName}:{e}"',
  2150. 'return caught',
  2151. ].join('\n'),
  2152. bindings: [{
  2153. global: 'tools',
  2154. functions: { fail: async () => { throw new Error('typed-nope') } },
  2155. errorClass: { name: 'ToolCallError', memberNameProperty: 'toolName' },
  2156. }],
  2157. })
  2158. expect(result.error).toBeUndefined()
  2159. expect(result.value).toBe('ToolCallError:fail:typed-nope')
  2160. })
  2161. it('keeps the declared error class catching when Exception and setattr are rebound', async () => {
  2162. // _make_error_class's minted __init__ def-time captures Exception and
  2163. // setattr, so a program rebinding __main__.Exception/__main__.setattr
  2164. // cannot break the rejection constructor: `except ToolCallError` must still
  2165. // catch and read the member property.
  2166. const { runtime } = await setup()
  2167. const result = await runtime.run({
  2168. program: [
  2169. 'import __main__',
  2170. 'def boom(*a, **k):',
  2171. ' raise RuntimeError("hijacked")',
  2172. '__main__.Exception = boom',
  2173. '__main__.setattr = boom',
  2174. 'caught = ""',
  2175. 'try:',
  2176. ' await tools.fail({})',
  2177. 'except ToolCallError as e:',
  2178. ' caught = f"{type(e).__name__}:{e.toolName}"',
  2179. 'return caught',
  2180. ].join('\n'),
  2181. bindings: [{
  2182. global: 'tools',
  2183. functions: { fail: async () => { throw new Error('typed-nope') } },
  2184. errorClass: { name: 'ToolCallError', memberNameProperty: 'toolName' },
  2185. }],
  2186. })
  2187. expect(result.error).toBeUndefined()
  2188. expect(result.value).toBe('ToolCallError:fail')
  2189. }, 15_000)
  2190. it('rejects an errorClass name colliding with its namespace global at the seam', async () => {
  2191. const { runtime } = await setup()
  2192. await expect(runtime.run({
  2193. program: 'return 1',
  2194. bindings: [{ global: 'tools', functions: {}, errorClass: { name: 'tools', memberNameProperty: 'toolName' } }],
  2195. })).rejects.toThrow(/collides with another injected global/)
  2196. })
  2197. it('rejects a namespace global colliding with a runtime-owned name at the seam', async () => {
  2198. // `__dsh_main__` passes the identifier check, but exec()ing the generated
  2199. // wrapper would silently overwrite the binding after injection. `console`
  2200. // is the WORKER backend's slot — refused here too so a namespace list
  2201. // valid on one backend is valid on all.
  2202. const { runtime } = await setup()
  2203. // `__debug__` is refused for a different reason than a collision: CPython
  2204. // compiles a bare `__debug__` reference to the constant True and refuses to
  2205. // assign the name at compile time, so an injected global under it is
  2206. // unreachable from the program — accepted by the seam, unusable here.
  2207. for (const global of ['__dsh_main__', 'console', '__debug__']) {
  2208. await expect(runtime.run({
  2209. program: 'x = 1',
  2210. bindings: [{ global, functions: {} }],
  2211. })).rejects.toThrow(/collides with a runtime-owned global/)
  2212. }
  2213. })
  2214. it('accepts a non-identifier memberNameProperty and rejects only an empty one', async () => {
  2215. // The seam permits any non-empty own property except the reserved
  2216. // members; Python setattr/getattr carry exotic names like `tool-name`,
  2217. // and the worker backend accepts them, so this backend must too.
  2218. const { runtime } = await setup()
  2219. const result = await runtime.run({
  2220. program: [
  2221. 'try:',
  2222. ' await tools.boom({})',
  2223. 'except ToolCallError as e:',
  2224. ' return getattr(e, "tool-name")',
  2225. ].join('\n'),
  2226. bindings: [{
  2227. global: 'tools',
  2228. functions: { boom: async () => { throw new Error('nope') } },
  2229. errorClass: { name: 'ToolCallError', memberNameProperty: 'tool-name' },
  2230. }],
  2231. })
  2232. expect(result.error).toBeUndefined()
  2233. expect(result.value).toBe('boom')
  2234. await expect(runtime.run({
  2235. program: 'return 1',
  2236. bindings: [{ global: 'tools', functions: {}, errorClass: { name: 'ToolCallError', memberNameProperty: '' } }],
  2237. })).rejects.toThrow(/memberNameProperty must be a non-empty attribute name/)
  2238. })
  2239. it('resolves a basename pythonBin to an absolute path (runs a real program)', async () => {
  2240. // A bare `python3` basename must resolve against PATH and actually launch
  2241. // under the empty-env spawn — exercises the accessSync success branch.
  2242. const { runtime } = await setup({ pythonBin: 'python3' })
  2243. const result = await runtime.run({ program: 'return 7', bindings: [] })
  2244. expect(result.error).toBeUndefined()
  2245. expect(result.value).toBe(7)
  2246. })
  2247. it('rejects at load a basename pythonBin with no PATH match', async () => {
  2248. // resolvePythonBin turns a basename into an absolute path before the
  2249. // empty-env spawn; a basename with no PATH match must fail at load (like an
  2250. // empty or NUL pythonBin) rather than silently falling to execvp's
  2251. // platform default PATH and starting a system interpreter the caller never
  2252. // asked for.
  2253. const ctx = new Context()
  2254. await expect(ctx.plugin(PythonCodeRuntime, { pythonBin: 'definitely-no-such-python-xyz' }))
  2255. .rejects.toThrow(/does not resolve on PATH/)
  2256. })
  2257. it('rejects a memberNameProperty naming a constrained BaseException attribute', async () => {
  2258. // `__dict__`/`__class__` are constrained descriptors alongside
  2259. // `__traceback__` — setattr of a string raises TypeError while
  2260. // constructing the rejection — so every dunder is refused at the seam.
  2261. const { runtime } = await setup()
  2262. // name/message/stack are the seam's own exclusions (CodeBindingErrorClass
  2263. // forbids replacing them; the worker backend rejects them identically).
  2264. for (const member of ['__traceback__', '__dict__', '__class__', 'args', 'name', 'message', 'stack']) {
  2265. await expect(runtime.run({
  2266. program: 'return 1',
  2267. bindings: [{ global: 'tools', functions: {}, errorClass: { name: 'ToolCallError', memberNameProperty: member } }],
  2268. })).rejects.toThrow(/reserved error member/)
  2269. }
  2270. })
  2271. it('rejects a lossy binding resolution (NaN) instead of coercing it to null', async () => {
  2272. // JSON.stringify would turn NaN into null and drop undefined fields; the
  2273. // seam requires a descriptive rejection so data cannot silently corrupt.
  2274. const { runtime } = await setup()
  2275. const result = await runtime.run({
  2276. program: [
  2277. 'caught = ""',
  2278. 'try:',
  2279. ' await tools.bad({})',
  2280. 'except RuntimeError as e:',
  2281. ' caught = str(e)',
  2282. 'return caught',
  2283. ].join('\n'),
  2284. bindings: tools({ bad: async () => Number.NaN }),
  2285. })
  2286. expect(result.error).toBeUndefined()
  2287. expect(result.value).toContain('lossless JSON')
  2288. })
  2289. it('contains a forged pathological done value without crashing the host', async () => {
  2290. // A ~20k-deep nested array forged onto fd 3 would overflow a recursive
  2291. // JSON.stringify; the host's iterative encoder measures it stack-safely
  2292. // and fails it deterministically on the byte budget (40 kB > 32 KiB).
  2293. const { runtime } = await setup()
  2294. const result = await runtime.run({
  2295. program: [
  2296. 'import os, json',
  2297. 'depth = 20000',
  2298. 'payload = "[" * depth + "]" * depth',
  2299. 'os.write(3, b\'{"type":"done","value":\' + payload.encode() + b\'}\\n\')',
  2300. 'import time',
  2301. 'time.sleep(5)',
  2302. ].join('\n'),
  2303. bindings: [],
  2304. })
  2305. expect(result.value).toBeUndefined()
  2306. expect(result.error?.kind).toBe('output-limit')
  2307. })
  2308. it('preserves a deeply nested completion value below the byte budget', async () => {
  2309. // CodeJsonValue has no depth limit: a 10000-deep nested list is only
  2310. // ~20 kB — under maxValueBytes — and must cross intact. That depth
  2311. // overflows BOTH recursive serializers the pipeline used to rely on
  2312. // (CPython's json.dumps recursion limit ~1000s, V8's JSON.stringify), so
  2313. // it proves the child-side _encode_json_plain and the host-side
  2314. // encodeJsonPlain together. The host JSON.parse of the frame is iterative
  2315. // in V8 for arrays, so only the two encoders were at risk.
  2316. const { runtime } = await setup()
  2317. const result = await runtime.run({
  2318. program: [
  2319. 'v = None',
  2320. 'for _ in range(10000):',
  2321. ' v = [v]',
  2322. 'return v',
  2323. ].join('\n'),
  2324. bindings: [],
  2325. })
  2326. expect(result.error).toBeUndefined()
  2327. // Walk down iteratively (a recursive toEqual would itself overflow).
  2328. let depth = 0
  2329. let cursor: unknown = result.value
  2330. while (Array.isArray(cursor)) {
  2331. expect(cursor).toHaveLength(1)
  2332. cursor = cursor[0]
  2333. depth++
  2334. }
  2335. expect(depth).toBe(10000)
  2336. expect(cursor).toBeNull()
  2337. })
  2338. it('bridges a deeply nested binding resolution back into the program stack-safely', async () => {
  2339. // A binding resolution has no seam-level depth or byte cap; neither the
  2340. // host's reply serialization nor the CHILD's reply decode may die on
  2341. // recursion (json.loads raises RecursionError ~10k levels deep; the
  2342. // bootstrap decodes frames iteratively). 12000 levels sits past that
  2343. // limit while staying tiny in bytes.
  2344. const { runtime } = await setup()
  2345. const deep = ((): unknown => {
  2346. let v: unknown = null
  2347. for (let i = 0; i < 12000; i++) v = [v]
  2348. return v
  2349. })()
  2350. const result = await runtime.run({
  2351. program: [
  2352. 'v = await tools.deep({})',
  2353. 'depth = 0',
  2354. 'while isinstance(v, list):',
  2355. ' v = v[0]',
  2356. ' depth += 1',
  2357. 'return depth',
  2358. ].join('\n'),
  2359. bindings: tools({ deep: async () => deep as never }),
  2360. })
  2361. expect(result.error).toBeUndefined()
  2362. expect(result.value).toBe(12000)
  2363. })
  2364. it('rejects a reserved errorClass name at the seam', async () => {
  2365. const { runtime } = await setup()
  2366. await expect(runtime.run({
  2367. program: 'return 1',
  2368. bindings: [{
  2369. global: 'tools',
  2370. functions: {},
  2371. errorClass: { name: 'class', memberNameProperty: 'toolName' },
  2372. }],
  2373. })).rejects.toThrow(/errorClass.name "class" is not a usable Python identifier/)
  2374. })
  2375. it('routes a declared inherited-attribute name through the bridge via subscript', async () => {
  2376. // __class__ resolves on `object` before any fallback hook; the proxy's
  2377. // __getattribute__ intercepts declared names first, and subscript access
  2378. // is the SDK-advertised route for underscore names.
  2379. const { runtime } = await setup()
  2380. const seen: string[] = []
  2381. const result = await runtime.run({
  2382. program: [
  2383. 'a = await tools["__class__"]({"via": "subscript"})',
  2384. 'b = await tools.__class__({"via": "dot"})',
  2385. 'return [a, b]',
  2386. ].join('\n'),
  2387. bindings: tools({
  2388. '__class__': async () => { seen.push('called'); return 'bridged' },
  2389. }),
  2390. })
  2391. expect(result.error).toBeUndefined()
  2392. expect(result.value).toEqual(['bridged', 'bridged'])
  2393. expect(seen).toEqual(['called', 'called'])
  2394. })
  2395. it('rejects NaN binding arguments immediately instead of hanging', async () => {
  2396. // Default json.dumps would emit a non-standard NaN token that the host
  2397. // JSON.parse drops silently, hanging the call until the wall clock;
  2398. // allow_nan=False raises in-program right away.
  2399. const { runtime } = await setup({ maxWallMs: 8000 })
  2400. const start = Date.now()
  2401. const result = await runtime.run({
  2402. program: [
  2403. 'caught = ""',
  2404. 'try:',
  2405. ' await tools.echo({"x": float("nan")})',
  2406. 'except RuntimeError as e:',
  2407. ' caught = str(e)',
  2408. 'return caught',
  2409. ].join('\n'),
  2410. bindings: tools({ echo: async args => args as CodeJsonValue }),
  2411. })
  2412. expect(result.error).toBeUndefined()
  2413. expect(result.value).toContain('lossless JSON')
  2414. expect(Date.now() - start).toBeLessThan(5000)
  2415. })
  2416. it('carries large binding arguments well past maxValueBytes', async () => {
  2417. // Binding traffic has no seam byte cap: a call frame far larger than the
  2418. // completion budget must reach the host intact (the fd-3 ceiling is a
  2419. // fixed memory-safety bound, not an output budget).
  2420. const maxValueBytes = 4096
  2421. const { runtime } = await setup({ maxValueBytes })
  2422. let receivedLength = 0
  2423. const result = await runtime.run({
  2424. program: [
  2425. `big = "B" * ${maxValueBytes * 50}`,
  2426. 'r = await tools.measure({"payload": big})',
  2427. 'return r',
  2428. ].join('\n'),
  2429. bindings: tools({
  2430. measure: async (args) => {
  2431. receivedLength = ((args as { payload: string }).payload).length
  2432. return receivedLength
  2433. },
  2434. }),
  2435. })
  2436. expect(result.error).toBeUndefined()
  2437. expect(receivedLength).toBe(maxValueBytes * 50)
  2438. expect(result.value).toBe(maxValueBytes * 50)
  2439. })
  2440. it('rejects an unknown binding name inside the program with a matching error', async () => {
  2441. const { runtime } = await setup()
  2442. const result = await runtime.run({
  2443. program: [
  2444. 'caught = ""',
  2445. 'try:',
  2446. ' await tools.nope({})',
  2447. 'except (AttributeError, RuntimeError) as e:',
  2448. ' caught = str(e)',
  2449. 'return caught',
  2450. ].join('\n'),
  2451. bindings: tools({ known: async () => 'ok' }),
  2452. })
  2453. expect(result.error).toBeUndefined()
  2454. expect(result.value).toContain('nope')
  2455. })
  2456. it('bounds an unknown-binding diagnostic built from a forged call frame', async () => {
  2457. // `call.global` and `call.name` carry no byte cap of their own, only the
  2458. // 64 MiB fd-3 frame parse cap, and the reply interpolated them raw: one copy
  2459. // into the template result, one into the `JSON.stringify` escape, one into
  2460. // the `encodeJsonPlain` frame, one into the pipe write. Slicing each field
  2461. // to `maxValueBytes` code units first makes an 8 MiB forged name a
  2462. // 128-byte reply. The observable effect is the reply the child then has to
  2463. // READ: its fd-3 reader is unbuffered, so `readline` consumes an oversized
  2464. // reply one `read(2)` per byte and the run's own legitimate call never gets
  2465. // answered — measured under a 60 s ceiling, the 8 MiB case timed out and a
  2466. // 64 MiB case cost the host 509.9 MiB of heap against 120.3 MiB with the
  2467. // slices in place. The child's address space stays generous enough to BUILD
  2468. // the forgery, which is not what is under test.
  2469. const { runtime } = await setup({ maxValueBytes: 128, addressSpaceMb: 1024, maxWallMs: 20_000 })
  2470. const result = await runtime.run({
  2471. program: [
  2472. 'import os',
  2473. 'frame = b\'{"type":"call","id":9001,"global":"tools","name":"\' + b"n" * (8 * 1024 * 1024) + b\'","args":{}}\\n\'',
  2474. // One os.write returns short past the pipe buffer, and a partial frame
  2475. // would glue itself to the next one and be dropped as malformed, so the
  2476. // forgery goes out through a drain loop.
  2477. 'view = memoryview(frame)',
  2478. 'while view:',
  2479. ' view = view[os.write(3, view):]',
  2480. // A legitimate call after the forgery: its reply can only arrive once
  2481. // the child has read past whatever the forged frame was answered with.
  2482. 'await tools.known({})',
  2483. 'return "settled"',
  2484. ].join('\n'),
  2485. bindings: tools({ known: async () => 'ok' }),
  2486. })
  2487. expect(result.error).toBeUndefined()
  2488. expect(result.value).toBe('settled')
  2489. }, 40_000)
  2490. it('bridges a binding call reached via subscript access (tools["name"])', async () => {
  2491. // The SDK tells the model `await tools["my-tool"](args)` works for exotic
  2492. // names; the proxy's __getitem__ must route it through the bridge.
  2493. const { runtime } = await setup()
  2494. const result = await runtime.run({
  2495. program: [
  2496. 'r = await tools["my-tool"]({"n": 7})',
  2497. 'return r',
  2498. ].join('\n'),
  2499. bindings: tools({ 'my-tool': async args => ({ got: args as CodeJsonValue }) }),
  2500. })
  2501. expect(result.error).toBeUndefined()
  2502. expect(result.value).toEqual({ got: { n: 7 } })
  2503. })
  2504. it('raises KeyError for an undeclared subscript name', async () => {
  2505. const { runtime } = await setup()
  2506. const result = await runtime.run({
  2507. program: [
  2508. 'caught = ""',
  2509. 'try:',
  2510. ' await tools["absent"]({})',
  2511. 'except KeyError as e:',
  2512. ' caught = str(e)',
  2513. 'return caught',
  2514. ].join('\n'),
  2515. bindings: tools({ known: async () => 'ok' }),
  2516. })
  2517. expect(result.error).toBeUndefined()
  2518. expect(result.value).toContain('absent')
  2519. })
  2520. })
  2521. describe('PythonCodeRuntime — budgets, termination, disposal', () => {
  2522. it('kills a wall-clock runaway program via SIGTERM/SIGKILL and reports timeout', async () => {
  2523. const { runtime } = await setup({ maxWallMs: 500, graceMs: 200 })
  2524. const start = Date.now()
  2525. const result = await runtime.run({
  2526. program: 'import time\nwhile True: time.sleep(1)',
  2527. bindings: [],
  2528. })
  2529. const elapsed = Date.now() - start
  2530. // The wall timer may fire first or the exit-after-signal may resolve; both are ok.
  2531. expect(['timeout', 'worker-exit']).toContain(result.error?.kind)
  2532. // We got somewhere in the neighborhood of maxWallMs, not the underlying `sleep(1)`.
  2533. expect(elapsed).toBeLessThan(2000)
  2534. }, 5000)
  2535. it('aborts a run when the outer signal fires mid-flight', async () => {
  2536. const { runtime } = await setup({ maxWallMs: 10_000 })
  2537. const controller = new AbortController()
  2538. const settled: Promise<CodeRunResult> = runtime.run({
  2539. program: 'import time\nwhile True: time.sleep(0.1)',
  2540. bindings: [],
  2541. signal: controller.signal,
  2542. })
  2543. setTimeout(() => { controller.abort('outer-abort') }, 200)
  2544. const result = await settled
  2545. expect(['abort', 'worker-exit']).toContain(result.error?.kind)
  2546. }, 5000)
  2547. it('settles the run when a mid-flight abort reason cannot be converted', async () => {
  2548. // The listener converted the reason before calling `finish()`, so a hostile
  2549. // reason threw from inside an `AbortSignal` listener. Node reports that as an
  2550. // uncaught exception — it can terminate the host — and `finish()` never ran,
  2551. // so the run stayed live until the wall ceiling and misreported as `timeout`
  2552. // (observed) instead of the caller's cancellation. `maxWallMs` is short so
  2553. // that misreport is a fast assertion failure rather than a suite timeout.
