runtime.spec.ts 252 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131213221332134213521362137213821392140214121422143214421452146214721482149215021512152215321542155215621572158215921602161216221632164216521662167216821692170217121722173217421752176217721782179218021812182218321842185218621872188218921902191219221932194219521962197219821992200220122022203220422052206220722082209221022112212221322142215221622172218221922202221222222232224222522262227222822292230223122322233223422352236223722382239224022412242224322442245224622472248224922502251225222532254225522562257225822592260226122622263226422652266226722682269227022712272227322742275227622772278227922802281228222832284228522862287228822892290229122922293229422952296229722982299230023012302230323042305230623072308230923102311231223132314231523162317231823192320232123222323232423252326232723282329233023312332233323342335233623372338233923402341234223432344234523462347234823492350235123522353235423552356235723582359236023612362236323642365236623672368236923702371237223732374237523762377237823792380238123822383238423852386238723882389239023912392239323942395239623972398239924002401240224032404240524062407240824092410241124122413241424152416241724182419242024212422242324242425242624272428242924302431243224332434243524362437243824392440244124422443244424452446244724482449245024512452245324542455245624572458245924602461246224632464246524662467246824692470247124722473247424752476247724782479248024812482248324842485248624872488248924902491249224932494249524962497249824992500250125022503250425052506250725082509251025112512251325142515251625172518251925202521252225232524252525262527252825292530253125322533253425352536253725382539254025412542254325442545254625472548254925502551255225532554255525562557255825592560256125622563256425652566256725682569257025712572257325742575257625772578257925802581258225832584258525862587258825892590259125922593259425952596259725982599260026012602260326042605260626072608260926102611261226132614261526162617261826192620262126222623262426252626262726282629263026312632263326342635263626372638263926402641264226432644264526462647264826492650265126522653265426552656265726582659266026612662266326642665266626672668266926702671267226732674267526762677267826792680268126822683268426852686268726882689269026912692269326942695269626972698269927002701270227032704270527062707270827092710271127122713271427152716271727182719272027212722272327242725272627272728272927302731273227332734273527362737273827392740274127422743274427452746274727482749275027512752275327542755275627572758275927602761276227632764276527662767276827692770277127722773277427752776277727782779278027812782278327842785278627872788278927902791279227932794279527962797279827992800280128022803280428052806280728082809281028112812281328142815281628172818281928202821282228232824282528262827282828292830283128322833283428352836283728382839284028412842284328442845284628472848284928502851285228532854285528562857285828592860286128622863286428652866286728682869287028712872287328742875287628772878287928802881288228832884288528862887288828892890289128922893289428952896289728982899290029012902290329042905290629072908290929102911291229132914291529162917291829192920292129222923292429252926292729282929293029312932293329342935293629372938293929402941294229432944294529462947294829492950295129522953295429552956295729582959296029612962296329642965296629672968296929702971297229732974297529762977297829792980298129822983298429852986298729882989299029912992299329942995299629972998299930003001300230033004300530063007300830093010301130123013301430153016301730183019302030213022302330243025302630273028302930303031303230333034303530363037303830393040304130423043304430453046304730483049305030513052305330543055305630573058305930603061306230633064306530663067306830693070307130723073307430753076307730783079308030813082308330843085308630873088308930903091309230933094309530963097309830993100310131023103310431053106310731083109311031113112311331143115311631173118311931203121312231233124312531263127312831293130313131323133313431353136313731383139314031413142314331443145314631473148314931503151315231533154315531563157315831593160316131623163316431653166316731683169317031713172317331743175317631773178317931803181318231833184318531863187318831893190319131923193319431953196319731983199320032013202320332043205320632073208320932103211321232133214321532163217321832193220322132223223322432253226322732283229323032313232323332343235323632373238323932403241324232433244324532463247324832493250325132523253325432553256325732583259326032613262326332643265326632673268326932703271327232733274327532763277327832793280328132823283328432853286328732883289329032913292329332943295329632973298329933003301330233033304330533063307330833093310331133123313331433153316331733183319332033213322332333243325332633273328332933303331333233333334333533363337333833393340334133423343334433453346334733483349335033513352335333543355335633573358335933603361336233633364336533663367336833693370337133723373337433753376337733783379338033813382338333843385338633873388338933903391339233933394339533963397339833993400340134023403340434053406340734083409341034113412341334143415341634173418341934203421342234233424342534263427342834293430343134323433343434353436343734383439344034413442344334443445344634473448344934503451345234533454345534563457345834593460346134623463346434653466346734683469347034713472347334743475347634773478347934803481348234833484348534863487348834893490349134923493349434953496349734983499350035013502350335043505350635073508350935103511351235133514351535163517351835193520352135223523352435253526352735283529353035313532353335343535353635373538353935403541354235433544354535463547354835493550355135523553355435553556355735583559356035613562356335643565356635673568356935703571357235733574357535763577357835793580358135823583358435853586358735883589359035913592359335943595359635973598359936003601360236033604360536063607360836093610361136123613361436153616361736183619362036213622362336243625362636273628362936303631363236333634363536363637363836393640364136423643364436453646364736483649365036513652365336543655365636573658365936603661366236633664366536663667366836693670367136723673367436753676367736783679368036813682368336843685368636873688368936903691369236933694369536963697369836993700370137023703370437053706370737083709371037113712371337143715371637173718371937203721372237233724372537263727372837293730373137323733373437353736373737383739374037413742374337443745374637473748374937503751375237533754375537563757375837593760376137623763376437653766376737683769377037713772377337743775377637773778377937803781378237833784378537863787378837893790379137923793379437953796379737983799380038013802380338043805380638073808380938103811381238133814381538163817381838193820382138223823382438253826382738283829383038313832383338343835383638373838383938403841384238433844384538463847384838493850385138523853385438553856385738583859386038613862386338643865386638673868386938703871387238733874387538763877387838793880388138823883388438853886388738883889389038913892389338943895389638973898389939003901390239033904390539063907390839093910391139123913391439153916391739183919392039213922392339243925392639273928392939303931393239333934393539363937393839393940394139423943394439453946394739483949395039513952395339543955395639573958395939603961396239633964396539663967396839693970397139723973397439753976397739783979398039813982398339843985398639873988398939903991399239933994399539963997399839994000400140024003400440054006400740084009401040114012401340144015401640174018401940204021402240234024402540264027402840294030403140324033403440354036403740384039404040414042404340444045404640474048404940504051405240534054405540564057405840594060406140624063406440654066406740684069407040714072407340744075407640774078407940804081408240834084408540864087408840894090409140924093409440954096409740984099410041014102410341044105410641074108410941104111411241134114411541164117411841194120412141224123412441254126412741284129413041314132413341344135413641374138413941404141414241434144414541464147414841494150415141524153415441554156415741584159416041614162416341644165416641674168416941704171417241734174417541764177417841794180418141824183418441854186418741884189419041914192419341944195419641974198419942004201420242034204420542064207420842094210421142124213421442154216421742184219422042214222422342244225422642274228422942304231423242334234423542364237423842394240424142424243424442454246424742484249425042514252425342544255425642574258425942604261426242634264426542664267426842694270427142724273427442754276427742784279428042814282428342844285428642874288428942904291429242934294429542964297429842994300430143024303430443054306430743084309431043114312431343144315431643174318431943204321432243234324432543264327432843294330433143324333433443354336433743384339434043414342434343444345434643474348434943504351435243534354435543564357435843594360436143624363436443654366436743684369437043714372437343744375437643774378437943804381438243834384438543864387438843894390439143924393439443954396439743984399440044014402440344044405440644074408440944104411441244134414441544164417441844194420442144224423442444254426442744284429443044314432443344344435443644374438443944404441444244434444444544464447444844494450445144524453445444554456445744584459446044614462446344644465446644674468446944704471447244734474447544764477447844794480448144824483448444854486448744884489449044914492449344944495449644974498449945004501450245034504450545064507450845094510451145124513451445154516451745184519452045214522452345244525452645274528452945304531453245334534453545364537453845394540454145424543454445454546454745484549455045514552455345544555455645574558455945604561456245634564456545664567456845694570457145724573457445754576457745784579458045814582458345844585458645874588458945904591459245934594459545964597459845994600460146024603460446054606460746084609461046114612461346144615461646174618461946204621462246234624462546264627462846294630463146324633463446354636463746384639464046414642464346444645464646474648464946504651465246534654465546564657465846594660466146624663466446654666466746684669467046714672467346744675467646774678467946804681468246834684468546864687468846894690469146924693469446954696469746984699470047014702470347044705470647074708470947104711471247134714471547164717471847194720472147224723472447254726472747284729473047314732473347344735473647374738473947404741474247434744474547464747474847494750475147524753475447554756475747584759476047614762476347644765476647674768476947704771477247734774477547764777477847794780478147824783478447854786478747884789479047914792479347944795479647974798479948004801480248034804480548064807480848094810481148124813481448154816481748184819482048214822482348244825482648274828482948304831483248334834483548364837483848394840484148424843484448454846484748484849485048514852485348544855485648574858485948604861486248634864486548664867486848694870487148724873487448754876487748784879488048814882488348844885488648874888488948904891489248934894489548964897489848994900490149024903490449054906490749084909491049114912491349144915491649174918491949204921492249234924492549264927492849294930493149324933493449354936493749384939494049414942494349444945494649474948494949504951495249534954495549564957495849594960496149624963496449654966496749684969497049714972497349744975497649774978497949804981498249834984498549864987498849894990499149924993499449954996499749984999500050015002500350045005500650075008500950105011501250135014501550165017501850195020502150225023502450255026502750285029503050315032503350345035503650375038503950405041504250435044504550465047504850495050505150525053505450555056505750585059506050615062506350645065506650675068506950705071507250735074507550765077507850795080508150825083508450855086508750885089509050915092509350945095509650975098509951005101510251035104510551065107510851095110511151125113511451155116511751185119512051215122512351245125512651275128512951305131513251335134513551365137513851395140514151425143514451455146514751485149515051515152515351545155515651575158515951605161516251635164516551665167516851695170517151725173517451755176517751785179518051815182518351845185518651875188518951905191519251935194519551965197519851995200520152025203520452055206520752085209521052115212521352145215521652175218521952205221522252235224522552265227522852295230523152325233523452355236523752385239524052415242
  1. import { existsSync, readdirSync, realpathSync, statSync } 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 } from '../src/index.ts'
  8. import { logTruncationMarker } from '../src/protocol.ts'
  9. import type { Config } from '../src/index.ts'
  10. import type { CodeBindingFunction, CodeJsonValue, CodeRunResult } from '@deepseek-ai/dsh-code-runtime'
  11. /**
  12. * Names one `py/` script whose `copyFileSync` must fail, for the partial-staging
  13. * case. A real disk-full or missing-asset failure mid-copy cannot be produced
  14. * from a test, and the leak only shows when `mkdtempSync` has already succeeded.