  2554. const uncaught: unknown[] = []
  2555. const record = (error: unknown): void => { uncaught.push(error) }
  2556. process.on('uncaughtException', record)
  2557. try {
  2558. const { runtime } = await setup({ maxWallMs: 4_000, graceMs: 200 })
  2559. const controller = new AbortController()
  2560. const settled: Promise<CodeRunResult> = runtime.run({
  2561. program: 'import time\nwhile True: time.sleep(0.1)',
  2562. bindings: [],
  2563. signal: controller.signal,
  2564. })
  2565. setTimeout(() => {
  2566. controller.abort({ [Symbol.toPrimitive]() { throw new Error('reason blew up') } })
  2567. }, 200)
  2568. const result = await settled
  2569. expect(result.error?.kind).toBe('abort')
  2570. expect(result.error?.message).toBe('<unrenderable rejection value>')
  2571. expect(uncaught).toEqual([])
  2572. } finally {
  2573. process.off('uncaughtException', record)
  2574. }
  2575. }, 15_000)
  2576. it('disposes to quiescence: an in-flight run resolves as abort and the child exits', async () => {
  2577. const { fiber, runtime } = await setup({ maxWallMs: 10_000 })
  2578. const pending = runtime.run({
  2579. program: 'import time\nwhile True: time.sleep(0.1)',
  2580. bindings: [],
  2581. })
  2582. // Give the process time to spawn and start running.
  2583. await new Promise(resolve => setTimeout(resolve, 200))
  2584. await fiber.dispose()
  2585. const result = await pending
  2586. expect(['abort', 'worker-exit']).toContain(result.error?.kind)
  2587. }, 5000)
  2588. it('reports a spawn failure via a bogus python binary as worker-exit', async () => {
  2589. const { runtime } = await setup({ pythonBin: '/nonexistent/python-binary', maxWallMs: 3000 })
  2590. const result = await runtime.run({
  2591. program: 'return 1',
  2592. bindings: [],
  2593. })
  2594. expect(result.error?.kind).toBe('worker-exit')
  2595. }, 8000)
  2596. it('applies the strictest of the configured and inherited resource limits', async () => {
  2597. // This case used to drive the bootstrap's `applying resource limits failed`
  2598. // handler with `cpuSeconds: 2 ** 63`, asserting that a cap the child cannot
  2599. // apply fails the run rather than running it uncapped. That premise no longer
  2600. // holds, for two independent reasons, so the test now pins what is actually
  2601. // guaranteed instead of a path no admissible input reaches.
  2602. //
  2603. // First, `2 ** 63` is not a safe integer, so it is now rejected at LOAD as a
  2604. // configuration error — it can never reach the child at all. Second, even the
  2605. // largest admissible values are applied successfully, because `_clamped`
  2606. // bounds every requested pair by the inherited hard limit: an unprivileged
  2607. // process may lower a hard limit but never raise one, so the child keeps the
  2608. // stricter of the two rather than asking for something `setrlimit` refuses.
  2609. // The failure handler remains as a substrate guard (a platform whose kernel
  2610. // refuses the call for its own reasons), but it is no longer reachable from
  2611. // configuration, and a test that pretends otherwise documents a contract the
  2612. // code does not have.
  2613. //
  2614. // What is observable: a very large cap still yields a working run, and the
  2615. // containment it promises is met by the inherited ceiling.
  2616. const { runtime } = await setup({ cpuSeconds: Number.MAX_SAFE_INTEGER - 1, maxWallMs: 10_000 })
  2617. const result = await runtime.run({ program: 'return 1', bindings: [] })
  2618. expect(result.error).toBeUndefined()
  2619. expect(result.value).toBe(1)
  2620. }, 20_000)
  2621. it('settles as worker-exit when the child exits before sending done (no hang)', async () => {
  2622. // Regression: settlement must key off `close` (process reaped AND stdio
  2623. // drained), not `exit`. With `exit`, finish() re-armed a second exit
  2624. // listener that never fired — run() hung forever whenever the exit event
  2625. // beat the final fd-3 data (deterministic on macOS, a lost race elsewhere).
  2626. const { runtime } = await setup({ maxWallMs: 10_000 })
  2627. const result = await runtime.run({
  2628. program: 'import os\nos._exit(7)',
  2629. bindings: [],
  2630. })
  2631. expect(result.error?.kind).toBe('worker-exit')
  2632. expect(result.error?.message).toContain('code=7')
  2633. }, 5000)
  2634. it('classifies RLIMIT_CPU soft-limit expiry (SIGXCPU) as a timeout', async () => {
  2635. // A CPU hot loop burns the soft limit; the kernel delivers SIGXCPU, whose
  2636. // close signal the host maps to `timeout`. macOS re-delivers SIGXCPU
  2637. // differently, so we assert only kind/message here — CI's darwin leg
  2638. // validates real delivery. cpuSeconds must be an integer for setrlimit.
  2639. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 20_000 })
  2640. const result = await runtime.run({
  2641. program: 'while True: pass',
  2642. bindings: [],
  2643. })
  2644. expect(result.error?.kind).toBe('timeout')
  2645. expect(result.error?.message).toContain('CPU time exhausted')
  2646. }, 8000)
  2647. it('keeps an early self-inflicted SIGKILL a worker-exit, not a CPU timeout', async () => {
  2648. // The unsolicited-SIGKILL-as-timeout classification applies only when the
  2649. // CPU budget could have expired (wall time >= cpuSeconds). A SIGKILL
  2650. // seconds before that (cgroup OOM, an operator, os.kill) is substrate
  2651. // death and stays worker-exit per the orthogonal taxonomy.
  2652. const { runtime } = await setup({ cpuSeconds: 60, maxWallMs: 10_000 })
  2653. const result = await runtime.run({
  2654. program: [
  2655. 'import os, signal',
  2656. 'os.kill(os.getpid(), signal.SIGKILL)',
  2657. ].join('\n'),
  2658. bindings: [],
  2659. })
  2660. expect(result.error?.kind).toBe('worker-exit')
  2661. expect(result.error?.message).toContain('SIGKILL')
  2662. })
  2663. it('charges a forked descendant against the run CPU budget', async () => {
  2664. // RLIMIT_CPU is per-process and every child inherits a FRESH budget, so a
  2665. // program that shells out multiplies `cpuSeconds` by the number of
  2666. // descendants it starts. Measured before the aggregate meter existed: with
  2667. // cpuSeconds 1, two sequential busy children burned 2.0 CPU-seconds
  2668. // (RUSAGE_CHILDREN) and the run still returned a SUCCESS completion. The
  2669. // settle-time check meters RUSAGE_SELF + RUSAGE_CHILDREN and converts the
  2670. // overrun into the same SIGXCPU the untrapped soft limit sends.
  2671. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 30_000 })
  2672. const result = await runtime.run({
  2673. program: [
  2674. 'import subprocess, sys',
  2675. 'for _ in range(2):',
  2676. ' subprocess.run([sys.executable, "-c", "import time\\nt=time.time()\\nwhile time.time()-t<1.2: pass"])',
  2677. 'return "escaped the cpu budget"',
  2678. ].join('\n'),
  2679. bindings: [],
  2680. })
  2681. // Darwin's SIGXCPU re-delivery differs, so accept either terminal
  2682. // classification; what must NOT happen is the completion crossing.
  2683. expect(['timeout', 'worker-exit']).toContain(result.error?.kind)
  2684. expect(result.value).toBeUndefined()
  2685. }, 40_000)
  2686. it('does not charge wall time or a cheap descendant against the CPU budget', async () => {
  2687. // The meter is CPU, not wall clock, and it must not fire on a child that
  2688. // burns almost nothing: a sleeping program and a trivial subprocess both
  2689. // have to complete normally, or the check would reject every program that
  2690. // shells out.
  2691. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 30_000 })
  2692. const slept = await runtime.run({
  2693. program: 'import time\ntime.sleep(1.5)\nreturn "slept"',
  2694. bindings: [],
  2695. })
  2696. expect(slept.error).toBeUndefined()
  2697. expect(slept.value).toBe('slept')
  2698. const cheap = await runtime.run({
  2699. program: [
  2700. 'import subprocess, sys',
  2701. 'subprocess.run([sys.executable, "-c", "pass"])',
  2702. 'return "cheap child"',
  2703. ].join('\n'),
  2704. bindings: [],
  2705. })
  2706. expect(cheap.error).toBeUndefined()
  2707. expect(cheap.value).toBe('cheap child')
  2708. }, 40_000)
  2709. it('spends no part of addressSpaceMb on bootstrap machinery', async () => {
  2710. // RLIMIT_AS counts RESERVED address space, so anything the bootstrap maps
  2711. // for its own accounting is subtracted from the program's `addressSpaceMb`.
  2712. // A sampling thread for the descendant-CPU meter cost 72 MiB here (an 8 MiB
  2713. // stack plus a 64 MiB glibc per-thread malloc arena reservation) and turned
  2714. // the 2-million-entry dict rejection below into a MemoryError under a
  2715. // 256 MiB cap on a slower runner. Assert the child's own mappings directly
  2716. // rather than inferring the budget from a near-cap allocation, so the bound
  2717. // is read from /proc instead of from how much headroom one machine happens
  2718. // to have; 48 MiB is well above the ~30 MiB a bare interpreter maps and
  2719. // well below the 102 MiB the thread produced. `addressSpaceMb` itself is
  2720. // skipped on darwin (the dyld shared cache makes any practical cap
  2721. // unsettable) and /proc/self/maps does not exist there, so the mapping
  2722. // assertion is Linux-only; the completion path is checked everywhere.
  2723. const { runtime } = await setup({ maxValueBytes: 4096, addressSpaceMb: 256 })
  2724. const mapped = await runtime.run({
  2725. program: [
  2726. 'import sys',
  2727. 'if sys.platform != "linux":',
  2728. ' return 0',
  2729. 'total = 0',
  2730. 'with open("/proc/self/maps") as handle:',
  2731. ' for line in handle:',
  2732. ' low, high = (int(part, 16) for part in line.split(" ", 1)[0].split("-"))',
  2733. ' total += high - low',
  2734. 'return total // (1024 * 1024)',
  2735. ].join('\n'),
  2736. bindings: [],
  2737. })
  2738. expect(mapped.error).toBeUndefined()
  2739. expect(mapped.value).toBeLessThan(48)
  2740. }, 20_000)
  2741. it('spends no part of addressSpaceMb on the reply pump, across a binding await', async () => {
  2742. // The test above measures BEFORE the program yields, so it could not see the
  2743. // reply pump's cost: `loop.run_in_executor(None, read_frame)` created the
  2744. // default executor's first thread on the first `await tools.*`, and that
  2745. // thread's 8 MiB stack plus a 64 MiB glibc per-thread malloc arena are
  2746. // charged to RLIMIT_AS while the limit is already in force — measured, the
  2747. // child went from 30.34 MiB to 102.39 MiB across one binding call. Under a
  2748. // small `addressSpaceMb` the thread cannot start and a legitimate call hangs
  2749. // to `maxWallMs`; under a larger one an allocation that should have fit dies
  2750. // as MemoryError. `loop.add_reader` watches the fd with no thread at all.
  2751. //
  2752. // Measuring both sides inside one run is what discriminates: a single
  2753. // after-the-fact number cannot separate the pump's cost from the
  2754. // interpreter's own footprint. Linux-only for the same reason as above.
  2755. const { runtime } = await setup({ addressSpaceMb: 256, maxWallMs: 20_000 })
  2756. const result = await runtime.run({
  2757. program: [
  2758. 'import sys',
  2759. 'def mapped():',
  2760. ' if sys.platform != "linux":',
  2761. ' return 0',
  2762. ' total = 0',
  2763. ' with open("/proc/self/maps") as handle:',
  2764. ' for line in handle:',
  2765. ' low, high = (int(part, 16) for part in line.split(" ", 1)[0].split("-"))',
  2766. ' total += high - low',
  2767. ' return total // (1024 * 1024)',
  2768. 'before = mapped()',
  2769. 'echoed = await tools.echo({"ping": True})',
  2770. 'return {"before": before, "after": mapped(), "echoed": echoed}',
  2771. ].join('\n'),
  2772. bindings: tools({ echo: async args => args as CodeJsonValue }),
  2773. })
  2774. expect(result.error).toBeUndefined()
  2775. const value = result.value as { before: number; after: number; echoed: unknown }
  2776. // The binding call really happened, so the pump really ran.
  2777. expect(value.echoed).toEqual({ ping: true })
  2778. // Awaiting a binding maps nothing extra. The 8 MiB allowance absorbs ordinary
  2779. // heap growth while staying far below the 72 MiB a pump thread cost.
  2780. expect(value.after - value.before).toBeLessThan(8)
  2781. }, 30_000)
  2782. it('still terminates a program that ignores SIGXCPU (hard-limit backstop)', async () => {
  2783. // A hot loop under SIG_IGN burns through the soft limit; the kernel's
  2784. // hard limit (cpuSeconds + 1) SIGKILLs it. Only a kernel-authoritative
  2785. // SIGXCPU close classifies as the CPU timeout — a bare SIGKILL is
  2786. // indistinguishable from a cgroup OOM kill, so it reports worker-exit
  2787. // (Darwin re-delivers SIGXCPU instead, where the wall clock settles it
  2788. // as timeout). Either way the run TERMINATES within the budget — the
  2789. // backstop holds even when the classification is the opaque one.
  2790. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 6_000 })
  2791. const result = await runtime.run({
  2792. program: [
  2793. 'import signal',
  2794. 'signal.signal(signal.SIGXCPU, signal.SIG_IGN)',
  2795. 'while True: pass',
  2796. ].join('\n'),
  2797. bindings: [],
  2798. })
  2799. expect(['timeout', 'worker-exit']).toContain(result.error?.kind)
  2800. }, 12_000)
  2801. it('enforces the CPU budget even when the program monkeypatches the enforcement primitives', async () => {
  2802. // The check uses import-time-captured references, so replacing
  2803. // resource.getrusage / signal.signal / os.kill on the modules cannot
  2804. // defang it: a trapping program that also swaps the callables and burns
  2805. // past the budget still dies by the authoritative SIGXCPU.
  2806. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 15_000 })
  2807. const result = await runtime.run({
  2808. program: [
  2809. 'import signal, os, resource, time',
  2810. 'signal.signal(signal.SIGXCPU, lambda *a: None)',
  2811. 'resource.getrusage = lambda *a: (_ for _ in ()).throw(RuntimeError("nope"))',
  2812. 'os.kill = lambda *a: None',
  2813. 'signal.signal = lambda *a: None',
  2814. 'deadline = time.process_time() + 1.05',
  2815. 'while time.process_time() < deadline: pass',
  2816. 'return "escaped"',
  2817. ].join('\n'),
  2818. bindings: [],
  2819. })
  2820. expect(result.error?.kind).toBe('timeout')
  2821. expect(result.value).toBeUndefined()
  2822. }, 15_000)
  2823. it('re-delivers SIGXCPU when a trapping program returns inside the soft-to-hard gap', async () => {
  2824. // A program can trap SIGXCPU and settle during the one-second gap; the
  2825. // bootstrap re-checks the kernel CPU meter (getrusage) after settlement
  2826. // and dies by SIGXCPU with the default disposition restored, so the host
  2827. // still classifies the exhausted budget as a timeout instead of success.
  2828. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 15_000 })
  2829. const result = await runtime.run({
  2830. program: [
  2831. 'import signal, time',
  2832. 'fired = []',
  2833. 'signal.signal(signal.SIGXCPU, lambda *a: fired.append(1))',
  2834. 'deadline = time.process_time() + 1.05',
  2835. 'while time.process_time() < deadline: pass',
  2836. 'return "escaped"',
  2837. ].join('\n'),
  2838. bindings: [],
  2839. })
  2840. expect(result.error?.kind).toBe('timeout')
  2841. expect(result.error?.message).toContain('CPU time exhausted')
  2842. expect(result.value).toBeUndefined()
  2843. }, 15_000)
  2844. it('enforces the CPU budget when the program rebinds the enforcer on __main__', async () => {
  2845. // The bootstrap IS `__main__`, so `import __main__` reaches its globals.
  2846. // The enforcement callable holds its primitives in closure cells (not
  2847. // module attributes) and `_run` reads the callable into a frame local
  2848. // before the program starts, so neither replacing the global nor swapping
  2849. // the module's captured names changes what runs after settlement.
  2850. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 15_000 })
  2851. const result = await runtime.run({
  2852. program: [
  2853. 'import signal, time, __main__',
  2854. 'signal.signal(signal.SIGXCPU, lambda *a: None)',
  2855. '__main__._DIE_IF_CPU_EXHAUSTED = lambda *_: None',
  2856. 'deadline = time.process_time() + 1.05',
  2857. 'while time.process_time() < deadline: pass',
  2858. 'return "escaped"',
  2859. ].join('\n'),
  2860. bindings: [],
  2861. })
  2862. expect(result.error?.kind).toBe('timeout')
  2863. expect(result.value).toBeUndefined()
  2864. }, 15_000)
  2865. it('bounds a program that defeats the post-check by writing its closure cell', async () => {
  2866. // The closure-cell capture raises the cost of defeating the post-check; it
  2867. // does NOT make it unreachable, and nothing in-process could: a cell is
  2868. // writable through `fn.__closure__[i].cell_contents`, and `sys._getframe`
  2869. // reads _run's frame locals. This program does exactly that — walks to
  2870. // _run's frame, takes the enforcement callable, and replaces its captured
  2871. // `getrusage` with one reporting zero CPU used — then burns past cpuSeconds
  2872. // with SIGXCPU trapped. The run must still fail, because the bound that
  2873. // model code cannot forge is outside the interpreter: the RLIMIT_CPU HARD
  2874. // limit at cpuSeconds + 1, whose SIGKILL admits no handler. No success is
  2875. // reportable either way.
  2876. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 20_000 })
  2877. const start = Date.now()
  2878. const result = await runtime.run({
  2879. program: [
  2880. 'import signal, sys, time',
  2881. 'signal.signal(signal.SIGXCPU, lambda *a: None)',
  2882. // Walk out of __dsh_main__ to _run's frame and take its local.
  2883. 'die = None',
  2884. 'depth = 1',
  2885. 'while depth < 12:',
  2886. ' frame = sys._getframe(depth)',
  2887. ' if "die_if_cpu_exhausted" in frame.f_locals:',
  2888. ' die = frame.f_locals["die_if_cpu_exhausted"]',
  2889. ' break',
  2890. ' depth += 1',
  2891. 'assert die is not None, "enforcer not reachable from the frame chain"',
  2892. 'class Zero:',
  2893. ' ru_utime = 0.0',
  2894. ' ru_stime = 0.0',
  2895. 'names = die.__code__.co_freevars',
  2896. 'die.__closure__[names.index("getrusage")].cell_contents = lambda *a: Zero()',
  2897. // Burn well past the soft limit into the hard limit's SIGKILL.
  2898. 'while True: pass',
  2899. ].join('\n'),
  2900. bindings: [],
  2901. })
  2902. // What holds on EVERY platform: the tampering bought no success. The run
  2903. // failed, carried no value, and the reported kind is one of the two
  2904. // kernel-level outcomes — never a completion.
  2905. expect(result.value).toBeUndefined()
  2906. expect(result.error?.kind === 'worker-exit' || result.error?.kind === 'timeout').toBe(true)
  2907. if (process.platform === 'linux') {
  2908. // Linux enforces the RLIMIT_CPU HARD limit at cpuSeconds + 1 promptly, so
  2909. // the CPU bound — not the 20 s wall ceiling — is what stops the program.
  2910. // Its SIGKILL is not SIGXCPU, so the orthogonal-failure taxonomy reports
  2911. // `worker-exit`: a bare SIGKILL is not evidence of CPU burn.
  2912. expect(result.error?.kind).toBe('worker-exit')
  2913. expect(Date.now() - start).toBeLessThan(15_000)
  2914. } else {
  2915. // Darwin does not deliver the hard limit's SIGKILL on the same schedule;
  2916. // observed on the macOS lane, a program that patches the post-check runs
  2917. // to the WALL ceiling instead. The CPU budget is therefore not the
  2918. // binding constraint against a tampering program there — the wall clock
  2919. // is. Asserted rather than skipped so the difference stays visible.
  2920. expect(result.error?.kind).toBe('timeout')
  2921. }
  2922. }, 30_000)
  2923. it('keeps a finished-but-not-closed run live so dispose awaits the child\'s death', async () => {
  2924. // finish() no longer drops the run from `live`; settle() (at close) does.
  2925. // A SIGTERM-trapping program with a small graceMs sits in the grace window
  2926. // after finish() fires — dispose() must not resolve until the SIGKILL
  2927. // backstop actually reaps the child. The program prints its pid (captured
  2928. // as a log even on abort); once dispose() resolves, that pid must be dead
  2929. // (process.kill(pid, 0) throws ESRCH).
  2930. const { fiber, runtime } = await setup({ maxWallMs: 10_000, graceMs: 400 })
  2931. // Deterministic readiness: the program reports its pid through a binding
  2932. // AFTER installing the trap, so dispose cannot race the spawn (a fixed
  2933. // sleep lost that race on slow CI runners — SIGTERM landed pre-trap).
  2934. let reportedPid!: (pid: number) => void
  2935. const trapReady = new Promise<number>((resolve) => { reportedPid = resolve })
  2936. const pending = runtime.run({
  2937. program: [
  2938. 'import signal, time, os',
  2939. 'signal.signal(signal.SIGTERM, lambda *a: None)',
  2940. 'await tools.ready({"pid": os.getpid()})',
  2941. 'while True: time.sleep(0.05)',
  2942. ].join('\n'),
  2943. bindings: tools({
  2944. ready: async (args) => {
  2945. reportedPid((args as { pid: number }).pid)
  2946. return 'ok'
  2947. },
  2948. }),
  2949. })
  2950. const pid = await trapReady
  2951. const start = Date.now()
  2952. await fiber.dispose()
  2953. const elapsed = Date.now() - start
  2954. const result = await pending
  2955. expect(['abort', 'worker-exit', 'timeout']).toContain(result.error?.kind)
  2956. // dispose() returned only after the grace window elapsed (the SIGTERM trap
  2957. // forces the SIGKILL backstop path), proving the run stayed live past finish().
  2958. expect(elapsed).toBeGreaterThanOrEqual(300)
  2959. expect(Number.isInteger(pid) && pid > 0).toBe(true)
  2960. // The child is fully reaped by the time dispose() resolved.
  2961. expect(() => process.kill(pid, 0)).toThrow(/ESRCH/)
  2962. }, 8000)
  2963. it('settles on the decided result even when a setsid-escaped orphan holds stdio open past close', async () => {
  2964. // `close` only fires once every inherited stdio stream drains. A descendant
  2965. // started with start_new_session=True escapes the child's process group, so
  2966. // the SIGTERM/SIGKILL aimed at that group never reaches it; if it inherited
  2967. // our stdout/stderr/fd 3 and outlives the run, `close` would never fire and
  2968. // run() would hang forever. The close-deadline backstop (graceMs + margin)
  2969. // must force settlement on the value the `done` frame already decided.
  2970. const { runtime } = await setup({ graceMs: 100 })
  2971. const start = Date.now()
  2972. const result = await runtime.run({
  2973. program: [
  2974. 'import subprocess, sys',
  2975. // Orphan in a fresh session, inheriting our stdout/stderr/fd 3, alive
  2976. // past the close-deadline so `close` cannot fire on its own. Its own
  2977. // 5 s self-exit is the leak ceiling AND the discriminator: it must stay
  2978. // ABOVE the < 4000 ms upper-bound assertion below, so if the deadline
  2979. // backstop failed to settle, settlement could only come from this
  2980. // self-exit at ~5 s and blow the bound — a sharper signal than the wall
  2981. // ceiling would give.
  2982. 'subprocess.Popen([sys.executable, "-c", "import time; time.sleep(5)"],',
  2983. ' start_new_session=True)',
  2984. 'return "escaped"',
  2985. ].join('\n'),
  2986. bindings: [],
  2987. })
  2988. const elapsed = Date.now() - start
  2989. // The done frame decided the value; the deadline settled it despite the
  2990. // orphan pinning the pipes open.
  2991. expect(result.error).toBeUndefined()
  2992. expect(result.value).toBe('escaped')
  2993. // Settlement waited for the backstop (graceMs + CLOSE_REAP_MARGIN_MS ≈ 2.1s),
  2994. // not the orphan's 5 s self-exit — proving the deadline, not a fallback, fired.
  2995. expect(elapsed).toBeGreaterThanOrEqual(1_500)
  2996. expect(elapsed).toBeLessThan(4_000)
  2997. }, 8000)
  2998. it('flushes a newline-free diagnostic when the closeDeadline forces settlement', async () => {
  2999. // A leader that writes an unterminated diagnostic via `os.write(1, ...)` and
  3000. // then exits, leaving a setsid orphan holding the pipes open, settles through
  3001. // the closeDeadline destroy() path — which fires no `end`. The residual must
  3002. // be flushed before destroy() drops it, or the diagnostic is lost from
  3003. // `logs`. The value is decided by the done frame; the diagnostic must survive.
  3004. const { runtime } = await setup({ graceMs: 100 })
  3005. const result = await runtime.run({
  3006. program: [
  3007. 'import os, subprocess, sys',
  3008. 'os.write(1, b"leader-diagnostic-no-newline")',
  3009. 'subprocess.Popen([sys.executable, "-c", "import time; time.sleep(5)"],',
  3010. ' start_new_session=True)',
  3011. 'return "escaped"',
  3012. ].join('\n'),
  3013. bindings: [],
  3014. })
  3015. expect(result.error).toBeUndefined()
  3016. expect(result.value).toBe('escaped')
  3017. expect(result.logs).toContain('leader-diagnostic-no-newline')
  3018. }, 8000)
  3019. it('closes the child stdin so a program read sees EOF instead of blocking', async () => {
  3020. // The host closes the child's stdin write handle immediately after spawn
  3021. // (the program is an async body that reads nothing from fd 0; a live pipe
  3022. // would hold a host-side handle open past the run). A program that DOES
  3023. // read fd 0 therefore sees EOF at once. Fail-before: with the handle left
  3024. // open and no data written, `sys.stdin.read()` blocks and the run would
  3025. // hang to maxWallMs as a timeout.
  3026. const { runtime } = await setup({ maxWallMs: 8_000 })
  3027. const result = await runtime.run({
  3028. program: [
  3029. 'import sys',
  3030. 'data = sys.stdin.read()',
  3031. 'return "read: " + repr(data)',
  3032. ].join('\n'),
  3033. bindings: [],
  3034. })
  3035. expect(result.error).toBeUndefined()
  3036. expect(result.value).toBe("read: ''")
  3037. }, 15_000)
  3038. it('keeps runtime type annotations as live classes, not PEP 563 strings, when the program reads them', async () => {
  3039. // bootstrap.py imports `from __future__ import annotations`; without
  3040. // dont_inherit=True on compile(), that PEP 563 flag leaks into the program's
  3041. // compiled code and stringifies its type annotations, changing the semantics
  3042. // of a legal program that reads `f.__annotations__` at runtime.
  3043. const { runtime } = await setup()
  3044. const result = await runtime.run({
  3045. program: [
  3046. 'def f(x: int) -> int:',
  3047. ' return x',
  3048. 'return f.__annotations__["x"].__name__',
  3049. ].join('\n'),
  3050. bindings: [],
  3051. })
  3052. expect(result.error).toBeUndefined()
  3053. expect(result.value).toBe('int')
  3054. }, 15_000)
  3055. it('reaps a same-group child that ignores SIGTERM and releases the pipes before close', async () => {
  3056. // The same-group counterpart to the setsid-orphan case above. A descendant
  3057. // left in the child's OWN process group (no setsid, so `kill(-pid)` reaches
  3058. // it) can ignore SIGTERM yet still release the inherited stdout/stderr/fd 3
  3059. // it does not hold — here by giving the Popen child DEVNULL streams and
  3060. // letting close_fds drop fd 3. The leader then writes `done` and exits, its
  3061. // `close` fires because the pipes drained, and settle() runs while that
  3062. // descendant is still alive. settle() then keeps a REF'd poll alive until the
  3063. // grace-window SIGKILL has emptied the whole process group, so the host cannot
  3064. // exit and reparent the survivor to init: no subprocess outlives the fiber.
  3065. //
  3066. // The descendant must have SIG_IGN installed BEFORE the host sends SIGTERM,
  3067. // or it dies from the default SIGTERM whether the fix is present or not — so
  3068. // it writes a readiness marker after trapping and the leader waits for that
  3069. // marker before returning. While alive it bumps a heartbeat file every 50 ms;
  3070. // the test asserts the heartbeat STOPS, which is what "no longer executing"
  3071. // means whether the killed descendant is reaped or lingers as a zombie (a
  3072. // SIGKILL'd process runs no more code either way). It sleeps 30 s as a safety
  3073. // net so a broken fix cannot leak it forever.
  3074. const handoff = await mkdtemp(join(tmpdir(), 'dsh-samegroup-'))
  3075. const readyMarker = join(handoff, 'ready')
  3076. const heartbeat = join(handoff, 'heartbeat')
  3077. const { runtime } = await setup({ maxWallMs: 10_000, graceMs: 300 })
  3078. const result = await runtime.run({
  3079. program: [
  3080. 'import subprocess, sys, os, time',
  3081. `marker = ${JSON.stringify(readyMarker)}`,
  3082. `heartbeat = ${JSON.stringify(heartbeat)}`,
  3083. // Same group (no start_new_session); ignores SIGTERM; holds none of the
  3084. // leader's pipes (DEVNULL std streams, close_fds drops fd 3). It writes
  3085. // the marker (argv[1]) only AFTER the trap is installed — so the leader
  3086. // cannot return, and the host cannot send SIGTERM, before it is ignored —
  3087. // then rewrites the heartbeat (argv[2]) every 50 ms for up to 30 s.
  3088. 'code = ("import signal, sys, time\\n"',
  3089. ' "signal.signal(signal.SIGTERM, signal.SIG_IGN)\\n"',
  3090. ' "open(sys.argv[1], \'w\').close()\\n"',
  3091. ' "end = time.time() + 30\\n"',
  3092. ' "while time.time() < end:\\n"',
  3093. ' " open(sys.argv[2], \'w\').close()\\n"',
  3094. ' " time.sleep(0.05)\\n")',
  3095. 'child = subprocess.Popen([sys.executable, "-c", code, marker, heartbeat],',
  3096. ' stdin=subprocess.DEVNULL,',
  3097. ' stdout=subprocess.DEVNULL,',
  3098. ' stderr=subprocess.DEVNULL)',
  3099. 'deadline = time.time() + 5',
  3100. 'while not os.path.exists(marker) and time.time() < deadline:',
  3101. ' time.sleep(0.02)',
  3102. 'return "spawned"',
  3103. ].join('\n'),
  3104. bindings: [],
  3105. })
  3106. expect(result.error).toBeUndefined()
  3107. expect(result.value).toBe('spawned')
  3108. // The trap really installed before the leader returned, so this is the
  3109. // SIGTERM-ignoring descendant, not one that would have died to the default.
  3110. expect(existsSync(readyMarker)).toBe(true)
  3111. // The grace-window SIGKILL (graceMs 300 + reap margin) empties the group. Once
  3112. // it has, the descendant stops bumping the heartbeat. Poll the heartbeat's
  3113. // mtime: two consecutive reads far enough apart with no change means it is no
  3114. // longer executing — true whether it was reaped or lingers as a zombie, so
  3115. // the assertion holds in a container whose init does not wait() orphans. The
  3116. // window (well under the 30 s self-timeout) proves the SIGKILL did the work.
  3117. const mtime = (): number => { try { return statSync(heartbeat).mtimeMs } catch { return 0 } }
  3118. const stopDeadline = Date.now() + 8_000
  3119. let last = mtime()
  3120. let still = false
  3121. while (Date.now() < stopDeadline) {
  3122. await new Promise(resolve => setTimeout(resolve, 400))
  3123. const now = mtime()
  3124. if (now === last && now !== 0) { still = true; break }
  3125. last = now
  3126. }
  3127. expect(still).toBe(true)
  3128. }, 20_000)
  3129. it('dispose awaits reaping of a same-group survivor from a completed run', async () => {
  3130. // The quiescence contract also holds for a run that ALREADY resolved: the run
  3131. // stays tracked in `live` until its process group is reaped, so a `dispose()`
  3132. // that races a just-returned run() still awaits the survivor rather than
  3133. // snapshotting an empty `live` and returning while it lives. Here the run
  3134. // completes (leaving a SIGTERM-ignoring same-group descendant), then dispose()
  3135. // is called; the heartbeat must be stale BY THE TIME dispose() resolves —
  3136. // proving teardown waited for the reap, not merely that the reap eventually
  3137. // happened.
  3138. const handoff = await mkdtemp(join(tmpdir(), 'dsh-dispose-quiesce-'))
  3139. const readyMarker = join(handoff, 'ready')
  3140. const heartbeat = join(handoff, 'heartbeat')
  3141. const { runtime, fiber } = await setup({ maxWallMs: 10_000, graceMs: 300 })
  3142. const result = await runtime.run({
  3143. program: [
  3144. 'import subprocess, sys, os, time',
  3145. `marker = ${JSON.stringify(readyMarker)}`,
  3146. `heartbeat = ${JSON.stringify(heartbeat)}`,
  3147. 'code = ("import signal, sys, time\\n"',
  3148. ' "signal.signal(signal.SIGTERM, signal.SIG_IGN)\\n"',
  3149. ' "open(sys.argv[1], \'w\').close()\\n"',
  3150. ' "end = time.time() + 30\\n"',
  3151. ' "while time.time() < end:\\n"',
  3152. ' " open(sys.argv[2], \'w\').close()\\n"',
  3153. ' " time.sleep(0.05)\\n")',
  3154. 'child = subprocess.Popen([sys.executable, "-c", code, marker, heartbeat],',
  3155. ' stdin=subprocess.DEVNULL,',
  3156. ' stdout=subprocess.DEVNULL,',
  3157. ' stderr=subprocess.DEVNULL)',
  3158. 'deadline = time.time() + 5',
  3159. 'while not os.path.exists(marker) and time.time() < deadline:',
  3160. ' time.sleep(0.02)',
  3161. 'return "spawned"',
  3162. ].join('\n'),
  3163. bindings: [],
  3164. })
  3165. expect(result.error).toBeUndefined()
  3166. expect(existsSync(readyMarker)).toBe(true)
  3167. // dispose() must not return until the group is reaped. After it resolves, the
  3168. // heartbeat must already be stale: read its mtime, wait past the heartbeat
  3169. // interval, and confirm it did not advance — the descendant is no longer
  3170. // executing (reaped or zombie), so teardown was genuinely quiescent.
  3171. await fiber.dispose()
  3172. const mtime = (): number => { try { return statSync(heartbeat).mtimeMs } catch { return 0 } }
  3173. const afterDispose = mtime()
  3174. // Pin the assertion to a heartbeat that actually ran: mtime() returns 0 when
  3175. // the file never existed, so without this the `toBe` below would pass
  3176. // vacuously (0 === 0) if the survivor never wrote a heartbeat at all.
  3177. expect(afterDispose).toBeGreaterThan(0)
  3178. await new Promise(resolve => setTimeout(resolve, 500))
  3179. expect(mtime()).toBe(afterDispose)
  3180. }, 20_000)
  3181. it('sends SIGKILL at the poll deadline when the event loop was blocked past both timers', async () => {
  3182. // If the host event loop is blocked (a big synchronous computation) from
  3183. // before the group-reap poll was scheduled until after the deadline, both the
  3184. // poll timer and the grace-window SIGKILL timer are overdue when the loop
  3185. // resumes. Node runs the earlier-scheduled poll first, so the SIGKILL timer
  3186. // may not have fired yet. The deadline arm must then send SIGKILL ITSELF
  3187. // rather than cancel the unfired escalation — otherwise a SIGTERM-ignoring
  3188. // same-group survivor is released for good. A synchronous busy-loop after
  3189. // run() resolves reproduces the block deterministically.
  3190. const handoff = await mkdtemp(join(tmpdir(), 'dsh-deadline-'))
  3191. const readyMarker = join(handoff, 'ready')
  3192. const heartbeat = join(handoff, 'heartbeat')
  3193. const graceMs = 300
  3194. const { runtime } = await setup({ maxWallMs: 10_000, graceMs })
  3195. const result = await runtime.run({
  3196. program: [
  3197. 'import subprocess, sys, os, time',
  3198. `marker = ${JSON.stringify(readyMarker)}`,
  3199. `heartbeat = ${JSON.stringify(heartbeat)}`,
  3200. 'code = ("import signal, sys, time\\n"',
  3201. ' "signal.signal(signal.SIGTERM, signal.SIG_IGN)\\n"',
  3202. ' "open(sys.argv[1], \'w\').close()\\n"',
  3203. ' "end = time.time() + 30\\n"',
  3204. ' "while time.time() < end:\\n"',
  3205. ' " open(sys.argv[2], \'w\').close()\\n"',
  3206. ' " time.sleep(0.05)\\n")',
  3207. 'child = subprocess.Popen([sys.executable, "-c", code, marker, heartbeat],',
  3208. ' stdin=subprocess.DEVNULL,',
  3209. ' stdout=subprocess.DEVNULL,',
  3210. ' stderr=subprocess.DEVNULL)',
  3211. 'deadline = time.time() + 5',
  3212. 'while not os.path.exists(marker) and time.time() < deadline:',
  3213. ' time.sleep(0.02)',
  3214. 'return "spawned"',
  3215. ].join('\n'),
  3216. bindings: [],
  3217. })
  3218. expect(result.error).toBeUndefined()
  3219. expect(existsSync(readyMarker)).toBe(true)
  3220. // Block the event loop synchronously past graceMs + CLOSE_REAP_MARGIN_MS
  3221. // (2000) with margin, so both timers are overdue when the loop resumes.
  3222. const blockUntil = Date.now() + graceMs + 2_000 + 800
  3223. while (Date.now() < blockUntil) { /* busy-wait, no yield */ }
  3224. // Yield: the overdue poll runs (group still non-empty, deadline passed) and
  3225. // must send SIGKILL itself. The survivor then stops bumping the heartbeat.
  3226. const mtime = (): number => { try { return statSync(heartbeat).mtimeMs } catch { return 0 } }
  3227. const stopDeadline = Date.now() + 5_000
  3228. let last = mtime()
  3229. let stopped = false
  3230. while (Date.now() < stopDeadline) {
  3231. await new Promise(resolve => setTimeout(resolve, 400))
  3232. const now = mtime()
  3233. if (now === last && now !== 0) { stopped = true; break }
  3234. last = now
  3235. }
  3236. expect(stopped).toBe(true)
  3237. }, 20_000)
  3238. })
  3239. describe('PythonCodeRuntime — hostile peer', () => {
  3240. it('drops garbage bytes and unknown-shape frames posted directly to fd 3', async () => {
  3241. // The model program can reach fd 3 and write anything. We inject a
  3242. // non-JSON line, a valid JSON but unknown-shape frame, and a broken done
  3243. // frame; the host must not crash, and the real `done` still settles the run.
  3244. const { runtime } = await setup()
  3245. const result = await runtime.run({
  3246. program: [
  3247. 'import os',
  3248. 'os.write(3, b"not-json\\n")',
  3249. 'os.write(3, b\'{"type":"unknown"}\\n\')',
  3250. 'os.write(3, b\'{"type":"done","error":{"message":42}}\\n\')',
  3251. 'return "survived"',
  3252. ].join('\n'),
  3253. bindings: [],
  3254. })
  3255. expect(result.error).toBeUndefined()
  3256. expect(result.value).toBe('survived')
  3257. })
  3258. it('answers a forged call frame for an unknown binding and never crashes', async () => {
  3259. // The unknown-binding reply path, driven through the id the host expects:
  3260. // the program lets its own first call claim id 0 and forges id 1, which the
  3261. // host answers with the `unknown binding` rejection the honest call would
  3262. // have received. A forged id out of sequence is dropped instead — that is
  3263. // the id-bound test below, not this one.
  3264. const { runtime } = await setup({ maxWallMs: 8_000 })
  3265. let seenLegitCall = false
  3266. const result = await runtime.run({
  3267. program: [
  3268. 'import os, json',
  3269. 'x = await tools.echo({"ping": True})',
  3270. 'os.write(3, json.dumps({"type":"call","id":1,"global":"tools","name":"forged","args":{}}).encode() + b"\\n")',
  3271. // The forged frame is answered, but nothing in the child awaits id 1, so
  3272. // the reply is ignored and the run completes on its own value.