  15. */
  16. const { failNextCopyOf } = vi.hoisted(() => ({ failNextCopyOf: { value: undefined as string | undefined } }))
  17. vi.mock('node:fs', async (importOriginal) => {
  18. const actual = await importOriginal<typeof import('node:fs')>()
  19. return {
  20. ...actual,
  21. copyFileSync(source: string, destination: string): void {
  22. if (failNextCopyOf.value !== undefined && basename(source) === failNextCopyOf.value) {
  23. failNextCopyOf.value = undefined
  24. throw Object.assign(new Error('simulated ENOSPC on copy'), { code: 'ENOSPC' })
  25. }
  26. actual.copyFileSync(source, destination)
  27. },
  28. }
  29. })
  30. /**
  31. * Integration suite over REAL python3 subprocesses (no subprocess mocks — it is
  32. * cheap and local, per docs/testing.md's real-over-mock policy; the only mock is
  33. * `node:fs.copyFileSync` for the staging-failure cases). Each test builds a fresh
  34. * runtime so budgets can be tuned per case.
  35. */
  36. async function setup(config: Config = {}) {
  37. const ctx = new Context()
  38. const fiber = await ctx.plugin(PythonCodeRuntime, config)
  39. const runtime = ctx.codeRuntime as PythonCodeRuntime
  40. return { ctx, fiber, runtime }
  41. }
  42. /** Convenience: one namespace `tools` with the given functions. */
  43. function tools(functions: Record<string, CodeBindingFunction>) {
  44. return [{ global: 'tools', functions }]
  45. }
  46. describe('PythonCodeRuntime — seam descriptors and misuse', () => {
  47. it('registers the seam descriptors', async () => {
  48. const { runtime } = await setup()
  49. expect(runtime.language).toBe('python')
  50. expect(runtime.isolation).toBe('process')
  51. })
  52. it('rejects non-positive config as seam misuse', async () => {
  53. const ctx = new Context()
  54. await expect(ctx.plugin(PythonCodeRuntime, { cpuSeconds: 0 }))
  55. .rejects.toThrow(/cpuSeconds must be a positive number/)
  56. await expect(ctx.plugin(PythonCodeRuntime, { maxWallMs: -1 }))
  57. .rejects.toThrow(/maxWallMs must be a positive number/)
  58. })
  59. it('rejects a non-integer cpuSeconds at load (setrlimit needs an int)', async () => {
  60. const ctx = new Context()
  61. await expect(ctx.plugin(PythonCodeRuntime, { cpuSeconds: 1.5 }))
  62. .rejects.toThrow(/cpuSeconds must be a positive integer, got 1.5/)
  63. })
  64. it('rejects a non-integer byte budget at load (the child int()-truncates it)', async () => {
  65. // maxLogBytes/maxValueBytes cross to the child, which reads them through
  66. // int(...): a float would floor there while the host meters the fraction, so
  67. // the two sides would enforce different public config. Reject at load.
  68. const ctxLog = new Context()
  69. await expect(ctxLog.plugin(PythonCodeRuntime, { maxLogBytes: 3.5 }))
  70. .rejects.toThrow(/maxLogBytes must be a positive integer/)
  71. const ctxValue = new Context()
  72. await expect(ctxValue.plugin(PythonCodeRuntime, { maxValueBytes: 1024.5 }))
  73. .rejects.toThrow(/maxValueBytes must be a positive integer/)
  74. })
  75. it('rejects finite numeric config that cannot cross as an exact rlimit integer', async () => {
  76. // `Number.isFinite` and `Number.isInteger` both admit values that cannot
  77. // round-trip. `addressSpaceMb: 1e308` overflows to `Infinity` once multiplied
  78. // by 1 MiB, and `encodeJsonPlain` renders that as `null`, so the child gets no
  79. // limit at all; `cpuSeconds: 1e100` clears `Number.isInteger` while sitting
  80. // far past the safe range, so `setrlimit` receives a different number than was
  81. // configured. Both used to end every run in a bootstrap exception instead of
  82. // failing at load, where a self-contained configuration error belongs.
  83. const ctx = new Context()
  84. await expect(ctx.plugin(PythonCodeRuntime, { addressSpaceMb: 1e308 }))
  85. .rejects.toThrow(/addressSpaceMb must be at most \d+ .*exact integer/)
  86. await expect(ctx.plugin(PythonCodeRuntime, { cpuSeconds: 1e100 }))
  87. .rejects.toThrow(/cpuSeconds must be at most \d+ .*exact integers/)
  88. // The boundary values still load: the bound rejects what cannot be encoded,
  89. // not everything large.