  3273. 'return x',
  3274. ].join('\n'),
  3275. bindings: tools({
  3276. echo: async (args) => { seenLegitCall = true; return args as CodeJsonValue },
  3277. }),
  3278. })
  3279. expect(result.error).toBeUndefined()
  3280. expect(result.value).toEqual({ ping: true })
  3281. expect(seenLegitCall).toBe(true)
  3282. }, 15_000)
  3283. it('drops forged call frames whose ids are not the next in sequence, retaining no per-id state', async () => {
  3284. // The host used to remember every answered id in a Set, so a program could
  3285. // write an unbounded run of unique forged ids — each frame far below the
  3286. // 64 MiB cap, so nothing rejected them — and grow host memory for the
  3287. // whole run. Ids are consecutive from 0, so one counter replaces the set.
  3288. //
  3289. // The discriminator is that the forgeries must not be answered. Each names a
  3290. // binding that does exist, so a host answering them would run `echo` once
  3291. // per forgery; the count proves only the legitimate call was dispatched.
  3292. // Ids also run DESCENDING, so a high-water-mark test would drop the honest
  3293. // call that follows rather than the forgeries.
  3294. const { runtime } = await setup()
  3295. let echoCalls = 0
  3296. const result = await runtime.run({
  3297. program: [
  3298. 'import os, json',
  3299. 'for i in range(2000, 0, -1):',
  3300. ' os.write(3, json.dumps({"type":"call","id":i,"global":"tools","name":"echo","args":{"forged":i}}).encode() + b"\\n")',
  3301. 'x = await tools.echo({"ping": True})',
  3302. 'return x',
  3303. ].join('\n'),
  3304. bindings: tools({
  3305. echo: async (args) => { echoCalls += 1; return args as CodeJsonValue },
  3306. }),
  3307. })
  3308. expect(result.error).toBeUndefined()
  3309. expect(result.value).toEqual({ ping: true })
  3310. expect(echoCalls).toBe(1)
  3311. }, 15_000)
  3312. it('keeps answering calls a program makes after one with unserializable arguments', async () => {
  3313. // The child claims an id only once its write succeeds, so a call rejected
  3314. // child-side for non-lossless arguments leaves no gap. Were a gap possible,
  3315. // the host's exact-successor test would drop every later call and the run
  3316. // would hang to the wall ceiling instead of completing.
  3317. const { runtime } = await setup({ maxWallMs: 8_000 })
  3318. const seen: unknown[] = []
  3319. const result = await runtime.run({
  3320. program: [
  3321. 'caught = ""',
  3322. 'try:',
  3323. ' await tools.echo({"bad": float("inf")})',
  3324. 'except RuntimeError as e:',
  3325. ' caught = str(e)',
  3326. 'after = await tools.echo({"ok": True})',
  3327. 'return {"caught": caught, "after": after}',
  3328. ].join('\n'),
  3329. bindings: tools({
  3330. echo: async (args) => { seen.push(args); return args as CodeJsonValue },
  3331. }),
  3332. })
  3333. expect(result.error).toBeUndefined()
  3334. const value = result.value as { caught: string; after: unknown }
  3335. expect(value.caught).toContain('lossless JSON')
  3336. expect(value.after).toEqual({ ok: true })
  3337. // The rejected call never reached the host; the one after it did.
  3338. expect(seen).toEqual([{ ok: true }])
  3339. }, 15_000)
  3340. it('drops a forged frame carrying an integer outside JavaScript safe range', async () => {
  3341. // JSON.parse would silently round 9007199254740993 to ...992 BEFORE any
  3342. // validation, corrupting a dispatched argument or completion. The host
  3343. // scans the raw line and drops such frames as hostile traffic; the honest
  3344. // child cannot produce one (its validator rejects unsafe ints).
  3345. const { runtime } = await setup()
  3346. let dispatched: unknown
  3347. const result = await runtime.run({
  3348. program: [
  3349. 'import os',
  3350. // Forged call frame with an unsafe int argument, then a forged done
  3351. // frame with an unsafe int value — both must be dropped whole.
  3352. 'os.write(3, b\'{"type":"call","id":7,"global":"tools","name":"echo","args":9007199254740993}\\n\')',
  3353. 'os.write(3, b\'{"type":"done","value":9007199254740993}\\n\')',
  3354. 'x = await tools.echo({"ok": True})',
  3355. 'return x',
  3356. ].join('\n'),
  3357. bindings: tools({
  3358. echo: async (args) => { dispatched = args; return args as CodeJsonValue },
  3359. }),
  3360. })
  3361. expect(result.error).toBeUndefined()
  3362. // The forged done did not settle the run; the legit call and completion did.
  3363. expect(result.value).toEqual({ ok: true })
  3364. expect(dispatched).toEqual({ ok: true })
  3365. })
  3366. it('truncates host-side logs once the budget is exhausted and emits the marker', async () => {
  3367. // Set a tiny host-side budget; the Python side has a much larger one, so
  3368. // its LogBuffer will not truncate — the host ledger fires first.
  3369. const { runtime } = await setup({ maxLogBytes: 128 })
  3370. const result = await runtime.run({
  3371. program: [
  3372. 'for _ in range(50):',
  3373. ' print("aaaaaaaaaa")',
  3374. 'return "done"',
  3375. ].join('\n'),
  3376. bindings: [],
  3377. })
  3378. expect(result.error).toBeUndefined()
  3379. const markers = result.logs.filter(line => line.includes('log capture truncated at 128 bytes'))
  3380. expect(markers.length).toBeGreaterThanOrEqual(1)
  3381. })
  3382. it('reports an exception whose message holds an unpaired surrogate instead of stranding to the wall clock', async () => {
  3383. // A strict UTF-8 encode of "\ud800" throws while BUILDING the failure
  3384. // frame; the run would then hang to maxWallMs and misreport as timeout.
  3385. const { runtime } = await setup({ maxWallMs: 8_000 })
  3386. const result = await runtime.run({
  3387. program: String.raw`raise Exception("bad \ud800 surrogate")`,
  3388. bindings: [],
  3389. })
  3390. expect(result.error?.kind).toBe('exception')
  3391. expect(result.error?.message).toContain('bad')
  3392. expect(result.error?.message).toContain('surrogate')
  3393. })
  3394. it('carries a lone-surrogate completion string across the wire as its JSON escape', async () => {
  3395. // UTF-8 has no encoding for a lone surrogate, but JSON does: the ASCII
  3396. // `\ud800` escape, which JSON.parse reads back as the same UTF-16 code
  3397. // unit. `CodeJsonValue`, `snapshotJsonValue`, and the worker backend all
  3398. // accept such a string, so this backend must not narrow the shared seam.
  3399. const { runtime } = await setup()
  3400. const result = await runtime.run({
  3401. program: String.raw`return {"lone": "a\ud800b", "spelled": "😀"}`,
  3402. bindings: [],
  3403. })
  3404. expect(result.error).toBeUndefined()
  3405. // The lone half survives as the code unit itself; a spelled-out high-low
  3406. // PAIR folds into the astral character the host would hold for it.
  3407. expect(result.value).toEqual({ lone: 'a\ud800b', spelled: '\u{1f600}' })
  3408. })
  3409. it('meters a lone surrogate at its six escaped bytes, matching the host', async () => {
  3410. // The child and the host share maxValueBytes, so the child must charge the
  3411. // escape's six ASCII bytes (plus two quotes): eight fits, nine does not.
  3412. const { runtime } = await setup({ maxValueBytes: 8 })
  3413. const ok = await runtime.run({ program: String.raw`return "\ud800"`, bindings: [] })
  3414. expect(ok.error).toBeUndefined()
  3415. expect(ok.value).toBe('\ud800')
  3416. const over = await setup({ maxValueBytes: 7 })
  3417. const result = await over.runtime.run({ program: String.raw`return "\ud800"`, bindings: [] })
  3418. expect(result.error?.kind).toBe('output-limit')
  3419. })
  3420. it('passes a lone-surrogate binding argument through instead of failing the call', async () => {
  3421. // The argument validator shared the same over-narrow rejection; a host
  3422. // binding must receive the code unit the program passed.
  3423. const seen: unknown[] = []
  3424. const { runtime } = await setup()
  3425. const result = await runtime.run({
  3426. program: String.raw`return await tools.echo({"text": "x\udfff"})`,
  3427. bindings: tools({ echo: async (args: unknown) => { seen.push(args); return args as CodeJsonValue } }),
  3428. })
  3429. expect(result.error).toBeUndefined()
  3430. expect(seen).toEqual([{ text: 'x\udfff' }])
  3431. expect(result.value).toEqual({ text: 'x\udfff' })
  3432. })
  3433. it('meters a non-ASCII completion in UTF-8 JSON bytes, matching the host', async () => {
  3434. // json.dumps' default \uXXXX escaping would count "é" as 8 bytes while
  3435. // the host meter counts its UTF-8 JSON form (4); the shared budget must
  3436. // agree, so a 4-byte-fitting value passes a maxValueBytes of 4.
  3437. const { runtime } = await setup({ maxValueBytes: 4 })
  3438. const ok = await runtime.run({ program: 'return "é"', bindings: [] })
  3439. expect(ok.error).toBeUndefined()
  3440. expect(ok.value).toBe('é')
  3441. const over = await runtime.run({ program: 'return "éx"', bindings: [] })
  3442. expect(over.error?.kind).toBe('output-limit')
  3443. })
  3444. it('filters bootstrap frames from exception-group members (TaskGroup)', async () => {
  3445. // Python 3.11+ stores member stacks under TracebackException.exceptions;
  3446. // the <model>-frame filter must recurse into them too.
  3447. const { runtime } = await setup()
  3448. const result = await runtime.run({
  3449. program: [
  3450. 'import asyncio, sys',
  3451. 'if sys.version_info < (3, 11):',
  3452. ' raise ValueError("skip-old <model>")',
  3453. 'async def boom():',
  3454. ' raise ValueError("group-member")',
  3455. 'async with asyncio.TaskGroup() as tg:',
  3456. ' tg.create_task(boom())',
  3457. ].join('\n'),
  3458. bindings: [],
  3459. })
  3460. expect(result.error?.kind).toBe('exception')
  3461. expect(result.error?.message).toContain('<model>')
  3462. expect(result.error?.message).not.toContain('bootstrap.py')
  3463. })
  3464. it('keeps frames intact when a model thread floods logs while a large frame drains', async () => {
  3465. // os.write releases the GIL and a frame beyond PIPE_BUF is not atomic:
  3466. // without the writer lock + full-write loop, the printing thread could
  3467. // interleave bytes mid-frame and the host would drop the malformed JSON,
  3468. // hanging the run to the wall clock (or losing the completion).
  3469. const { runtime } = await setup({ maxValueBytes: 1024 * 1024, maxLogBytes: 4 * 1024 * 1024, maxWallMs: 15_000 })
  3470. const result = await runtime.run({
  3471. program: [
  3472. 'import threading',
  3473. 'stop = False',
  3474. 'def spam():',
  3475. ' while not stop:',
  3476. ' print("spam-line-" + "y" * 100)',
  3477. 't = threading.Thread(target=spam)',
  3478. 't.start()',
  3479. // A ~300 KiB completion — several PIPE_BUF units — while spam runs.
  3480. 'big = "x" * (300 * 1024)',
  3481. 'stop = True',
  3482. 't.join()',
  3483. 'return big',
  3484. ].join('\n'),
  3485. bindings: [],
  3486. })
  3487. expect(result.error).toBeUndefined()
  3488. expect(result.value).toBe('x'.repeat(300 * 1024))
  3489. }, 20_000)
  3490. it('settles cleanly while a daemon thread keeps writing unterminated log text', async () => {
  3491. // WARNING regression: the settlement `flush_out()/flush_err()` on the main
  3492. // coroutine read and clear `_LogStream._pending` and the shared LogBuffer
  3493. // ledger with NO lock, while a model daemon thread's `print`/`write` mutate
  3494. // the same state. Capturing the bound method (`out_stream.flush_line`) only
  3495. // fixes WHICH callable runs, not what it reads mid-flight: the flush could
  3496. // interleave with a concurrent write and join a `_pending` list being
  3497. // mutated under it, corrupting the ledger and costing the `done` frame — the
  3498. // run would then strand to the wall clock instead of completing. The shared
  3499. // re-entrant lock serializes them.
  3500. //
  3501. // A pure data race has no single bad input to reject deterministically, so
  3502. // this maximizes overlap: daemon threads emit UNTERMINATED writes (which
  3503. // pile into `_pending` rather than flushing per line) right up to the moment
  3504. // the body returns and settlement flushes. Repeated so the interleave lands.
  3505. for (let attempt = 0; attempt < 5; attempt++) {
  3506. const { runtime, fiber } = await setup({ maxLogBytes: 4 * 1024 * 1024, maxWallMs: 15_000 })
  3507. const result = await runtime.run({
  3508. program: [
  3509. 'import sys, threading',
  3510. 'stop = False',
  3511. 'def spam():',
  3512. ' while not stop:',
  3513. // No newline: the text accumulates in the stream's `_pending`, which is
  3514. // exactly the state the settlement flush also touches.
  3515. ' sys.stdout.write("tail-fragment-" + "z" * 64)',
  3516. 'workers = [threading.Thread(target=spam, daemon=True) for _ in range(4)]',
  3517. 'for t in workers: t.start()',
  3518. // Let the daemons build up pending writes, then return so settlement
  3519. // flushes while they are still mid-write.
  3520. 'import time; time.sleep(0.05)',
  3521. 'return "settled"',
  3522. ].join('\n'),
  3523. bindings: [],
  3524. })
  3525. expect(result.error).toBeUndefined()
  3526. expect(result.value).toBe('settled')
  3527. await fiber.dispose()
  3528. }
  3529. }, 30_000)
  3530. it('completes a binding called from a worker thread on its own event loop', async () => {
  3531. // A binding reply Future is created on the loop that ran `dispatch`. When the
  3532. // model calls a binding from a worker THREAD via `asyncio.run(tools.x(...))`,
  3533. // that Future belongs to the thread's loop, not the main loop where
  3534. // `_pump_replies` reads the reply. `asyncio.Future` is not thread-safe:
  3535. // completing it from another thread does not wake its own loop, so a direct
  3536. // `set_result` would strand the awaiting thread and the run would degrade to a
  3537. // wall-clock timeout. The pump must schedule completion on the Future's own
  3538. // loop via `call_soon_threadsafe`. The tight maxWallMs makes the pre-fix
  3539. // failure a fast timeout rather than a hang.
  3540. //
  3541. // The main coroutine yields with `await asyncio.sleep` while the worker runs,
  3542. // rather than a synchronous `t.join()`: joining would block the main thread,
  3543. // so the main loop could not run `_pump_replies` and the call would deadlock
  3544. // regardless of the fix — that blocks the pump, not the cross-loop delivery
  3545. // this test pins.
  3546. const { runtime } = await setup({ maxWallMs: 8_000 })
  3547. const seen: unknown[] = []
  3548. const result = await runtime.run({
  3549. program: [
  3550. 'import asyncio, threading',
  3551. 'result = {}',
  3552. 'def worker():',
  3553. // A fresh loop in this thread; the binding Future is created here.
  3554. ' result["value"] = asyncio.run(tools.echo({"from": "thread"}))',
  3555. 't = threading.Thread(target=worker)',
  3556. 't.start()',
  3557. 'while t.is_alive():',
  3558. ' await asyncio.sleep(0.02)',
  3559. 'return result["value"]',
  3560. ].join('\n'),
  3561. bindings: tools({
  3562. echo: async (args) => { seen.push(args); return args as CodeJsonValue },
  3563. }),
  3564. })
  3565. expect(result.error).toBeUndefined()
  3566. expect(result.value).toEqual({ from: 'thread' })
  3567. // The host binding actually ran (the reply round-tripped), not a timeout.
  3568. expect(seen).toEqual([{ from: 'thread' }])
  3569. }, 15_000)
  3570. it('keeps the reply pump alive when a late reply targets a closed thread loop', async () => {
  3571. // A binding called from a worker thread that ABANDONS the call (its
  3572. // `asyncio.run` is cancelled) leaves the pending entry holding that thread's
  3573. // loop, which `asyncio.run` closes on return. When the host later answers
  3574. // that call, `_pump_replies` schedules the completion onto the closed loop —
  3575. // `call_soon_threadsafe` raises `RuntimeError('Event loop is closed')`.
  3576. // Unguarded, that RuntimeError ends the pump task and strands every later
  3577. // reply; the guard drops the moot reply and keeps the pump serving.
  3578. //
  3579. // The ordering is a STRUCTURAL guarantee, not a timing window: the worker
  3580. // closes its loop before the main coroutine signals `closed`; the host
  3581. // answers the abandoned `slow` call (hitting the closed loop) before it
  3582. // answers `release`, because `release`'s handler only resolves `slow` first
  3583. // and then yields a microtask. So the pump provably meets the closed loop on
  3584. // `slow`'s reply before it must deliver `release`'s. Fail-before: the pump
  3585. // dies on `slow`, `release`'s reply is never read, and `await tools.release`
  3586. // hangs to the (small) maxWallMs as a timeout.
  3587. let releaseSlow!: () => void
  3588. const slowGate = new Promise<void>((resolve) => { releaseSlow = resolve })
  3589. const { runtime } = await setup({ maxWallMs: 6_000 })
  3590. const result = await runtime.run({
  3591. program: [
  3592. 'import asyncio, threading',
  3593. 'closed = threading.Event()',
  3594. 'def worker():',
  3595. ' async def body():',
  3596. // Abandon the call: wait_for cancels it, but the pending host-side entry
  3597. // survives (dispatch does not pop on cancellation), holding this loop.
  3598. ' try:',
  3599. ' await asyncio.wait_for(tools.slow({}), timeout=0.1)',
  3600. ' except asyncio.TimeoutError:',
  3601. ' pass',
  3602. ' asyncio.run(body())', // closes the thread's loop on return
  3603. ' closed.set()',
  3604. 't = threading.Thread(target=worker)',
  3605. 't.start()',
  3606. 'while not closed.is_set():',
  3607. ' await asyncio.sleep(0.02)',
  3608. // The loop is closed. Now the host answers slow (dead-loop reply) then
  3609. // release; the pump must survive the first to deliver the second.
  3610. 'after = await tools.release({})',
  3611. 'return after',
  3612. ].join('\n'),
  3613. bindings: tools({
  3614. slow: async () => {
  3615. // Answer only once the worker has closed its loop AND the main
  3616. // coroutine is awaiting release, so this reply reaches the pump against
  3617. // the closed loop.
  3618. await slowGate
  3619. return 'late'
  3620. },
  3621. release: async () => {
  3622. // Let slow's reply be written first, then yield a microtask so the
  3623. // pump processes the dead-loop reply before release's own reply lands.
  3624. releaseSlow()
  3625. await new Promise(resolve => setImmediate(resolve))
  3626. return 'released'
  3627. },
  3628. }),
  3629. })
  3630. expect(result.error).toBeUndefined()
  3631. // The pump survived the closed-loop reply and delivered the later binding.
  3632. expect(result.value).toBe('released')
  3633. }, 15_000)
  3634. it('keeps the reply pump alive when RuntimeError is rebound before the program runs', async () => {
  3635. // `_pump_replies` catches a closed-loop scheduling failure with `except
  3636. // _RuntimeError`. If that name were bound as a pump BODY local, it would be
  3637. // captured at pump-start — but `_run` reaches the model's top-level
  3638. // statements (which run before the pump's first step, since there is no
  3639. // suspension point between `create_task` and `await __dsh_main__`) with the
  3640. // rebind already applied, so `_RuntimeError` would capture the REBOUND class
  3641. // and the closed-loop `RuntimeError` would escape, killing the pump. Binding
  3642. // it as a DEF-TIME default argument captures the original before any model
  3643. // code runs. This rebinds `__main__.RuntimeError` as the very first program
  3644. // statement and drives the closed-loop worker pattern: the pump must survive
  3645. // the dead-loop reply and deliver the later binding.
  3646. let releaseSlow!: () => void
  3647. const slowGate = new Promise<void>((resolve) => { releaseSlow = resolve })
  3648. const { runtime } = await setup({ maxWallMs: 6_000 })
  3649. const result = await runtime.run({
  3650. program: [
  3651. 'import __main__',
  3652. '__main__.RuntimeError = ValueError',
  3653. 'import asyncio, threading',
  3654. 'closed = threading.Event()',
  3655. 'def worker():',
  3656. ' async def body():',
  3657. ' try:',
  3658. ' await asyncio.wait_for(tools.slow({}), timeout=0.1)',
  3659. ' except asyncio.TimeoutError:',
  3660. ' pass',
  3661. ' asyncio.run(body())',
  3662. ' closed.set()',
  3663. 't = threading.Thread(target=worker)',
  3664. 't.start()',
  3665. 'while not closed.is_set():',
  3666. ' await asyncio.sleep(0.02)',
  3667. 'after = await tools.release({})',
  3668. 'return after',
  3669. ].join('\n'),
  3670. bindings: tools({
  3671. slow: async () => { await slowGate; return 'late' },
  3672. release: async () => { releaseSlow(); await new Promise(resolve => setImmediate(resolve)); return 'released' },
  3673. }),
  3674. })
  3675. expect(result.error).toBeUndefined()
  3676. expect(result.value).toBe('released')
  3677. }, 15_000)
  3678. it('keeps a successful completion when _done_with_value is rebound', async () => {
  3679. // `_run` calls `_done_with_value(value, max_value_bytes)` after the program
  3680. // returns. The name is a module global, and this bootstrap IS `__main__`, so
  3681. // `__main__._done_with_value = boom` as a program statement would otherwise
  3682. // be resolved at call time and a legitimate success would be rewritten into
  3683. // an `exception`. `_run` now binds `done_with_value_bound = _done_with_value`
  3684. // before the program runs, so the entry name is immune; the run must still
  3685. // report the success value.
  3686. const { runtime } = await setup({ maxWallMs: 10_000 })
  3687. const result = await runtime.run({
  3688. program: [
  3689. 'import __main__',
  3690. 'def boom(*a, **k):',
  3691. ' raise RuntimeError("hijacked")',
  3692. '__main__._done_with_value = boom',
  3693. 'return 1',
  3694. ].join('\n'),
  3695. bindings: [],
  3696. })
  3697. expect(result.error).toBeUndefined()
  3698. expect(result.value).toBe(1)
  3699. }, 15_000)
  3700. it('round-trips an exactly representable large integer through a binding echo', async () => {
  3701. // The reply serializer must print BigInt digits for a beyond-safe
  3702. // integral double: String(2**60) emits a rounded form, and the child
  3703. // would receive a DIFFERENT integer than the binding resolved.
  3704. const { runtime } = await setup()
  3705. const result = await runtime.run({
  3706. program: [
  3707. 'v = await tools.echo(2**60)',
  3708. 'return v == 2**60',
  3709. ].join('\n'),
  3710. bindings: tools({ echo: async args => args as never }),
  3711. })
  3712. expect(result.error).toBeUndefined()
  3713. expect(result.value).toBe(true)
  3714. })
  3715. it('preserves an exactly representable large integer and rejects a rounding one', async () => {
  3716. // The canonical boundary accepts every JS-double-exact value: 2**53 and
  3717. // 2**60 round-trip exactly and must cross (matching the worker backend);
  3718. // 2**53+1 rounds and must fail as invalid-output.
  3719. const { runtime } = await setup()
  3720. const exact = await runtime.run({ program: 'return [2**53, 2**60]', bindings: [] })
  3721. expect(exact.error).toBeUndefined()
  3722. expect(exact.value).toEqual([2 ** 53, 2 ** 60])
  3723. const lossy = await runtime.run({ program: 'return 2**53 + 1', bindings: [] })
  3724. expect(lossy.error?.kind).toBe('invalid-output')
  3725. expect(lossy.error?.message).toContain('not exactly representable')
  3726. })
  3727. it('rejects a container subclass whose overridden methods hide its contents', async () => {
  3728. // A dict subclass returning [] from items() passes an isinstance check but
  3729. // serializes as {}, so the host would receive a value the program did not
  3730. // compute. Exact-type matching fails it as invalid-output instead. The
  3731. // worker backend rejects the prototype-equivalent shapes the same way.
  3732. const { runtime } = await setup()
  3733. const hidden = await runtime.run({
  3734. program: [
  3735. 'class Sneaky(dict):',
  3736. ' def items(self): return []',
  3737. ' def keys(self): return []',
  3738. ' def __iter__(self): return iter([])',
  3739. ' def __len__(self): return 0',
  3740. 'return Sneaky(secret="kept")',
  3741. ].join('\n'),
  3742. bindings: [],
  3743. })
  3744. expect(hidden.error?.kind).toBe('invalid-output')
  3745. expect(hidden.error?.message).toContain('unsupported type (Sneaky)')
  3746. // A list subclass is refused on the same rule.
  3747. const listish = await runtime.run({
  3748. program: ['class L(list):', ' def __iter__(self): return iter([])', 'return L([1, 2, 3])'].join('\n'),
  3749. bindings: [],
  3750. })
  3751. expect(listish.error?.kind).toBe('invalid-output')
  3752. expect(listish.error?.message).toContain('unsupported type (L)')
  3753. // The exact built-in containers still cross unchanged.
  3754. const plain = await runtime.run({ program: 'return {"secret": [1, 2]}', bindings: [] })
  3755. expect(plain.error).toBeUndefined()
  3756. expect(plain.value).toEqual({ secret: [1, 2] })
  3757. })
  3758. it('rejects a scalar subclass whose overrides disagree with what gets serialized', async () => {
  3759. // The validators checked scalars with isinstance, so a subclass passed
  3760. // every check by its real value while the ENCODER read an override — the
  3761. // host then received a value the walk never approved. Each case below is a
  3762. // distinct override reaching a distinct reader.
  3763. const { runtime } = await setup()
  3764. // _dump_float spells a float from repr(value), so an overridden __repr__
  3765. // decides the digits: F(2.5) serialized as 1.
  3766. const floated = await runtime.run({
  3767. program: [
  3768. 'class F(float):',
  3769. ' def __repr__(self): return "1.0"',
  3770. 'return F(2.5)',
  3771. ].join('\n'),
  3772. bindings: [],
  3773. })
  3774. expect(floated.error?.kind).toBe('invalid-output')
  3775. expect(floated.error?.message).toContain('unsupported type (F)')
  3776. // The JS-safe-range bound is two comparisons, so overriding them admits an
  3777. // int whose true digits (json.dumps reads the C-level value) the host's
  3778. // JSON.parse rounds: 9007199254740993 arrives as ...992.
  3779. const inted = await runtime.run({
  3780. program: [
  3781. 'class I(int):',
  3782. ' def __gt__(self, other): return False',
  3783. ' def __lt__(self, other): return False',
  3784. 'return I(2 ** 53 + 1)',
  3785. ].join('\n'),
  3786. bindings: [],
  3787. })
  3788. expect(inted.error?.kind).toBe('invalid-output')
  3789. expect(inted.error?.message).toContain('unsupported type (I)')
  3790. // The pre-encode size bound reads len(), so overriding it to 0 admits a
  3791. // string of any length past maxValueBytes.
  3792. const stringed = await runtime.run({
  3793. program: [
  3794. 'class S(str):',
  3795. ' def __len__(self): return 0',
  3796. 'return S("Q" * 100000)',
  3797. ].join('\n'),
  3798. bindings: [],
  3799. })
  3800. expect(stringed.error?.kind).toBe('invalid-output')
  3801. expect(stringed.error?.message).toContain('unsupported type (S)')
  3802. // A str-subclass dict KEY reaches the same len() bound.
  3803. const keyed = await runtime.run({
  3804. program: [
  3805. 'class S(str):',
  3806. ' def __len__(self): return 0',
  3807. 'return {S("Q" * 100000): 1}',
  3808. ].join('\n'),
  3809. bindings: [],
  3810. })
  3811. expect(keyed.error?.kind).toBe('invalid-output')
  3812. expect(keyed.error?.message).toContain('non-string dict key (S)')
  3813. // bool is an int subclass that IS lossless JSON, and the exact scalars all
  3814. // still cross unchanged.
  3815. const plain = await runtime.run({
  3816. program: 'return {"t": True, "f": False, "n": None, "i": 7, "d": 2.5, "s": "ok"}',
  3817. bindings: [],
  3818. })
  3819. expect(plain.error).toBeUndefined()
  3820. expect(plain.value).toEqual({ t: true, f: false, n: null, i: 7, d: 2.5, s: 'ok' })
  3821. })
  3822. it('rejects a scalar subclass passed as a binding argument', async () => {
  3823. // The uncapped binding-argument validator shares the exact-type rule, so
  3824. // the call fails through its rejection contract instead of dispatching a
  3825. // float whose digits come from an override.
  3826. const { runtime } = await setup()
  3827. const seen: CodeJsonValue[] = []
  3828. const result = await runtime.run({
  3829. program: [
  3830. 'class F(float):',
  3831. ' def __repr__(self): return "1.0"',
  3832. 'try:',
  3833. ' await tools.echo({"v": F(2.5)})',
  3834. 'except Exception as exc:',
  3835. ' return str(exc)',
  3836. ].join('\n'),
  3837. bindings: tools({ echo: async (args) => {
  3838. seen.push(args as CodeJsonValue)
  3839. return null
  3840. } }),
  3841. })
  3842. expect(result.error).toBeUndefined()
  3843. expect(result.value).toContain('unsupported type (F)')
  3844. expect(seen).toEqual([])
  3845. })
  3846. it('rejects a container subclass passed as a binding argument', async () => {
  3847. // Binding arguments run the uncapped validator, which must apply the same
  3848. // exact-type rule: the call fails descriptively instead of dispatching a
  3849. // value whose serialization disagrees with what was validated.
  3850. const { runtime } = await setup()
  3851. const seen: CodeJsonValue[] = []
  3852. const result = await runtime.run({
  3853. program: [
  3854. 'class Sneaky(dict):',
  3855. ' def items(self): return []',
  3856. 'try:',
  3857. ' await tools.echo(Sneaky(secret="kept"))',
  3858. 'except Exception as exc:',
  3859. ' return str(exc)',
  3860. ].join('\n'),
  3861. bindings: tools({ echo: async (args) => {
  3862. seen.push(args as CodeJsonValue)
  3863. return null
  3864. } }),
  3865. })
  3866. expect(result.error).toBeUndefined()
  3867. expect(result.value).toContain('unsupported type (Sneaky)')
  3868. expect(seen).toEqual([])
  3869. })
  3870. it('fails an oversized completion as output-limit without materializing its encoding', async () => {
  3871. // A 100 MiB string under maxValueBytes: 1024 must fail as output-limit.
  3872. // The address-space cap leaves room for the program to BUILD the string
  3873. // (one copy + interpreter) but not for the old full pre-check encode,
  3874. // which materialized chunk fragments plus the joined copy (~2 more
  3875. // copies) and died on RLIMIT_AS as MemoryError/worker-exit.
  3876. const { runtime } = await setup({ maxValueBytes: 1024, addressSpaceMb: 384, maxWallMs: 15_000 })
  3877. const result = await runtime.run({
  3878. program: 'return "x" * (100 * 1024 * 1024)',
  3879. bindings: [],
  3880. })
  3881. expect(result.error?.kind).toBe('output-limit')
  3882. expect(result.error?.message).toContain('exceeded 1024 bytes')
  3883. }, 20_000)
  3884. it('rejects a control-heavy oversized completion on its length, not its escaped copy', async () => {
  3885. // Every "\x00" escapes to the six bytes "\u0000", so the escaped form of a
  3886. // 40 MB string is ~240 MB. The walk must refuse on the cheap
  3887. // `len(current) + 2` lower bound; the 384 MiB address space holds the raw
  3888. // string but not its escaped expansion, so a pre-escape check dies on
  3889. // RLIMIT_AS instead of returning output-limit.
  3890. const { runtime } = await setup({ maxValueBytes: 1024, addressSpaceMb: 384, maxWallMs: 15_000 })
  3891. const result = await runtime.run({
  3892. program: 'return "\\x00" * (40 * 1024 * 1024)',
  3893. bindings: [],
  3894. })
  3895. expect(result.error?.kind).toBe('output-limit')
  3896. expect(result.error?.message).toContain('exceeded 1024 bytes')
  3897. }, 20_000)
  3898. it('truncates a single print far above maxLogBytes instead of dying on the encode', async () => {
  3899. // LogBuffer must reject via the cheap char-count lower bound BEFORE
  3900. // UTF-8-encoding the whole string: the full encode of a ~100 MB line
  3901. // would double the allocation and can breach RLIMIT_AS. 256 MiB
  3902. // address space comfortably holds one copy of the 100 MB string but
  3903. // not the pre-fix double allocation plus interpreter overhead spikes.
  3904. const { runtime } = await setup({ maxLogBytes: 1024, addressSpaceMb: 256, maxWallMs: 15_000 })
  3905. const result = await runtime.run({
  3906. program: [
  3907. 'print("x" * (100 * 1024 * 1024))',
  3908. 'return "done"',
  3909. ].join('\n'),
  3910. bindings: [],
  3911. })
  3912. expect(result.error).toBeUndefined()
  3913. expect(result.value).toBe('done')
  3914. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  3915. }, 20_000)
  3916. it('stops host capture at the child ledger truncation, keeping exactly one marker', async () => {
  3917. // The two ledgers exhaust independently. One child entry larger than
  3918. // `maxLogBytes` sends ONLY the marker, so the host budget is still nearly
  3919. // untouched — and the marker used to arrive as an ordinary `log` frame the
  3920. // host could not tell from program output. Text written afterwards was
  3921. // therefore retained AFTER the marker, contradicting the stop-after-
  3922. // truncation contract, and a later host-side exhaustion could append a
  3923. // second marker. The frame now carries `truncated: true`.
  3924. //
  3925. // `os.write(1, ...)` bypasses the child's own stream, so those bytes reach
  3926. // the host as stray stdout and take the host ledger path rather than the
  3927. // child's — which is exactly the route that leaked past the marker.
  3928. const { runtime } = await setup({ maxLogBytes: 64, maxWallMs: 10_000 })
  3929. const result = await runtime.run({
  3930. program: [
  3931. 'import os',
  3932. 'print("y" * 70000)',
  3933. 'os.write(1, b"AFTER")',
  3934. 'return "done"',
  3935. ].join('\n'),
  3936. bindings: [],
  3937. })
  3938. expect(result.error).toBeUndefined()
  3939. expect(result.value).toBe('done')
  3940. const markers = result.logs.filter(line => line.includes('log capture truncated'))
  3941. expect(markers).toHaveLength(1)
  3942. // The marker is the LAST entry: nothing was retained after truncation.
  3943. expect(result.logs.at(-1)).toBe(markers[0])
  3944. expect(result.logs.join('\n')).not.toContain('AFTER')
  3945. }, 20_000)
  3946. it('keeps one marker when a program forges repeated truncation frames', async () => {
  3947. // `truncated` is attacker-reachable: the program owns fd 3 and can write the
  3948. // flag itself, so the field is a hostile input rather than a trusted signal.
  3949. // Repeats must collapse to the single marker the contract promises, and only
  3950. // the literal `true` counts — a forged `"yes"` is rebuilt away by
  3951. // validateChildFrame, so that frame stays ordinary text.
  3952. const { runtime } = await setup({ maxLogBytes: 4096, maxWallMs: 10_000 })
  3953. const result = await runtime.run({
  3954. program: [
  3955. 'import os, json',
  3956. 'os.write(3, json.dumps({"type":"log","text":"first","truncated":"yes"}).encode() + b"\\n")',
  3957. 'os.write(3, json.dumps({"type":"log","text":"MARK-A","truncated":True}).encode() + b"\\n")',
  3958. 'os.write(3, json.dumps({"type":"log","text":"MARK-B","truncated":True}).encode() + b"\\n")',
  3959. 'return "done"',
  3960. ].join('\n'),
  3961. bindings: [],
  3962. })
  3963. expect(result.error).toBeUndefined()
  3964. expect(result.value).toBe('done')
  3965. // The non-boolean flag did not truncate, so its text was captured normally.
  3966. expect(result.logs).toContain('first')
  3967. // The first genuine flag stopped capture and emitted the HOST's own marker;
  3968. // the frame's own text is discarded, so neither payload appears.
  3969. expect(result.logs).not.toContain('MARK-A')
  3970. expect(result.logs).not.toContain('MARK-B')
  3971. expect(result.logs.at(-1)).toBe(logTruncationMarker(4096))
  3972. expect(result.logs.filter(line => line.includes('log capture truncated'))).toHaveLength(1)
  3973. }, 20_000)
  3974. it('discards the text of a forged truncation frame instead of retaining it', async () => {
  3975. // The marker branch bypasses `admit`, so retaining the frame's own text put
  3976. // attacker-controlled bytes into `logs` with no cap at all: measured, a 1 MiB
  3977. // forged text was retained whole under `maxLogBytes: 64`, and the only bound
  3978. // left was the 64 MiB frame parse cap. The host emits its own marker instead,
  3979. // so the retained size is fixed regardless of what the program sent.
  3980. const forgedBytes = 1024 * 1024
  3981. const { runtime } = await setup({ maxLogBytes: 64, maxWallMs: 20_000 })
  3982. const result = await runtime.run({
  3983. program: [
  3984. 'import os, json',
  3985. `big = "A" * ${forgedBytes}`,
  3986. 'os.write(3, json.dumps({"type":"log","truncated":True,"text":big}).encode() + b"\\n")',
  3987. 'return "done"',
  3988. ].join('\n'),
  3989. bindings: [],
  3990. })
  3991. expect(result.error).toBeUndefined()
  3992. expect(result.value).toBe('done')
  3993. // Only the host marker is kept, so the total stays orders of magnitude below
  3994. // what the forgery carried — and below the cap it was trying to escape.
  3995. expect(result.logs).toEqual([logTruncationMarker(64)])
  3996. expect(result.logs.join('').length).toBeLessThan(forgedBytes / 1000)
  3997. }, 30_000)
  3998. it('coalesces unframed fd-3 fragments without recopying the sealed prefix', async () => {
  3999. // The frame ceiling meters payload BYTES, but each retained chunk is its own
  4000. // Buffer with object and backing-store overhead the byte count cannot see:
  4001. // 5000 single-byte newline-free writes produced 5000 chunks holding 5031
  4002. // bytes, so a program pacing such writes could accumulate millions of objects
  4003. // inside the wall budget and exhaust the host heap far below 256 MiB.
  4004. //
  4005. // The observable behavior is that the run still completes normally: the
  4006. // fragments are coalesced rather than rejected, since a slow trickle of bytes
  4007. // is not itself a protocol violation.
  4008. //
  4009. // `Buffer.concat` is wrapped for the duration so the cumulative copy volume
  4010. // is measured rather than inferred: that total is what separates sealing into
  4011. // blocks from re-merging the whole buffer, and both shapes pass every
  4012. // behavioral assertion below.
  4013. //
  4014. // The trickle is terminated with its own newline before the real frame is
  4015. // written. Without that, those 5000 bytes prefix the frame on the SAME line,
  4016. // which then parses as junk and is dropped — correct framing behavior, but it
  4017. // would leave this test asserting the wrong thing.
  4018. // Bound at capture: `Buffer.concat` is a static method, and taking a bare
  4019. // reference to one trips no-unbound-method.
  4020. const realConcat = Buffer.concat.bind(Buffer)
  4021. let copied = 0
  4022. Buffer.concat = (list: readonly Uint8Array[], total?: number): Buffer<ArrayBuffer> => {
  4023. for (const part of list) copied += part.length
  4024. return realConcat(list, total)
  4025. }
  4026. const program = [
  4027. 'import os',
  4028. // Newline-free single-byte writes, spaced so each lands as its own read.
  4029. // 60000 rather than 5000: the trickle has to cross the seal threshold
  4030. // enough times for the two shapes to separate. At 5000 writes there are
  4031. // only four seals, so even the quadratic form copies well under a
  4032. // megabyte and the budget below could not tell them apart.
  4033. 'for _ in range(60000):',
  4034. ' os.write(3, b"x")',
  4035. ' os.sched_yield()',
  4036. 'os.write(3, b"\\n")',
  4037. // A real frame after the trickle proves framing still works on the
  4038. // coalesced residual.