  90. const okMb = await ctx.plugin(PythonCodeRuntime, { addressSpaceMb: Math.floor(Number.MAX_SAFE_INTEGER / (1024 * 1024)) })
  91. await okMb.dispose()
  92. const okCpu = await ctx.plugin(PythonCodeRuntime, { cpuSeconds: Number.MAX_SAFE_INTEGER - 1 })
  93. await okCpu.dispose()
  94. })
  95. it('rejects an output cap whose payload could not cross the frame ceiling', async () => {
  96. // The caps budget a payload that must arrive inside ONE fd-3 frame, and the
  97. // 64 MiB frame parse cap is fixed. A larger cap is unsatisfiable rather
  98. // than generous: a completion the cap admits arrives as an over-ceiling
  99. // frame and fails the run as `worker-exit`, inverting the `output-limit`
  100. // the cap describes. Both budgets are metered in already-escaped serialized
  101. // bytes, so a payload occupies at most `cap + envelope` on the wire; the
  102. // bound is `parse-cap - envelope`, not `(ceiling - envelope) / 6` (that
  103. // divided in escape expansion the charge already counts). The receive path
  104. // rejects raw frames past the 64 MiB parse cap (the run settles as a
  105. // worker-exit), so a budget above it would admit a config whose honest
  106. // child frames the host then rejects.
  107. const admissible = 64 * 1024 * 1024 - 64
  108. const ctx = new Context()
  109. await expect(ctx.plugin(PythonCodeRuntime, { maxLogBytes: admissible + 1 }))
  110. .rejects.toThrow(/maxLogBytes must not exceed 67108800/)
  111. await expect(ctx.plugin(PythonCodeRuntime, { maxValueBytes: admissible + 1 }))
  112. .rejects.toThrow(/maxValueBytes must not exceed 67108800/)
  113. // The boundary value itself loads: the bound is the largest cap a frame can
  114. // still carry, not one below it. It needs an address space large enough to
  115. // clear the separate maxValueBytes/addressSpaceMb worst-case gate (the cap
  116. // times the 12x Unicode expansion must fit), so this pairs it with a 4 GiB
  117. // addressSpaceMb — the two load-time bounds are independent.
  118. const boundary = await ctx.plugin(PythonCodeRuntime, { maxValueBytes: admissible, addressSpaceMb: 4096 })
  119. await boundary.dispose()
  120. })
  121. it('rejects a pythonBin that spawn() would throw on, at load', async () => {
  122. // Both values pass the string schema and both make `spawn` throw
  123. // SYNCHRONOUSLY from inside run() — ERR_INVALID_ARG_VALUE for the empty
  124. // path, ERR_INVALID_ARG_TYPE for the NUL — so run() would REJECT instead of
  125. // resolving the worker-exit the seam promises for a child that cannot
  126. // start. Both are self-contained configuration errors, so they fail here.
  127. const ctx = new Context()
  128. await expect(ctx.plugin(PythonCodeRuntime, { pythonBin: '' }))
  129. .rejects.toThrow(/pythonBin must be a non-empty path without NUL bytes/)
  130. await expect(ctx.plugin(PythonCodeRuntime, { pythonBin: 'py\u0000thon3' }))
  131. .rejects.toThrow(/pythonBin must be a non-empty path without NUL bytes/)
  132. })
  133. it('skips relative PATH entries when resolving a basename pythonBin', async () => {
  134. // resolvePythonBin must return an absolute path: a RELATIVE PATH entry
  135. // ('.' here) would otherwise resolve the basename against the host CWD.
  136. // This run's CWD holds no executable named python3, so both the relative
  137. // skip and the accessSync-miss fall through to the absolute entry — the
  138. // case pins the contract (absolute candidate wins over a relative PATH
  139. // prefix), not a worker-exit distinction, which would need an executable
  140. // named python3 in the test CWD.
  141. const cp = await import('node:child_process')
  142. const nodePath = await import('node:path')
  143. const pythonDir = nodePath.dirname(cp.execFileSync('which', ['python3'], { encoding: 'utf8' }).trim())
  144. vi.stubEnv('PATH', `.:${pythonDir}`)
  145. try {
  146. const { runtime, fiber } = await setup({ pythonBin: 'python3', maxWallMs: 30_000 })
  147. const result = await runtime.run({ program: 'return 1', bindings: [] })
  148. expect(result.error).toBeUndefined()
  149. expect(result.value).toBe(1)
  150. await fiber.dispose()
  151. } finally {
  152. vi.unstubAllEnvs()
  153. }
  154. }, 45_000)
  155. it('ignores a forged second boot-ack without re-sending the run frame', async () => {
  156. // The run frame is sent once, from the first boot-ack; a program that
  157. // forges an extra boot-ack frame on fd 3 must not re-enter the gate (a
  158. // second run frame would confuse the child's frame reader). The honest
  159. // child sends exactly one ack; the forged one exercises the re-entry
  160. // guard.
  161. const { runtime } = await setup()
  162. const result = await runtime.run({
  163. program: [
  164. 'import os',
  165. // One forged boot-ack after the program starts; the run already went
  166. // out on the real ack.
  167. "os.write(3, b'{\"type\":\"boot-ack\"}\\n')",
  168. 'return "done"',
  169. ].join('\n'),
  170. bindings: [],
  171. })
  172. expect(result.error).toBeUndefined()
  173. expect(result.value).toBe('done')
  174. }, 15_000)
  175. it('skips a PATH entry that is an executable DIRECTORY named like the interpreter', async () => {
  176. // accessSync(X_OK) succeeds on directories, so without the isFile guard a
  177. // PATH entry like a `python3` directory would be chosen over a later real
  178. // interpreter. The stub PATH puts such a directory first and asserts the
  179. // real interpreter is used.