  4039. 'print("after-trickle")',
  4040. 'return "done"',
  4041. ].join('\n')
  4042. let result: CodeRunResult
  4043. try {
  4044. const { runtime } = await setup({ maxWallMs: 30_000 })
  4045. result = await runtime.run({ program, bindings: [] })
  4046. } finally {
  4047. Buffer.concat = realConcat
  4048. }
  4049. expect(result.error).toBeUndefined()
  4050. expect(result.value).toBe('done')
  4051. expect(result.logs).toContain('after-trickle')
  4052. // Sealing appends a finished block rather than re-merging everything held, so
  4053. // each byte is copied once. Re-concatenating the whole buffer at every
  4054. // threshold made the cumulative copy volume quadratic — 10 MiB trickled a
  4055. // byte at a time copies 53.7 GB that way. A per-byte-copied budget is the
  4056. // discriminator, and it is measured rather than reasoned about: this shape
  4057. // copies about 119 KB for 60000 trickled bytes, the re-merging shape about
  4058. // 540 KB. 256 KiB sits between them with margin on both sides — most writes
  4059. // are coalesced by the pipe before they reach us, so the observed ratio is
  4060. // smaller than the asymptotic one, and the threshold has to sit where a real
  4061. // measurement lands rather than where the asymptote suggests.
  4062. expect(copied).toBeLessThan(256 * 1024)
  4063. }, 40_000)
  4064. it('seals trickled stray fragments into blocks without recopying the sealed prefix', async () => {
  4065. // The stray-capture buffer has the same object-overhead exposure as the fd-3
  4066. // reader above: each newline-free `data` chunk is its own Buffer, so a
  4067. // program pacing single-byte `os.write(1, ...)` accumulates one object per
  4068. // write, which the serialized-cost counter cannot see. Past MAX_PENDING_CHUNKS
  4069. // the fragments seal into a finished block; re-merging the whole residual at
  4070. // each threshold instead would copy the sealed prefix again and again, making
  4071. // the cumulative copy volume quadratic. `Buffer.concat` is wrapped to measure
  4072. // that volume — both shapes admit the same final log entry, so the copy total
  4073. // is the discriminator. maxLogBytes is raised so the trickle is retained,
  4074. // not truncated, which is what forces the fragments to accumulate and seal.
  4075. const realConcat = Buffer.concat.bind(Buffer)
  4076. let copied = 0
  4077. Buffer.concat = (list: readonly Uint8Array[], total?: number): Buffer<ArrayBuffer> => {
  4078. for (const part of list) copied += part.length
  4079. return realConcat(list, total)
  4080. }
  4081. let result: CodeRunResult
  4082. try {
  4083. const { runtime } = await setup({ maxLogBytes: 200_000, maxWallMs: 30_000 })
  4084. result = await runtime.run({
  4085. program: [
  4086. 'import os',
  4087. 'for _ in range(60000):',
  4088. ' os.write(1, b"x")',
  4089. ' os.sched_yield()',
  4090. 'os.write(1, b"\\n")',
  4091. 'return "done"',
  4092. ].join('\n'),
  4093. bindings: [],
  4094. })
  4095. } finally {
  4096. Buffer.concat = realConcat
  4097. }
  4098. expect(result.error).toBeUndefined()
  4099. expect(result.value).toBe('done')
  4100. // The trickle coalesces into one log line (no interior newlines). Its exact
  4101. // length depends on pipe coalescing, but it is one entry and non-empty.
  4102. expect(result.logs.length).toBe(1)
  4103. expect((result.logs[0] as string).length).toBeGreaterThan(0)
  4104. // Sealing appends a finished block rather than re-merging everything held, so
  4105. // each byte is copied a bounded number of times. Re-merging the whole
  4106. // residual at every seal threshold instead makes the cumulative copy volume
  4107. // quadratic. Measured like the fd-3 sibling above rather than reasoned about:
  4108. // this sealed shape copies about 120 KB for 60000 trickled bytes, the
  4109. // re-merging shape about 538 KB (the stray path adds one whole-residual
  4110. // concat at the terminating newline over the fd-3 sibling's 119/540, landing
  4111. // at the same order). 256 KiB sits between them with margin on both sides, so
  4112. // reverting the seal to a re-merge turns this assertion red.
  4113. expect(copied).toBeLessThan(256 * 1024)
  4114. }, 40_000)
  4115. it('caps a huge exception diagnostic child-side before it crosses the wire', async () => {
  4116. // A program can raise with a multi-megabyte message; the child must cap
  4117. // it at maxValueBytes before formatting/sending, not ship the whole
  4118. // payload for the host to truncate after parsing.
  4119. const { runtime } = await setup({ maxValueBytes: 1024 })
  4120. const result = await runtime.run({
  4121. program: 'raise ValueError("boom-" + "x" * (8 * 1024 * 1024))',
  4122. bindings: [],
  4123. })
  4124. expect(result.error?.kind).toBe('exception')
  4125. expect(result.error?.message).toContain('boom-')
  4126. expect(result.error?.message.endsWith('… [truncated]')).toBe(true)
  4127. expect(Buffer.byteLength(result.error?.message ?? '', 'utf8')).toBeLessThan(2048)
  4128. })
  4129. it('caps a control-heavy exception diagnostic by its serialized cost, not raw bytes', async () => {
  4130. // The diagnostic crosses fd 3 inside a JSON frame where a control character
  4131. // escapes sixfold (a NUL is one raw byte, six as `\u0000`). Capping by raw
  4132. // UTF-8 length would let a NUL-heavy message near maxValueBytes serialize to
  4133. // ~6x that and breach the frame ceiling — the silent worker-exit inversion
  4134. // the load-time cap check exists to prevent. The child meters the diagnostic
  4135. // by its serialized cost, so a NUL flood is truncated to fit the frame and
  4136. // the run still reports the exception rather than a worker-exit.
  4137. const { runtime } = await setup({ maxValueBytes: 4096 })
  4138. const result = await runtime.run({
  4139. // 512 KiB of NUL: ~3 MiB once escaped, far past the 4 KiB cap.
  4140. program: 'raise ValueError("\\x00" * (512 * 1024))',
  4141. bindings: [],
  4142. })
  4143. expect(result.error?.kind).toBe('exception')
  4144. expect(result.error?.message.endsWith('… [truncated]')).toBe(true)
  4145. // The SERIALIZED form (what the frame carried) fits the budget, so its raw
  4146. // length is well under it too — a raw-byte cap would have admitted ~4 KiB of
  4147. // NULs that serialize to ~24 KiB.
  4148. const serialized = JSON.stringify(result.error?.message ?? '')
  4149. expect(Buffer.byteLength(serialized, 'utf8')).toBeLessThanOrEqual(4096 + 8)
  4150. })
  4151. it('bounds a newline-free partial-line flood while the program is still running', async () => {
  4152. // print("x", end="") never completes a line, so nothing reaches the
  4153. // Python LogBuffer until settlement — the buffered tail must still hit
  4154. // the budget mid-run instead of growing without bound to RLIMIT/timeout.
  4155. const { runtime } = await setup({ maxLogBytes: 1024, maxWallMs: 15_000 })
  4156. const result = await runtime.run({
  4157. program: [
  4158. 'for _ in range(100000):',
  4159. ' print("xxxxxxxxxx", end="")',
  4160. 'return "done"',
  4161. ].join('\n'),
  4162. bindings: [],
  4163. })
  4164. expect(result.error).toBeUndefined()
  4165. expect(result.value).toBe('done')
  4166. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  4167. // The retained text is bounded by the budget, not the 1 MB the program wrote.
  4168. expect(result.logs.join('\n').length).toBeLessThan(4096)
  4169. }, 20_000)
  4170. it('discards empty writes instead of buffering one list slot each', async () => {
  4171. // An empty chunk adds no character, so the mid-run budget check (which
  4172. // compares buffered CHARS against the remaining ledger) can never fire on
  4173. // it. Buffering empty strings therefore grew `_pending` without bound —
  4174. // millions of slots per CPU second — until RLIMIT_AS turned an append into
  4175. // a MemoryError, long after the log ledger was exhausted. Two million
  4176. // empty writes must instead settle normally and contribute NO log entry,
  4177. // proving the chunk was dropped rather than joined at flush_line.
  4178. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 256, maxWallMs: 20_000 })
  4179. const result = await runtime.run({
  4180. program: [
  4181. 'import sys',
  4182. 'for _ in range(2000000):',
  4183. ' sys.stdout.write("")',
  4184. 'return "done"',
  4185. ].join('\n'),
  4186. bindings: [],
  4187. })
  4188. expect(result.error).toBeUndefined()
  4189. expect(result.value).toBe('done')
  4190. expect(result.logs).toEqual([])
  4191. }, 30_000)
  4192. it('stops scanning a single-write newline flood once the log ledger truncates', async () => {
  4193. // One write carrying half a million newlines: the offset scan must exit the
  4194. // instant LogBuffer truncates rather than re-slicing and pushing every
  4195. // remaining line. If it kept scanning it would exhaust the CPU/wall budget;
  4196. // the run instead settles quickly with exactly one truncation marker.
  4197. const { runtime } = await setup({ maxLogBytes: 256, maxWallMs: 10_000 })
  4198. const start = Date.now()
  4199. const result = await runtime.run({
  4200. program: ['print("x\\n" * 500000, end="")', 'return "done"'].join('\n'),
  4201. bindings: [],
  4202. })
  4203. expect(result.error).toBeUndefined()
  4204. expect(result.value).toBe('done')
  4205. expect(result.logs.filter(line => line.includes('log capture truncated'))).toHaveLength(1)
  4206. expect(Date.now() - start).toBeLessThan(8_000)
  4207. }, 15_000)
  4208. it('bounds an oversized newline-terminated write before joining and slicing it', async () => {
  4209. // The newline branch slices the first line out of the write before
  4210. // `LogBuffer.push` can apply its cheap budget rejection, so a single
  4211. // over-budget write cost a full extra copy of itself in peak address space —
  4212. // the amplification that bound exists to avoid, applied one layer too late.
  4213. // Measured under a 400 MiB addressSpaceMb with the slice unbounded: writes
  4214. // of 200 MiB and up died on MemoryError inside `sys.stdout.write`, reported
  4215. // as the PROGRAM's own exception rather than the promised truncation marker.
  4216. // `"\\n".rjust(n, "A")` is a single allocation ending in the newline, so the
  4217. // payload itself fits and the only remaining allocation is the stream's own
  4218. // slice; 340 MiB of a 400 MiB cap cannot survive one more copy of it.
  4219. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 400, maxWallMs: 30_000 })
  4220. const result = await runtime.run({
  4221. program: [
  4222. 'import sys',
  4223. 'payload = "\\n".rjust(340 * 1024 * 1024, "A")',
  4224. 'sys.stdout.write(payload)',
  4225. 'return "done"',
  4226. ].join('\n'),
  4227. bindings: [],
  4228. })
  4229. expect(result.error).toBeUndefined()
  4230. expect(result.value).toBe('done')
  4231. expect(result.logs).toEqual([logTruncationMarker(256)])
  4232. }, 40_000)
  4233. it('bounds a newline-free write against the already-buffered chunks before joining them', async () => {
  4234. // The newline-free arm buffers the write and then compared the buffered
  4235. // CHARACTER COUNT against the ledger — correct — but paid for the comparison
  4236. // with `"".join(self._pending)`, a second full copy of everything held. One
  4237. // buffered character is enough to make that join a copy of the whole
  4238. // following write. Measured under a 400 MiB addressSpaceMb with a 340 MiB
  4239. // second write: the join raised MemoryError inside `sys.stdout.write`, and
  4240. // because the oversized chunks stayed in `_pending` the settlement
  4241. // `flush_line` raised it again — that throw sits after the `except
  4242. // BaseException` block, so it costs the `done` frame and the run came back
  4243. // `timeout: wall-clock ceiling reached (30000ms)` with no logs at all. The
  4244. // bound must be applied BEFORE the join and the chunks dropped on that path,
  4245. // so the run settles with the truncation marker it promises.
  4246. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 400, maxWallMs: 30_000 })
  4247. const result = await runtime.run({
  4248. program: [
  4249. 'import sys',
  4250. // One unterminated character first, so `_pending` is non-empty and the
  4251. // large write cannot take the "buffered text IS the write" shortcut.
  4252. 'sys.stdout.write("x")',
  4253. 'payload = "A" * (340 * 1024 * 1024)',
  4254. 'sys.stdout.write(payload)',
  4255. 'return "done"',
  4256. ].join('\n'),
  4257. bindings: [],
  4258. })
  4259. expect(result.error).toBeUndefined()
  4260. expect(result.value).toBe('done')
  4261. expect(result.logs).toEqual([logTruncationMarker(256)])
  4262. }, 40_000)
  4263. it('bounds a newline-terminated write against the already-buffered chunks before joining them', async () => {
  4264. // Same allocation, reached through the newline arm: with chunks pending, the
  4265. // whole write used to be appended and joined so the offset scan could run
  4266. // over one string. Only the FIRST line needs those chunks, so a pending
  4267. // chunk plus a 340 MiB newline-terminated write under a 400 MiB
  4268. // addressSpaceMb died on MemoryError in the join before the per-line bound
  4269. // could reject anything, and the retained chunks made the settlement flush
  4270. // die the same way: measured, `timeout: wall-clock ceiling reached
  4271. // (30000ms)`. The reconstructed first line is now checked against the ledger
  4272. // and only a budget-sized prefix of it is copied; the rest of the write is
  4273. // scanned in place.
  4274. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 400, maxWallMs: 30_000 })
  4275. const result = await runtime.run({
  4276. program: [
  4277. 'import sys',
  4278. 'sys.stdout.write("x")',
  4279. 'payload = "\\n".rjust(340 * 1024 * 1024, "A")',
  4280. 'sys.stdout.write(payload)',
  4281. 'return "done"',
  4282. ].join('\n'),
  4283. bindings: [],
  4284. })
  4285. expect(result.error).toBeUndefined()
  4286. expect(result.value).toBe('done')
  4287. expect(result.logs).toEqual([logTruncationMarker(256)])
  4288. }, 40_000)
  4289. it('emits pending text on an explicit flush, before the run can be killed', async () => {
  4290. // `_LogStream` inherits TextIOBase's no-op `flush()`, so an explicit
  4291. // `print(..., flush=True)` or `sys.stdout.flush()` left the text in
  4292. // `_pending` with nothing to drain it but `flush_line` after settlement — a
  4293. // call a hanging or killed run never reaches. Measured: printing
  4294. // "before hang" with flush=True ahead of an infinite loop returned
  4295. // `logs: []`, losing the one diagnostic the program deliberately committed.
  4296. const { runtime } = await setup({ maxWallMs: 4_000 })
  4297. const result = await runtime.run({
  4298. program: [
  4299. 'import sys',
  4300. 'print("before hang", end="", flush=True)',
  4301. 'while True: pass',
  4302. ].join('\n'),
  4303. bindings: [],
  4304. })
  4305. expect(result.error?.kind).toBe('timeout')
  4306. expect(result.logs).toContain('before hang')
  4307. }, 15_000)
  4308. it('marks a dropped tail when the ledger lands on exactly zero remaining', async () => {
  4309. // One 100-character line costs 103 serialized bytes (quotes + separator),
  4310. // consuming a 104-byte budget minus the 1-byte array-envelope reservation
  4311. // (104 - 1 = 103) EXACTLY. Landing on zero never trips
  4312. // LogBuffer's "cost > remaining" branch, so `_truncated` stays unset and the
  4313. // stream's own `remaining > 0` guard silently discarded the unscanned tail —
  4314. // the run reported a complete log while dropping text. The tail must be
  4315. // pushed so the marker is emitted. (This surfaced only after empty writes
  4316. // stopped being buffered: `print` issues a trailing `write("")` whose
  4317. // buffered-empty path used to force the marker out incidentally.) A single
  4318. // wide line is used rather than many narrow ones so the CHILD ledger is the
  4319. // one that lands on zero: the host's identical ledger truncates first when
  4320. // many small entries precede the long marker text. `["y"*100]` serializes to
  4321. // exactly 104 bytes (103 payload + 1 envelope), so 104 is the smallest
  4322. // budget that admits the entry.
  4323. const { runtime } = await setup({ maxLogBytes: 104, maxWallMs: 10_000 })
  4324. const result = await runtime.run({
  4325. program: ['print("y" * 100 + "\\n" + "z" * 10, end="")', 'return "done"'].join('\n'),
  4326. bindings: [],
  4327. })
  4328. expect(result.error).toBeUndefined()
  4329. expect(result.value).toBe('done')
  4330. expect(result.logs).toContain('y'.repeat(100))
  4331. expect(result.logs.filter(line => line.includes('log capture truncated'))).toHaveLength(1)
  4332. // The dropped tail is not retained, but its loss is now reported.
  4333. expect(result.logs.some(line => line.includes('z'))).toBe(false)
  4334. }, 15_000)
  4335. it('keeps an admitted log within the serialized array envelope at the exact limit', async () => {
  4336. // Each entry is charged its JSON-string cost plus one separator byte, and
  4337. // the serialized outer logs array adds one more byte of envelope (two
  4338. // brackets and n-1 commas). The ledgers reserve that byte, so a result that
  4339. // exactly exhausts the ledger still serializes within the configured cap.
  4340. // At the 64-byte floor (the smallest admissible maxLogBytes): ledger 63,
  4341. // a 60-character line serializes as `"aaa...a"` (62 bytes) + 1 separator
  4342. // = 63, exactly exhausting the ledger and serializing as `["aaa...a"]`
  4343. // = 64 = the cap; a 61-character line costs 64 > 63 and truncates. The
  4344. // marker rides envelope, so the serialized logs run to cap + marker.
  4345. const { runtime } = await setup({ maxLogBytes: 64, maxWallMs: 10_000 })
  4346. const result = await runtime.run({
  4347. program: ['print("a" * 60 + "\\n" + "b" * 61, end="")', 'return "done"'].join('\n'),
  4348. bindings: [],
  4349. })
  4350. expect(result.error).toBeUndefined()
  4351. expect(result.value).toBe('done')
  4352. // The 60-character line was admitted; the 61-character line was not (a
  4353. // single 'b' would also match the marker's "bytes", so check for the line).
  4354. expect(result.logs).toContain('a'.repeat(60))
  4355. expect(result.logs.some(line => line.includes('b'.repeat(61)))).toBe(false)
  4356. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  4357. }, 15_000)
  4358. it('rejects a log budget too small to serialize the truncation marker', async () => {
  4359. // A maxLogBytes below 64 cannot serialize the truncation marker itself;
  4360. // it is rejected at construction so a marker-only truncated run cannot
  4361. // report more than the public cap. maxValueBytes keeps no floor beyond the
  4362. // positive-integer requirement (a completion can be 1 byte).
  4363. await expect(setup({ maxLogBytes: 63, maxWallMs: 10_000 })).rejects.toThrow(/must be at least 64/)
  4364. }, 15_000)
  4365. it('charges the JSON-escaped cost of control characters against the log ledger', async () => {
  4366. // A NUL renders as \u0000 (6 bytes) in the serialized outer logs; the
  4367. // ledger must charge that expansion, or a control-character flood admits
  4368. // 6x the configured cap.
  4369. const { runtime } = await setup({ maxLogBytes: 256 })
  4370. const result = await runtime.run({
  4371. program: [
  4372. 'for _ in range(500):',
  4373. ' print("\\x00" * 10)',
  4374. 'return "done"',
  4375. ].join('\n'),
  4376. bindings: [],
  4377. })
  4378. expect(result.error).toBeUndefined()
  4379. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  4380. // Serialized (escaped) size of retained entries stays in the budget's
  4381. // neighborhood: well under the ~30 kB an uncharged flood would retain.
  4382. const serialized = Buffer.byteLength(JSON.stringify(result.logs), 'utf8')
  4383. expect(serialized).toBeLessThan(1024)
  4384. })
  4385. it('charges the serialized cost child-side, so a control-heavy line truncates instead of being admitted whole', async () => {
  4386. // The child's ledger must charge what the entry costs on the wire, not its
  4387. // raw UTF-8 length: a NUL is one raw byte but six as its escape. A 24 MiB NUL
  4388. // line clears the cheap char-count lower bound (24 MiB < 32 MiB budget), so
  4389. // charging raw bytes would ADMIT it and emit a ~144 MiB escaped entry;
  4390. // charging the serialized cost (~144 MiB > the 32 MiB budget) rejects it
  4391. // before any encode and emits the marker instead. The address space (512 MiB,
  4392. // clearing the 12x load gate for a 32 MiB budget) is sized so the run loads;
  4393. // the gate separately guarantees a correctly-charged near-budget entry fits.