  180. const cp = await import('node:child_process')
  181. const nodePath = await import('node:path')
  182. const { mkdtempSync, mkdirSync } = await import('node:fs')
  183. const { tmpdir } = await import('node:os')
  184. const realPythonDir = nodePath.dirname(cp.execFileSync('which', ['python3'], { encoding: 'utf8' }).trim())
  185. const fakeDir = mkdtempSync(nodePath.join(tmpdir(), 'dsh-fake-bin-'))
  186. mkdirSync(nodePath.join(fakeDir, 'python3')) // A directory named python3, executable by default.
  187. vi.stubEnv('PATH', `${fakeDir}:${realPythonDir}`)
  188. try {
  189. const { runtime, fiber } = await setup({ pythonBin: 'python3', maxWallMs: 30_000 })
  190. const result = await runtime.run({ program: 'return 1', bindings: [] })
  191. expect(result.error).toBeUndefined()
  192. expect(result.value).toBe(1)
  193. await fiber.dispose()
  194. } finally {
  195. vi.unstubAllEnvs()
  196. }
  197. }, 45_000)
  198. it('rejects a timer budget setTimeout would silently clamp to 1 ms', async () => {
  199. // Node stores a setTimeout delay as a signed 32-bit value and substitutes
  200. // 1 ms for anything larger, inverting the knob's meaning: a huge maxWallMs
  201. // would time every run out at once, and a huge graceMs would SIGKILL one
  202. // millisecond after SIGTERM. Both must fail at load instead.
  203. const ctx = new Context()
  204. await expect(ctx.plugin(PythonCodeRuntime, { maxWallMs: 2_147_483_648 }))
  205. .rejects.toThrow(/maxWallMs must not exceed 2147483647/)
  206. // graceMs is bounded by the close deadline's added margin, not by the raw
  207. // timer maximum, because that sum is what gets armed.
  208. await expect(ctx.plugin(PythonCodeRuntime, { graceMs: 2_147_481_648 }))
  209. .rejects.toThrow(/graceMs must not exceed 2147481647/)
  210. // The exact maxima still load.
  211. await expect(ctx.plugin(PythonCodeRuntime, { maxWallMs: 2_147_483_647, graceMs: 2_147_481_647 }))
  212. .resolves.toBeDefined()
  213. })
  214. it('rejects loading this Unix-only backend on Windows', async () => {
  215. // The bootstrap needs the POSIX `resource` module, a positional fd 3, and
  216. // negative-PID process-group signals — none on Windows. The constructor
  217. // must throw at load rather than register ctx.codeRuntime and defer the
  218. // failure to the first run.
  219. const original = process.platform
  220. Object.defineProperty(process, 'platform', { value: 'win32', configurable: true })
  221. try {
  222. const ctx = new Context()
  223. await expect(ctx.plugin(PythonCodeRuntime, {})).rejects.toThrow(/requires a Unix platform/)
  224. } finally {
  225. Object.defineProperty(process, 'platform', { value: original, configurable: true })
  226. }
  227. })
  228. it('rejects a binding global that is not a Python identifier or is reserved', async () => {
  229. const { runtime } = await setup()
  230. await expect(runtime.run({
  231. program: 'return 1',
  232. bindings: [{ global: '1bad', functions: {} }],
  233. })).rejects.toThrow(/is not a usable Python identifier/)
  234. await expect(runtime.run({
  235. program: 'return 1',
  236. bindings: [{ global: 'class', functions: {} }],
  237. })).rejects.toThrow(/is not a usable Python identifier/)
  238. })
  239. it('rejects duplicate binding namespaces', async () => {
  240. const { runtime } = await setup()
  241. await expect(runtime.run({
  242. program: 'return 1',
  243. bindings: [
  244. { global: 'tools', functions: {} },
  245. { global: 'tools', functions: {} },
  246. ],
  247. })).rejects.toThrow(/duplicate binding global/)
  248. })
  249. it('rejects run() after disposal, and unregisters ctx.codeRuntime', async () => {
  250. const { ctx, fiber, runtime } = await setup()
  251. await fiber.dispose()
  252. await expect(runtime.run({ program: 'return 1', bindings: [] }))
  253. .rejects.toThrow(/after disposal/)
  254. expect(ctx.get('codeRuntime')).toBeUndefined()
  255. })
  256. it('short-circuits when the request signal is already aborted', async () => {
  257. const { runtime } = await setup()
  258. const signal = AbortSignal.abort('already-cancelled')
  259. const result = await runtime.run({ program: 'return 1', bindings: [], signal })
  260. expect(result.error?.kind).toBe('abort')
  261. expect(result.error?.message).toContain('already-cancelled')
  262. expect(result.logs).toEqual([])
  263. })
  264. it('short-circuits on an already-aborted signal whose reason cannot be converted', async () => {
  265. // The pre-flight arm converted the reason with a bare `String()`, so a
  266. // hostile reason threw out of `run()` — the seam promises to reject only for
  267. // misuse, and a caller's cancellation token is not misuse.