  4394. const { runtime } = await setup({ maxLogBytes: 32 * 1024 * 1024, addressSpaceMb: 512, maxWallMs: 20_000 })
  4395. const result = await runtime.run({
  4396. program: [
  4397. 'print("\\x00" * (24 * 1024 * 1024))',
  4398. 'return "done"',
  4399. ].join('\n'),
  4400. bindings: [],
  4401. })
  4402. expect(result.error).toBeUndefined()
  4403. expect(result.value).toBe('done')
  4404. expect(result.logs.filter(line => line.includes('log capture truncated'))).toHaveLength(1)
  4405. // Nothing of the line itself was retained: the ledger refused the whole entry.
  4406. expect(result.logs.every(line => !line.includes(String.fromCharCode(0)))).toBe(true)
  4407. }, 30_000)
  4408. it('bounds a huge unterminated tail after an early newline without copying it whole', async () => {
  4409. // The newline branch of _LogStream.write buffered the whole unterminated
  4410. // tail after the last newline into `_pending` before the flush trigger could
  4411. // bound it, so an early newline followed by a huge tail made a second full
  4412. // copy of the model's own string — a MemoryError the config gate cannot
  4413. // catch (the tail far exceeds maxLogBytes). The tail is now sliced to a
  4414. // budget-sized prefix, so the run truncates and completes. Linux-only RLIMIT_AS
  4415. // repro (Darwin skips the limit); on macOS this asserts the happy path.
  4416. //
  4417. // Sizing: the model builds `tail` (N) then the `"\n" + tail` write argument
  4418. // (another ~N), so construction peaks at ~2N — kept under the 384 MiB address
  4419. // space at N = 150 MiB (~300 MiB). The pre-fix code then buffered the whole
  4420. // ~150 MiB tail again, pushing past 384 MiB; the sliced prefix does not.
  4421. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 384, maxWallMs: 20_000 })
  4422. const result = await runtime.run({
  4423. program: [
  4424. 'import sys',
  4425. 'tail = "A" * (150 * 1024 * 1024)',
  4426. 'sys.stdout.write("\\n" + tail)',
  4427. 'return "done"',
  4428. ].join('\n'),
  4429. bindings: [],
  4430. })
  4431. expect(result.error).toBeUndefined()
  4432. expect(result.value).toBe('done')
  4433. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  4434. }, 30_000)
  4435. it('flushes logs before framing the value so their peaks do not add against RLIMIT_AS', async () => {
  4436. // The load gate bounds maxLogBytes and maxValueBytes INDEPENDENTLY against the
  4437. // address space, each at the 12x worst case. But the child framed the
  4438. // completion value (materializing its escaped form to meter it, then encoding
  4439. // the frame) while a newline-free log tail still sat unflushed in _pending.
  4440. // Those two peaks added: two budgets each admitted alone could together breach
  4441. // RLIMIT_AS, dying as worker-exit instead of settling. The flush now runs
  4442. // before the value is framed, so the log pending is freed first.
  4443. //
  4444. // Config: 32 MiB each against 512 MiB (each 32*12 = 384 MiB < 448 MiB
  4445. // budgetable, so both load). The program writes ~33M astral chars with no
  4446. // newline (buffered ~132 MB, under the char-count flush trigger) then returns
  4447. // ~33M astral chars — a ~132 MB serialized value that is itself OVER the 32 MiB
  4448. // maxValueBytes, so the correct outcome is `output-limit`. Pre-fix the
  4449. // unflushed 132 MB plus the value's build-and-encode (~396 MB) exceeded 512 MiB
  4450. // and OOM'd (reported as exception/worker-exit); flushing first lets the value
  4451. // check complete (~460 MB alone) and report output-limit. On Darwin (no
  4452. // RLIMIT_AS) the value is over budget too, so output-limit holds either way;
  4453. // the OOM the reorder prevents is the Linux-only failure.
  4454. const { runtime } = await setup({
  4455. maxLogBytes: 32 * 1024 * 1024,
  4456. maxValueBytes: 32 * 1024 * 1024,
  4457. addressSpaceMb: 512,
  4458. maxWallMs: 20_000,
  4459. })
  4460. const result = await runtime.run({
  4461. program: [
  4462. 'import sys',
  4463. 'sys.stdout.write("\\U0001F600" * 33_000_000)',
  4464. 'return "\\U0001F600" * 33_000_000',
  4465. ].join('\n'),
  4466. bindings: [],
  4467. })
  4468. expect(result.error?.kind).toBe('output-limit')
  4469. }, 30_000)
  4470. it('checks and encodes a wide completion value in O(depth), not O(width)', async () => {
  4471. // A wide flat list serializes to ~2 bytes per element but the pre-fix walk
  4472. // enqueued one traversal tuple per element (_check_done_value) and one stack
  4473. // entry plus a separator marker per element (_encode_json_plain) — ~56 bytes
  4474. // per element, ~28x the serialized size. A value the byte meter admits could
  4475. // therefore OOM on the checker's or encoder's own bookkeeping, the inversion
  4476. // the load gate exists to prevent (the gate reserves 12x, not 28x). Both now
  4477. // walk with an O(depth) cursor that pulls one child at a time, so the only
  4478. // width-proportional allocation is the output string the meter bounded.
  4479. //
  4480. // Config: maxValueBytes 20 MiB against 384 MiB (20*12 = 240 MiB < 320 MiB
  4481. // budgetable, so it loads). `[0] * 6_000_000` is ~12 MB of JSON, under the
  4482. // 20 MiB budget, so it must round-trip. Pre-fix the ~400 MB of per-element
  4483. // frames plus the interpreter exceeded 384 MiB and returned MemoryError as an
  4484. // exception. Linux-only RLIMIT_AS repro; on macOS the value round-trips
  4485. // either way, but the fixture stays within the address space so it is honest.
  4486. //
  4487. // `maxWallMs` is 60s, not the 20s the memory assertion alone needs: the O(depth)
  4488. // cursor pulls 6M elements one at a time through Python-level frames, which costs
  4489. // ~11s on an idle machine and more under the coverage lane's V8 instrumentation
  4490. // with several workers sharing a box. This budget bounds the run without letting a
  4491. // loaded runner's scheduling latency read as a `timeout` — what this test asserts
  4492. // is the O(depth) memory shape, not a speed claim.
  4493. const { runtime } = await setup({ maxValueBytes: 20 * 1024 * 1024, addressSpaceMb: 384, maxWallMs: 60_000 })
  4494. const result = await runtime.run({ program: 'return [0] * 6_000_000', bindings: [] })
  4495. expect(result.error).toBeUndefined()
  4496. expect(Array.isArray(result.value)).toBe(true)
  4497. expect((result.value as number[]).length).toBe(6_000_000)
  4498. }, 90_000)
  4499. it('validates wide binding arguments in O(depth), not O(width)', async () => {
  4500. // The completion-value walks are budgeted; this one is not. `dispatch` runs
  4501. // `_lossless_json_violation` on the arguments the MODEL built, and no
  4502. // child-side byte budget bounds them first: the frame ceiling is the host's
  4503. // and applies only after this validation returns. A per-member traversal
  4504. // frame therefore turned a legitimate call into the program's own
  4505. // MemoryError. Measured with tracemalloc on the two walk shapes over this
  4506. // exact argument (JSON ~17 MB): the cursor peaks at 0.0 MiB of auxiliary
  4507. // state, the pre-fix `stack.extend` at 459.1 MiB -- past the 384 MiB
  4508. // configured below, so the discriminating failure is real. It is Linux-only:
  4509. // Darwin skips RLIMIT_AS, so this case round-trips there either way.
  4510. //
  4511. // The binding echoes its argument's length back, so the assertion proves the
  4512. // call actually round-tripped rather than merely avoiding a crash.
  4513. const { runtime } = await setup({ addressSpaceMb: 384, maxWallMs: 60_000 })
  4514. const result = await runtime.run({
  4515. program: 'return await tools.width([0] * 6_000_000)',
  4516. bindings: [{
  4517. global: 'tools',
  4518. functions: { width: async (items: unknown) => (items as number[]).length },
  4519. }],
  4520. })
  4521. expect(result.error).toBeUndefined()
  4522. expect(result.value).toBe(6_000_000)
  4523. }, 90_000)
  4524. it('decodes a multi-megabyte binding reply without regex backtracking state', async () => {
  4525. // The child parses every host reply with `_decode_json_plain`. Its scalar
  4526. // regex matched strings with a `(?:[^"\\]|\\.)*` repetition, which makes
  4527. // CPython's backtracking engine retain state proportional to the string's
  4528. // WIDTH -- measured at ~146 MiB of engine state for a 1 MiB string and
  4529. // ~558 MiB for 4 MiB. A legitimate multi-megabyte reply therefore raised
  4530. // MemoryError inside `_pump_replies`; because that pump is the only settler
  4531. // of the call's future, the run stranded until the wall clock reported a
  4532. // `timeout` instead of returning the value the binding produced.
  4533. //
  4534. // Strings now scan chunk-to-chunk over a character class (no backtracking
  4535. // state). Measured on this exact 4 MiB reply: the pre-fix regex peaks at
  4536. // 557.8 MiB, past the default 512 MiB address space, while the scanner peaks
  4537. // at the 4.0 MiB result itself. Linux-only, like the other RLIMIT_AS repros:
  4538. // Darwin does not apply the limit, so the spike is merely allocated there.
  4539. const reply = 'A'.repeat(4 * 1024 * 1024)
  4540. const { runtime } = await setup({ maxWallMs: 60_000 })
  4541. const result = await runtime.run({
  4542. program: 'value = await tools.big({})\nreturn len(value)',
  4543. bindings: [{ global: 'tools', functions: { big: async () => reply } }],
  4544. })
  4545. expect(result.error).toBeUndefined()
  4546. expect(result.value).toBe(reply.length)
  4547. }, 90_000)
  4548. it('drops a late binding resolution before snapshotting it', async () => {
  4549. // `sendReply` checks `settled`, but only after the resolution has been walked
  4550. // and copied by `snapshotJsonValue`. Binding resolution carries no seam-level
  4551. // byte cap, so a binding that resolves a wide value AFTER the run already
  4552. // settled (here on `maxWallMs`) spent host heap building a frame that is then
  4553. // discarded. The check now runs before the snapshot.
  4554. //
  4555. // The binding resolves well after the 1s wall clock with a 2M-element array;
  4556. // the run must still report `timeout`, and the late value must not appear.
  4557. let resolvedLate = false
  4558. const { runtime } = await setup({ maxWallMs: 1_000 })
  4559. const result = await runtime.run({
  4560. program: 'return await tools.slow({})',
  4561. bindings: [{
  4562. global: 'tools',
  4563. functions: {
  4564. slow: async () => {
  4565. await new Promise(resolve => setTimeout(resolve, 2_500))
  4566. resolvedLate = true
  4567. return Array.from({ length: 2_000_000 }, () => 0)
  4568. },
  4569. },
  4570. }],
  4571. })
  4572. expect(result.error?.kind).toBe('timeout')
  4573. expect(result.value).toBeUndefined()
  4574. // Pin that the late path actually ran, so the assertion above is not vacuous.
  4575. await new Promise(resolve => setTimeout(resolve, 2_000))
  4576. expect(resolvedLate).toBe(true)
  4577. }, 90_000)
  4578. it('paces concurrent binding replies instead of queueing every frame at once', async () => {
  4579. // Binding resolution carries no seam-level byte cap. Before pacing, a program
  4580. // resolving several large values in one `asyncio.gather` round encoded them
  4581. // all in the same turn and queued every frame in fd 3's writable buffer,
  4582. // which exhausted the host heap and killed the whole process rather than
  4583. // failing the run. Replies are now encoded one at a time, waiting for
  4584. // `drain` when the pipe is full.
  4585. //
  4586. // Eight concurrent 4 MiB replies (32 MiB of frames) must all round-trip. The
  4587. // program sums the lengths, so the assertion proves every reply arrived and
  4588. // was matched to its own call -- pacing must not drop or misroute any. What
  4589. // this case cannot show is the peak itself, which lives in the stream's
  4590. // buffer: measured directly on a 64 KiB-highWaterMark pipe with this same
  4591. // 8x4 MiB shape, the unpaced writes buffered 32.0 MiB while the paced ones
  4592. // peaked at 0.0 MiB.
  4593. const chunk = 'A'.repeat(4 * 1024 * 1024)
  4594. const { runtime } = await setup({ maxWallMs: 60_000 })
  4595. const result = await runtime.run({
  4596. program: [
  4597. 'import asyncio',
  4598. 'parts = await asyncio.gather(*[tools.chunk({}) for _ in range(8)])',
  4599. 'return sum(len(p) for p in parts)',
  4600. ].join('\n'),
  4601. bindings: [{ global: 'tools', functions: { chunk: async () => chunk } }],
  4602. })
  4603. expect(result.error).toBeUndefined()
  4604. expect(result.value).toBe(8 * chunk.length)
  4605. }, 90_000)
  4606. it('bounds a flood of zero-byte log lines through the per-entry separator charge', async () => {
  4607. // Blank print() lines carry zero content bytes; without the +1 separator
  4608. // charge they would bypass maxLogBytes entirely and grow the retained
  4609. // array without bound. Each empty entry costs one byte, so a 64-byte
  4610. // budget retains at most 64 entries before the marker.
  4611. const { runtime } = await setup({ maxLogBytes: 64, maxWallMs: 10_000 })
  4612. const result = await runtime.run({
  4613. program: [
  4614. 'for _ in range(10000):',
  4615. ' print()',
  4616. 'return "done"',
  4617. ].join('\n'),
  4618. bindings: [],
  4619. })
  4620. expect(result.error).toBeUndefined()
  4621. expect(result.logs.length).toBeLessThanOrEqual(65)
  4622. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  4623. })
  4624. it('reassembles multibyte UTF-8 split across stray-output pipe chunks', async () => {
  4625. // A single os.write far past the 64 KiB pipe buffer forces multiple
  4626. // 'data' chunks; when the boundary lands inside a multibyte sequence,
  4627. // per-chunk decoding would corrupt it into replacement characters. Raw bytes
  4628. // are buffered and only decoded once a complete line (or the whole tail at
  4629. // flush) is assembled, so the split sequence is whole by the time it is
  4630. // decoded. The payload spans every valid multibyte lead class so
  4631. // accrueStrayCost's per-lead continuation ranges are all exercised: U+0900
  4632. // (E0 A4 80, the range-restricted E0 lead), U+4F60 and U+597D (E4/E5, plain
  4633. // 3-byte), U+1F600 (F0, the range-restricted F0 lead), and U+10FFFF (F4 8F
  4634. // BF BF, the range-restricted F4 lead).
  4635. const { runtime } = await setup({ maxLogBytes: 1024 * 1024 })
  4636. const result = await runtime.run({
  4637. program: [
  4638. 'import os',
  4639. // os.write is one syscall and returns a partial count on a full
  4640. // pipe, so loop until the whole payload (odd prefix -> a chunk
  4641. // boundary lands inside a multibyte sequence) is out.
  4642. String.raw`payload = b"a" * 65535 + "\u0900\u4f60\u597d\U0001f600\U0010ffff".encode("utf-8")`,
  4643. 'view = memoryview(payload)',
  4644. 'while view:',
  4645. ' view = view[os.write(1, view):]',
  4646. 'return "done"',
  4647. ].join('\n'),
  4648. bindings: [],
  4649. })
  4650. expect(result.error).toBeUndefined()
  4651. const text = result.logs.join('')
  4652. expect(text).toContain('\u0900\u4f60\u597d\u{1f600}\u{10ffff}')
  4653. expect(text).not.toContain('\ufffd')
  4654. })
  4655. it('flushes a stray-output byte sequence left incomplete when the pipe ends', async () => {
  4656. // The child writes the first two bytes of a 3-byte UTF-8 character to fd 1
  4657. // and exits, so the pipe closes with the sequence unfinished in the raw
  4658. // residual. The 'end' flush decodes the residual with `toString('utf8')`,
  4659. // which renders the stranded bytes as U+FFFD instead of dropping them.
  4660. const { runtime } = await setup({ maxLogBytes: 1024 * 1024 })
  4661. const result = await runtime.run({
  4662. program: [
  4663. 'import os',
  4664. // b"\xe4\xbd" is the leading two bytes of U+4F60; no continuation byte
  4665. // follows before exit.
  4666. String.raw`os.write(1, b"\xe4\xbd")`,
  4667. 'return "done"',
  4668. ].join('\n'),
  4669. bindings: [],
  4670. })
  4671. expect(result.error).toBeUndefined()
  4672. expect(result.logs.join('')).toContain('�')
  4673. })
  4674. it('rejects reserved words of EITHER backend language as binding globals', async () => {
  4675. // The seam's portable contract: `lambda` (Python keyword, legal JS name)
  4676. // and `typeof` (JS keyword, legal Python name) are both refused, so a
  4677. // namespace list valid on one backend is valid on every backend.
  4678. const { runtime } = await setup()
  4679. for (const global of ['lambda', 'typeof']) {
  4680. await expect(runtime.run({
  4681. program: 'return 1',
  4682. bindings: [{ global, functions: {} }],
  4683. })).rejects.toThrow(/is not a usable Python identifier/)
  4684. }
  4685. })
  4686. it('captures stray stdout bytes the child writes bypassing sys.stdout', async () => {
  4687. // Model code that writes to fd 1 via os.write() bypasses the Python-side
  4688. // LogBuffer, so the host's stray-byte capture on child.stdout is what
  4689. // records it.
  4690. const { runtime } = await setup()
  4691. const result = await runtime.run({
  4692. program: [
  4693. 'import os',
  4694. 'os.write(1, b"stray stdout\\n")',
  4695. 'os.write(2, b"stray stderr\\n")',
  4696. 'return "done"',
  4697. ].join('\n'),
  4698. bindings: [],
  4699. })
  4700. expect(result.error).toBeUndefined()
  4701. expect(result.logs.join('')).toContain('stray stdout')
  4702. expect(result.logs.join('')).toContain('stray stderr')
  4703. })
  4704. it('flushes bytes written through sys.__stdout__/sys.__stderr__ before the done frame', async () => {
  4705. // The bootstrap only replaces sys.stdout/sys.stderr with the _LogStream;
  4706. // sys.__stdout__/sys.__stderr__ are the original block-buffered wrappers
  4707. // over fd 1/2. A program that writes through them without an explicit flush
  4708. // would lose those bytes when the host SIGTERMs the child right after the
  4709. // done frame (the default SIGTERM disposition terminates without
  4710. // interpreter finalization). The settlement flush now drains the original
  4711. // std streams before sending the done frame, so the bytes land in the
  4712. // kernel pipe buffer and the host's stray capture records them.
  4713. const { runtime } = await setup()
  4714. const result = await runtime.run({
  4715. program: [
  4716. 'import sys',
  4717. // `-u` makes the streams write-through; re-enable block buffering so
  4718. // the bytes sit in the wrapper until the SETTLEMENT drain flushes them
  4719. // — the drain path, not the -u immediate write, is what this case pins.
  4720. 'if hasattr(sys.__stdout__, "reconfigure"):',
  4721. ' sys.__stdout__.reconfigure(write_through=False)',
  4722. ' sys.__stderr__.reconfigure(write_through=False)',
  4723. 'sys.__stdout__.write("orig stdout\\n")',
  4724. 'sys.__stderr__.write("orig stderr\\n")',
  4725. 'return "done"',
  4726. ].join('\n'),
  4727. bindings: [],
  4728. })
  4729. expect(result.error).toBeUndefined()
  4730. expect(result.value).toBe('done')
  4731. expect(result.logs.join('')).toContain('orig stdout')
  4732. expect(result.logs.join('')).toContain('orig stderr')
  4733. }, 15_000)
  4734. it('escalates to SIGKILL when the program traps SIGTERM and ignores the grace period', async () => {
  4735. // A program that traps SIGTERM should still die: the kill() escalation
  4736. // fires SIGKILL after graceMs. The full run reports either timeout (wall)
  4737. // or worker-exit depending on which finish reason wins the race.