  268. const { runtime } = await setup()
  269. const signal = AbortSignal.abort({
  270. [Symbol.toPrimitive]() { throw new Error('reason blew up') },
  271. })
  272. const result = await runtime.run({ program: 'return 1', bindings: [], signal })
  273. expect(result.error?.kind).toBe('abort')
  274. expect(result.error?.message).toBe('<unrenderable rejection value>')
  275. expect(result.logs).toEqual([])
  276. })
  277. it('runs the interpreter from materialized scripts outside the package, and removes them per run', async () => {
  278. // The interpreter is an EXTERNAL process, so it can only open paths the OS
  279. // resolves. Inside the single-file Python-SDK executable the packaged `py/`
  280. // directory lives in pkg's virtual filesystem, which Node reads through its
  281. // patched `fs` but `python3` cannot see, so spawning from that path fails
  282. // with ENOENT. The scripts are therefore copied to a real directory first.
  283. //
  284. // The path is read from the child's own `__main__` module, so it proves
  285. // where the interpreter actually loaded the entry script — asserting on a
  286. // host-side constant would only restate the source. The program namespace
  287. // seeds `__name__` but no `__file__`, hence the module lookup.
  288. // `protocol.py` must land in the SAME directory, since `bootstrap.py` puts
  289. // its own directory on `sys.path` to import it; the run completing at all
  290. // already exercises that import.
  291. const { runtime } = await setup()
  292. const entryOf = async (): Promise<string> => {
  293. const result = await runtime.run({ program: 'import sys\nreturn sys.modules["__main__"].__file__', bindings: [] })
  294. expect(result.error).toBeUndefined()
  295. return result.value as string
  296. }
  297. const entry = await entryOf()
  298. expect(entry.endsWith('/bootstrap.py')).toBe(true)
  299. const dir = dirname(entry)
  300. expect(dir.startsWith(realpathSync(tmpdir()))).toBe(true)
  301. expect(dir).not.toContain('/packages/')
  302. // Staging is per RUN and removed at settlement, so by the time `run()`
  303. // resolved the directory is already gone — nothing survives to be rewritten
  304. // by a later run. `protocol.py` had to be beside the entry script for the run
  305. // to complete at all, since `bootstrap.py` imports it off `sys.path`.
  306. expect(existsSync(dir)).toBe(false)
  307. // A second run stages its own copy rather than reusing the first.
  308. expect(dirname(await entryOf())).not.toBe(dir)
  309. })
  310. it('contains a program that rewrites its own bootstrap to the run that did it', async () => {
  311. // The child runs as the same UID as the host, so `0o700` does not stop model
  312. // code from rewriting the scripts it was started from —
  313. // `sys.modules['__main__'].__file__` names them. While all runs shared one
  314. // staged copy, a program that overwrote `bootstrap.py` broke the NEXT run
  315. // (measured: it settled as `worker-exit`), and substituted code would have
  316. // run before the resource limits were applied.
  317. const { runtime } = await setup({ maxWallMs: 10_000 })
  318. const sabotage = await runtime.run({
  319. program: [
  320. 'import sys',
  321. 'path = sys.modules["__main__"].__file__',
  322. 'open(path, "w").write("raise SystemExit(1)\\n")',
  323. 'return path',
  324. ].join('\n'),
  325. bindings: [],
  326. })
  327. expect(sabotage.error).toBeUndefined()
  328. // The damage stayed inside the run that caused it.
  329. const after = await runtime.run({ program: 'return 1 + 1', bindings: [] })
  330. expect(after.error).toBeUndefined()
  331. expect(after.value).toBe(2)
  332. }, 20_000)
  333. it('leaves no subprocess or scripts behind when disposal races the first run', async () => {
  334. // Staging runs SYNCHRONOUSLY so no async boundary opens between `run()` and
  335. // the point where `execute` registers the run in `live` and installs the
  336. // abort listener. With an `await` there, a disposal landing in that window
  337. // saw an empty `live`, returned, removed the script directory, and let the
  338. // continuation spawn a subprocess after the fiber was gone.
  339. //
  340. // `dispose()` is called in the same synchronous turn as `run()`, with no
  341. // `await` between them, so it lands exactly in that window.
  342. //
  343. // The leak assertion compares before and after rather than requiring an
  344. // empty tmpdir: other tests in this file build runtimes they never dispose,
  345. // so only the directories this test adds are its own evidence.
  346. const staged = (): string[] =>
  347. readdirSync(realpathSync(tmpdir())).filter(name => name.startsWith('dsh-code-runtime-python-'))
  348. const before = new Set(staged())
  349. const { fiber, runtime } = await setup({ maxWallMs: 8_000 })
  350. const pending = runtime.run({ program: 'import time\nwhile True: time.sleep(0.1)', bindings: [] })
  351. const disposed = fiber.dispose()
  352. const result = await pending
  353. await disposed
  354. // Whatever the run reports, it must be terminal and must not be a success.
  355. expect(result.value).toBeUndefined()
  356. expect(['abort', 'worker-exit', 'timeout']).toContain(result.error?.kind)
  357. // Disposal is to quiescence, so this run's directory is gone once it
  358. // resolves, and nothing recreated it afterwards.
  359. expect(staged().filter(name => !before.has(name))).toEqual([])
  360. }, 15_000)
  361. it('settles as abort when the signal fires in the same turn as the first run', async () => {
  362. // Same window, the other listener. `addEventListener('abort')` does not
  363. // replay an event that already fired, so an abort landing before the
  364. // listener was installed used to be missed entirely and the program ran to
  365. // success or the wall ceiling instead of resolving as `abort`. Synchronous
  366. // staging keeps the pre-flight check and the listener in one turn, leaving
  367. // no gap for the signal to slip through.
  368. const { runtime } = await setup({ maxWallMs: 4_000, graceMs: 200 })
  369. const controller = new AbortController()
  370. const pending = runtime.run({
  371. program: 'import time\nwhile True: time.sleep(0.1)',
  372. bindings: [],
  373. signal: controller.signal,
  374. })
  375. controller.abort('same-turn-abort')
  376. const result = await pending
  377. expect(result.error?.kind).toBe('abort')
  378. expect(result.error?.message).toContain('same-turn-abort')
  379. }, 15_000)
  380. it('reports a staging failure as worker-exit instead of rejecting run()', async () => {
  381. // Staging touches the filesystem, so it can fail for reasons that are not
  382. // the caller's doing: a full or read-only temp filesystem, or a deployment
  383. // that failed to ship the packaged scripts. Those are SUBSTRATE failures,
  384. // the same class as a child that cannot start, and the seam reserves
  385. // rejection for misuse — so `run()` must resolve, not throw.
  386. //
  387. // `TMPDIR` is the honest lever: `mkdtempSync` builds its path from
  388. // `os.tmpdir()`, so pointing it at a path that is not a directory makes the
  389. // real call fail without stubbing the module under test.
  390. const previous = process.env.TMPDIR
  391. const notADirectory = join(await mkdtemp(join(tmpdir(), 'dsh-staging-')), 'file')
  392. await writeFile(notADirectory, '')
  393. process.env.TMPDIR = notADirectory
  394. try {
  395. const { runtime } = await setup()
  396. const result = await runtime.run({ program: 'return 1', bindings: [] })
  397. expect(result.error?.kind).toBe('worker-exit')
  398. expect(result.error?.message).toContain('failed to stage the python bootstrap')
  399. expect(result.logs).toEqual([])
  400. } finally {
  401. if (previous === undefined) delete process.env.TMPDIR
  402. else process.env.TMPDIR = previous
  403. }
  404. })
  405. it('leaves no staging directory behind when a script copy fails', async () => {
  406. // `mkdtempSync` succeeding and a later `copyFileSync` failing is its own
  407. // case: the directory exists but is only partially populated. Recording it
  408. // before the copies would leak it, because `run` retries staging on the next
  409. // call and overwrites the single recorded path — teardown could then remove
  410. // only the newest attempt. Staging must clean up its own partial directory.
  411. //
  412. // Only `copyFileSync` is stubbed, and only for the second script, so
  413. // `mkdtempSync` really runs and the directory under assertion is real.
  414. const staged = (): string[] =>
  415. readdirSync(realpathSync(tmpdir())).filter(name => name.startsWith('dsh-code-runtime-python-'))
  416. const before = new Set(staged())
  417. failNextCopyOf.value = 'protocol.py'
  418. try {
  419. const { runtime } = await setup()
  420. const result = await runtime.run({ program: 'return 1', bindings: [] })
  421. expect(result.error?.kind).toBe('worker-exit')
  422. expect(result.error?.message).toContain('failed to stage the python bootstrap')
  423. // The partial directory is gone, so nothing accumulates across retries.
  424. expect(staged().filter(name => !before.has(name))).toEqual([])
  425. } finally {
  426. failNextCopyOf.value = undefined
  427. }
  428. }, 15_000)
  429. })
  430. describe('PythonCodeRuntime — process identity', () => {
  431. it('reads a live process start time and distinguishes it from an absent pid', () => {
  432. // The teardown guard signals `-child.pid` with a RAW `process.kill`, which
  433. // (unlike `child.kill()`) has no handle check, so it would reach a recycled
  434. // pgid during the window between the leader being reaped and `close` firing.
  435. // A pid alone cannot separate the original from its replacement -- both
  436. // answer `kill(pid, 0)` -- so the guard compares START TIME, and this pins
  437. // that the reading is stable for one process and absent for a pid that
  438. // cannot be read.
  439. const own = readProcessStart(process.pid)
  440. if (process.platform === 'linux') {
  441. // Same process, two reads: the identity must be stable, or the guard would
  442. // refuse to signal its own live group.
  443. expect(own).toBeDefined()
  444. expect(readProcessStart(process.pid)).toBe(own)
  445. // Pid 0 is never a readable /proc entry, so the guard degrades to
  446. // undefined rather than throwing on a teardown path. This is also the
  447. // reading a REAPED leader produces -- its /proc entry is gone while the
  448. // group it led can still hold survivors -- so `undefined` must NOT be
  449. // treated as an identity mismatch. Reading it as one refused the SIGKILL
  450. // that the same-group survivor tests depend on, which is why they went red
  451. // on Linux while passing on Darwin (where the reader always returns
  452. // undefined and the guard is inert).
  453. expect(readProcessStart(0)).toBeUndefined()
  454. } else {
  455. // Darwin has no /proc: the reader reports undefined, and `killGroup`
  456. // signals the pgid without the identity re-check instead of paying a `ps`
  457. // fork per signal.