  4738. const { runtime } = await setup({ maxWallMs: 400, graceMs: 200 })
  4739. const result = await runtime.run({
  4740. program: [
  4741. 'import signal, time',
  4742. 'signal.signal(signal.SIGTERM, lambda *a: None)',
  4743. 'while True: time.sleep(1)',
  4744. ].join('\n'),
  4745. bindings: [],
  4746. })
  4747. expect(['timeout', 'worker-exit']).toContain(result.error?.kind)
  4748. }, 6000)
  4749. it('bounds the fd-3 receive buffer against a newline-free flood', async () => {
  4750. // A program looping os.write(3, ...) with no newline would grow the host
  4751. // accumulator unbounded (the child's RLIMIT_AS does not cover the host
  4752. // string). The frame cap is a fixed 64 MiB memory-safety invariant —
  4753. // deliberately NOT derived from maxValueBytes, because legitimate binding
  4754. // call frames may be large. We flood slightly past it in 8 MiB writes so
  4755. // the test terminates promptly once the guard trips.
  4756. const ceiling = 64 * 1024 * 1024
  4757. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  4758. const start = Date.now()
  4759. const result = await runtime.run({
  4760. program: [
  4761. 'import os',
  4762. `for _ in range(${Math.ceil((ceiling * 1.1) / (8 * 1024 * 1024))}):`,
  4763. ' os.write(3, b"A" * (8 * 1024 * 1024))',
  4764. 'return "never"',
  4765. ].join('\n'),
  4766. bindings: [],
  4767. })
  4768. const elapsed = Date.now() - start
  4769. expect(result.error?.kind).toBe('worker-exit')
  4770. expect(result.error?.message).toContain(`protocol frame exceeded ${ceiling} bytes`)
  4771. // The breach ends the run before the wall ceiling (the run did not idle
  4772. // out); absolute pipe throughput varies too much under parallel suites
  4773. // for a tight bound.
  4774. expect(elapsed).toBeLessThan(30_000)
  4775. }, 45_000)
  4776. it('fails a forged oversized done value host-side as output-limit', async () => {
  4777. // The Python-side _done_with_value check is bypassable by writing a done
  4778. // frame straight to fd 3. The host re-enforces maxValueBytes; the seam
  4779. // forbids substituting a truncated value, so the run FAILS as output-limit
  4780. // instead of returning a lie.
  4781. const maxValueBytes = 64
  4782. const { runtime } = await setup({ maxValueBytes })
  4783. const result = await runtime.run({
  4784. program: [
  4785. 'import os, json',
  4786. 'big = "B" * 5000',
  4787. 'os.write(3, json.dumps({"type":"done","value":big}).encode() + b"\\n")',
  4788. // The real done never sends; the forged one settles the run.
  4789. 'import time',
  4790. 'time.sleep(5)',
  4791. ].join('\n'),
  4792. bindings: [],
  4793. })
  4794. expect(result.value).toBeUndefined()
  4795. expect(result.error?.kind).toBe('output-limit')
  4796. expect(result.error?.message).toContain('exceeded 64 bytes')
  4797. }, 8000)
  4798. it('drops a forged oversized log frame on its code-unit lower bound, before escaping it', async () => {
  4799. // A forged `log` frame carrying a control-heavy string: NULs escape
  4800. // several-fold (one NUL -> six bytes `\u0000`). The raw frame stays under
  4801. // the host's 64 MiB parse cap (4 MiB of `\u0000` text = 24 MiB raw) while
  4802. // the escaped form would be ~24 MiB. Charging it required building that
  4803. // escaped copy first, so a 32-byte maxLogBytes could still force a large
  4804. // host allocation. The cheap `length + 3` lower bound truncates it instead.
  4805. // The host's own heap is what is under test, so keep the child's address
  4806. // space generous enough to BUILD the frame.
  4807. const { runtime } = await setup({ maxLogBytes: 128, addressSpaceMb: 1024, maxWallMs: 60_000 })
  4808. const before = process.memoryUsage().heapUsed
  4809. const result = await runtime.run({
  4810. program: [
  4811. 'import os',
  4812. // Written as a raw frame so the child's own ledger never sees it.
  4813. 'os.write(3, b\'{"type":"log","text":"\' + b"\\\\u0000" * (4 * 1024 * 1024) + b\'"}\\n\')',
  4814. 'return "settled"',
  4815. ].join('\n'),
  4816. bindings: [],
  4817. })
  4818. expect(result.error).toBeUndefined()
  4819. expect(result.value).toBe('settled')
  4820. // The frame was dropped as one truncation marker, not retained.
  4821. expect(result.logs).toEqual([logTruncationMarker(128)])
  4822. // The escaped copy (~144 MiB) was never materialized.
  4823. expect(process.memoryUsage().heapUsed - before).toBeLessThan(256 * 1024 * 1024)
  4824. }, 90_000)
  4825. it('charges a forged log frame its escaped cost once past the code-unit lower bound', async () => {
  4826. // The cheap lower bound only rejects what cannot possibly fit; a SHORT
  4827. // control-heavy frame clears it and must still be charged what it costs on
  4828. // the wire. Eleven NULs are 14 against the 64-byte ledger's cheap bound
  4829. // (six bytes each, two quotes, one separator), so the full charge truncates.
  4830. const { runtime } = await setup({ maxLogBytes: 64 })
  4831. const result = await runtime.run({
  4832. program: [
  4833. 'import os',
  4834. 'os.write(3, b\'{"type":"log","text":"\' + b"\\\\u0000" * 11 + b\'"}\\n\')',
  4835. 'return "settled"',
  4836. ].join('\n'),
  4837. bindings: [],
  4838. })
  4839. expect(result.error).toBeUndefined()
  4840. expect(result.value).toBe('settled')
  4841. expect(result.logs).toEqual([logTruncationMarker(64)])
  4842. }, 8000)
  4843. it('caps a forged done error.message from its code-unit prefix, never encoding the whole message', async () => {
  4844. // `Buffer.from(message)` on a message near the frame ceiling allocates a
  4845. // full UTF-8 copy before maxValueBytes applies. Only the first
  4846. // maxValueBytes code units can fit the cap, so only that prefix is encoded
  4847. // — at most 3x the cap in bytes. The message here is 48 MiB of ASCII: its
  4848. // full encode would be another 48 MiB in the host.
  4849. const maxValueBytes = 64
  4850. const { runtime } = await setup({ maxValueBytes, addressSpaceMb: 1024, maxWallMs: 60_000 })
  4851. const before = process.memoryUsage().heapUsed
  4852. const result = await runtime.run({
  4853. program: [
  4854. 'import os',
  4855. 'os.write(3, b\'{"type":"done","error":{"kind":"exception","message":"\' + b"E" * (48 * 1024 * 1024) + b\'"}}\\n\')',
  4856. 'import time',
  4857. 'time.sleep(30)',
  4858. ].join('\n'),
  4859. bindings: [],
  4860. })
  4861. expect(result.error?.kind).toBe('exception')
  4862. const message = result.error?.message ?? ''
  4863. // The marker's 15 bytes come OUT of the 64-byte cap, so 49 E's precede it
  4864. // and the whole string is exactly 64 bytes — not 64 plus the marker.
  4865. expect(message).toBe(`${'E'.repeat(maxValueBytes - 15)}… [truncated]`)
  4866. expect(Buffer.byteLength(message, 'utf8')).toBe(maxValueBytes)
  4867. // JSON.parse already holds the 48 MiB string; the cap must not add a
  4868. // second full-length copy on top of it.
  4869. expect(process.memoryUsage().heapUsed - before).toBeLessThan(256 * 1024 * 1024)
  4870. }, 90_000)
  4871. it('keeps a capped diagnostic within maxValueBytes, marker included', async () => {
  4872. // The marker is part of the emitted diagnostic, so its bytes are reserved
  4873. // from the cap rather than appended past it — the host meters this same
  4874. // field downstream. Checked on BOTH producers: the child's own _cap_message
  4875. // (a raised exception) and the host's capMessage (a forged done frame).
  4876. const maxValueBytes = 40
  4877. const { runtime } = await setup({ maxValueBytes })
  4878. const raised = await runtime.run({
  4879. program: 'raise ValueError("R" * 100000)',
  4880. bindings: [],
  4881. })
  4882. expect(raised.error?.kind).toBe('exception')
  4883. const raisedMessage = raised.error?.message ?? ''
  4884. expect(raisedMessage.endsWith('… [truncated]')).toBe(true)
  4885. expect(Buffer.byteLength(raisedMessage, 'utf8')).toBeLessThanOrEqual(maxValueBytes)
  4886. const forged = await runtime.run({
  4887. program: [
  4888. 'import os, json',
  4889. 'msg = "F" * 100000',
  4890. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":msg}}).encode() + b"\\n")',
  4891. 'import time',
  4892. 'time.sleep(5)',
  4893. ].join('\n'),
  4894. bindings: [],
  4895. })
  4896. expect(forged.error?.kind).toBe('exception')
  4897. const forgedMessage = forged.error?.message ?? ''
  4898. expect(forgedMessage.endsWith('… [truncated]')).toBe(true)
  4899. expect(Buffer.byteLength(forgedMessage, 'utf8')).toBe(maxValueBytes)
  4900. }, 15_000)
  4901. it('emits the marker alone when the cap is smaller than the marker itself', async () => {
  4902. // With maxValueBytes below the marker's own 15 bytes there is no room for
  4903. // message text; the marker still goes out, so the truncation stays reported
  4904. // instead of the diagnostic silently becoming empty. Both producers agree.
  4905. const { runtime } = await setup({ maxValueBytes: 4 })
  4906. const raised = await runtime.run({ program: 'raise ValueError("R" * 500)', bindings: [] })
  4907. expect(raised.error?.kind).toBe('exception')
  4908. expect(raised.error?.message).toBe('… [truncated]')
  4909. const forged = await runtime.run({
  4910. program: [
  4911. 'import os, json',
  4912. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":"F" * 500}}).encode() + b"\\n")',
  4913. 'import time',
  4914. 'time.sleep(5)',
  4915. ].join('\n'),
  4916. bindings: [],
  4917. })
  4918. expect(forged.error?.kind).toBe('exception')
  4919. expect(forged.error?.message).toBe('… [truncated]')
  4920. }, 15_000)
  4921. it('caps a forged done error.message without splitting a surrogate pair', async () => {
  4922. // At exactly maxValueBytes code units the prefix can end on a high
  4923. // surrogate whose low half sits just outside it. `Buffer.from` encodes that
  4924. // orphan as U+FFFD — the same corruption a mid-sequence byte cut causes —
  4925. // and those three replacement bytes sit past the marker-reserved budget, so
  4926. // the byte trim-back drops them.
  4927. const maxValueBytes = 32
  4928. const { runtime } = await setup({ maxValueBytes })
  4929. const result = await runtime.run({
  4930. program: [
  4931. 'import os, json',
  4932. // 32 ASCII chars then astral characters: code unit 32 is the first
  4933. // character's high surrogate (Python spells it as one code point, so
  4934. // json.dumps emits the raw 4 bytes the host reads back as a pair).
  4935. 'msg = "A" * 32 + "\\U0001f600" * 4',
  4936. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":msg}}).encode() + b"\\n")',
  4937. 'import time',
  4938. 'time.sleep(5)',
  4939. ].join('\n'),
  4940. bindings: [],
  4941. })
  4942. expect(result.error?.kind).toBe('exception')
  4943. // 17 A's fill the marker-reserved budget; no orphaned half, no U+FFFD.
  4944. expect(result.error?.message).toBe(`${'A'.repeat(17)}… [truncated]`)
  4945. expect(Buffer.byteLength(result.error?.message ?? '', 'utf8')).toBe(maxValueBytes)
  4946. }, 8000)
  4947. it('returns a diagnostic under a third of the cap untouched, skipping the encode', async () => {
  4948. // Under maxValueBytes/3 code units a message cannot overflow the cap
  4949. // whatever it holds (3 bytes is the per-code-unit maximum), so the fast
  4950. // path returns it without encoding anything. Non-ASCII proves the bound is
  4951. // the code-unit count, not a byte assumption: 6 characters at 3 bytes each
  4952. // is 18 bytes, inside the 64-byte cap.
  4953. const { runtime } = await setup({ maxValueBytes: 64 })
  4954. const result = await runtime.run({
  4955. program: [
  4956. 'import os, json',
  4957. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":"中文中文中文"}}).encode() + b"\\n")',
  4958. 'import time',
  4959. 'time.sleep(5)',
  4960. ].join('\n'),
  4961. bindings: [],
  4962. })
  4963. expect(result.error?.kind).toBe('exception')
  4964. expect(result.error?.message).toBe('中文中文中文')
  4965. }, 8000)
  4966. it('re-caps a forged done error.message host-side on a UTF-8 boundary', async () => {
  4967. // A forged done frame can carry an arbitrarily long error message; the
  4968. // host caps it to maxValueBytes and appends the shared marker. The
  4969. // message is emoji-dense and the cap is chosen so the marker-reserved
  4970. // 51-byte cut lands INSIDE a 4-byte sequence (one ASCII byte then 4-byte
  4971. // runs, so only a cut at 1 + 4k is aligned) — the cap must trim back to a
  4972. // code-point boundary rather than decode a replacement character, which
  4973. // would also exceed the cap.
  4974. const maxValueBytes = 66
  4975. const { runtime } = await setup({ maxValueBytes })
  4976. const result = await runtime.run({
  4977. program: [
  4978. 'import os, json',
  4979. 'msg = "E" + "\\U0001f600" * 2000',
  4980. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":msg}}).encode() + b"\\n")',
  4981. 'import time',
  4982. 'time.sleep(5)',
  4983. ].join('\n'),
  4984. bindings: [],
  4985. })
  4986. expect(result.error?.kind).toBe('exception')
  4987. const message = result.error?.message ?? ''
  4988. expect(message.endsWith('… [truncated]')).toBe(true)
  4989. const marker = '… [truncated]'
  4990. const body = message.slice(0, message.length - marker.length)
  4991. // The WHOLE message, marker included, honors the cap.
  4992. expect(Buffer.byteLength(message, 'utf8')).toBeLessThanOrEqual(maxValueBytes)
  4993. // 'E' plus 12 emoji is 49 bytes: the trim-back walked the 51-byte budget
  4994. // down past two continuation bytes rather than splitting the 13th.
  4995. expect(body).toBe(`E${'\u{1f600}'.repeat(12)}`)
  4996. // The cut landed on a code-point boundary — no replacement character.
  4997. expect(body).not.toContain('\ufffd')
  4998. }, 8000)
  4999. it('bounds a single oversized newline-terminated line on fd 3', async () => {
  5000. // The same cap applies to one giant framed line. Write EXACTLY the
  5001. // cap with no newline — at the limit, not past it, so nothing trips —
  5002. // then a small newline tail, which is the chunk that crosses.
  5003. const ceiling = 64 * 1024 * 1024
  5004. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  5005. const result = await runtime.run({
  5006. program: [
  5007. 'import os',
  5008. 'chunk = b"A" * (8 * 1024 * 1024)',
  5009. `for _ in range(${ceiling / (8 * 1024 * 1024)}):`,
  5010. ' os.write(3, chunk)',
  5011. 'os.write(3, b"AAAA\\n")',
  5012. 'return "never"',
  5013. ].join('\n'),
  5014. bindings: [],
  5015. })
  5016. expect(result.error?.kind).toBe('worker-exit')
  5017. expect(result.error?.message).toContain(`protocol frame exceeded ${ceiling} bytes`)
  5018. }, 90_000)
  5019. it('rejects an over-cap newline-free fd-3 buffer without first joining it into one line', async () => {
  5020. // The cap has to be enforced on the byte COUNTER before Buffer.concat,
  5021. // not on the joined line afterwards: the join is a second copy of
  5022. // everything held, so a program could force roughly twice the advertised
  5023. // 64 MiB of host memory before anything rejected it.
  5024. //
  5025. // This program writes past the cap with no newline: the counter crosses on
  5026. // the 9th 8 MiB write (72 MiB) while the buffer is still a single unframed
  5027. // line, so the pre-join check rejects it without concat-ing a second copy.
  5028. // Checking the joined line instead would have produced a 72 MiB FIRST LINE
  5029. // that the per-line bound then dropped only after the doubling had happened.
  5030. const ceiling = 64 * 1024 * 1024
  5031. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  5032. const result = await runtime.run({
  5033. program: [
  5034. 'import os',
  5035. 'chunk = b"A" * (8 * 1024 * 1024)',
  5036. `for _ in range(${ceiling / (8 * 1024 * 1024) + 1}):`,
  5037. ' os.write(3, chunk)',
  5038. 'return "never"',
  5039. ].join('\n'),
  5040. bindings: [],
  5041. })
  5042. expect(result.value).toBeUndefined()
  5043. expect(result.error?.kind).toBe('worker-exit')
  5044. expect(result.error?.message).toContain(`protocol frame exceeded ${ceiling} bytes`)
  5045. }, 120_000)
  5046. it('keeps two within-cap frames whose combined buffer crosses the cap', async () => {
  5047. // The unframed byte counter charges the WHOLE buffer, which legitimately
  5048. // holds several frames each within FRAME_PARSE_CAP_BYTES. A first frame of
  5049. // exactly the cap followed by a second frame crosses the counter without
  5050. // either frame exceeding the cap; the first-frame check (not the counter)
  5051. // must let them through, or a legitimate near-cap frame plus a trailing
  5052. // frame would be misreported as a worker-exit.
  5053. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  5054. const result = await runtime.run({
  5055. program: [
  5056. 'import os',
  5057. 'chunk = b"A" * (8 * 1024 * 1024)',
  5058. // Exactly the cap, no newline — at the limit, so nothing trips.
  5059. 'for _ in range(8):',
  5060. ' os.write(3, chunk)',
  5061. // A newline, then a small legitimate log frame.
  5062. 'os.write(3, b"\\n{\\"type\\":\\"log\\",\\"text\\":\\"after-cap-frames\\"}\\n")',
  5063. 'return "done"',
  5064. ].join('\n'),
  5065. bindings: [],
  5066. })
  5067. expect(result.error).toBeUndefined()
  5068. expect(result.value).toBe('done')
  5069. expect(result.logs).toContain('after-cap-frames')
  5070. }, 120_000)
  5071. it('rejects an oversized first frame that lands on the sealing threshold with a newline', async () => {
  5072. // The fragment-count seal runs only on newline-free chunks (the ELSE half
  5073. // of the newline branch), so a chunk that carries the first newline always
  5074. // reaches the join and its first-frame check; sealing it into a block
  5075. // would empty pendingChunks, leave sawNewline false, and skip that check.
  5076. // Whether the pipe delivers exactly 1024 chunks is timing-dependent, but
  5077. // the oversized first frame (63.9 MiB of A's + 12289 more before the
  5078. // newline) exceeds FRAME_PARSE_CAP_BYTES no matter how it arrives — the
  5079. // case pins the worker-exit settlement, not a pre/post copy-count
  5080. // distinction (both orders reject an over-cap frame).
  5081. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  5082. const result = await runtime.run({
  5083. program: [
  5084. 'import os',
  5085. // 4 KiB writes are <= PIPE_BUF, so each os.write is atomic and the
  5086. // host sees one chunk per write; 16384 of them accumulate 64 MiB of
  5087. // newline-free bytes (16 fragment-count seals of 1024 chunks).
  5088. 'chunk = b"A" * 4096',
  5089. 'for _ in range(16384):',
  5090. ' os.write(3, chunk)',
  5091. // 12289 more A's push the first frame past 64 MiB; drain-loop so the
  5092. // write cannot truncate, then a newline and a small legitimate frame.
  5093. "data = b'A' * 12289 + b'\\n' + b'{\"type\":\"log\",\"text\":\"after-seal\"}\\n'",
  5094. 'view = memoryview(data)',
  5095. 'while view:',
  5096. ' view = view[os.write(3, view):]',
  5097. 'return "done"',
  5098. ].join('\n'),
  5099. bindings: [],
  5100. })
  5101. expect(result.error?.kind).toBe('worker-exit')
  5102. expect(result.error?.message).toContain('protocol frame exceeded')
  5103. }, 120_000)
  5104. })