  458. expect(own).toBeUndefined()
  459. }
  460. })
  461. })
  462. describe('PythonCodeRuntime — inherited resource limits', () => {
  463. it('runs under an inherited hard limit tighter than addressSpaceMb', async () => {
  464. // An unprivileged process may lower a hard rlimit but never raise it. Under
  465. // a harness started with `ulimit -v` below `addressSpaceBytes`, requesting
  466. // the configured cap made `setrlimit` raise `ValueError` and every run
  467. // returned a bootstrap exception — even though the inherited limit is
  468. // STRONGER than the one asked for. The bootstrap clamps to the inherited
  469. // hard limit instead, so the run proceeds under the stricter bound.
  470. //
  471. // `pythonBin` is the honest lever: a wrapper that lowers RLIMIT_AS and then
  472. // execs the real interpreter reproduces the inherited-limit condition
  473. // without touching this test process's own limits.
  474. const dir = await mkdtemp(join(tmpdir(), 'dsh-rlimit-'))
  475. const wrapper = join(dir, 'python3-capped')
  476. // 256 MiB, half the 512 MiB addressSpaceMb default, so the requested cap is
  477. // unambiguously above the inherited ceiling.
  478. await writeFile(wrapper, '#!/bin/sh\nulimit -v 262144\nexec python3 "$@"\n', { mode: 0o755 })
  479. const { runtime } = await setup({ pythonBin: wrapper })
  480. const result = await runtime.run({
  481. program: 'import resource\nreturn resource.getrlimit(resource.RLIMIT_AS)[1]',
  482. bindings: [],
  483. })
  484. expect(result.error).toBeUndefined()
  485. // The applied hard limit is the inherited one, not the configured 512 MiB.
  486. expect(result.value).toBe(256 * 1024 * 1024)
  487. }, 15_000)
  488. it('rejects at boot when an inherited RLIMIT_AS is too tight for the output budgets', async () => {
  489. // The host gate validates the output budgets against the CONFIGURED
  490. // addressSpaceMb, but a launch environment can inherit a STRICTER RLIMIT_AS
  491. // (a `ulimit -v` wrapper below addressSpaceMb), which the bootstrap clamps the
  492. // effective limit down to — leaving the budgets sized for a ceiling the child
  493. // never gets, so a near-budget output would OOM mid-run as an opaque
  494. // worker-exit. The bootstrap re-checks both budgets against the EFFECTIVE
  495. // clamped limit and fails loud at boot instead. A 128 MiB inherited limit
  496. // leaves 64 MiB budgetable (~5 MiB admissible under the 12x multiple), under
  497. // which a 32 MiB maxLogBytes — admitted by the 512 MiB configured default — is
  498. // rejected. The rejection surfaces as an 'exception' (bootstrap's
  499. // setrlimit-phase failure class), not a mid-run OOM. The repro is Linux-only
  500. // (macOS ignores `ulimit -v`); there the run proceeds.
  501. const dir = await mkdtemp(join(tmpdir(), 'dsh-rlimit-'))
  502. const wrapper = join(dir, 'python3-tight')
  503. await writeFile(wrapper, '#!/bin/sh\nulimit -v 131072\nexec python3 "$@"\n', { mode: 0o755 })
  504. const { runtime } = await setup({ pythonBin: wrapper, maxLogBytes: 32 * 1024 * 1024, addressSpaceMb: 512 })
  505. const result = await runtime.run({ program: 'return 1', bindings: [] })
  506. if (process.platform === 'darwin') {
  507. expect(result.error).toBeUndefined()
  508. } else {
  509. // The re-check raises inside bootstrap's resource-limit block, which
  510. // reports every setrlimit-phase failure as kind 'exception'; the message
  511. // discriminates this config rejection from a generic setrlimit error.
  512. expect(result.error?.kind).toBe('exception')
  513. expect(result.error?.message).toContain('too large for the inherited RLIMIT_AS')
  514. }
  515. }, 15_000)
  516. it('applies the configured limits when nothing tighter is inherited', async () => {
  517. // The clamp must not weaken the normal path: with an infinite inherited hard
  518. // limit there is nothing to clamp against, and RLIM_INFINITY compares as -1,
  519. // so treating it as a numeric bound would collapse every limit to -1.
  520. const { runtime } = await setup({ cpuSeconds: 42, addressSpaceMb: 400 })
  521. const result = await runtime.run({
  522. // `getrlimit` returns a tuple, which the lossless-JSON completion check
  523. // rejects; the pair is listed explicitly rather than converted.
  524. program: 'import resource\ncpu = resource.getrlimit(resource.RLIMIT_CPU)\nreturn [cpu[0], cpu[1], resource.getrlimit(resource.RLIMIT_AS)[1]]',
  525. bindings: [],
  526. })
  527. expect(result.error).toBeUndefined()
  528. // Soft at cpuSeconds, hard at +1 (the SIGKILL backstop), address space at
  529. // the configured megabytes — exactly what the unclamped path applied.
  530. expect(result.value).toEqual([42, 43, 400 * 1024 * 1024])
  531. }, 15_000)
  532. it('preserves an inherited soft limit stricter than the configured cap', async () => {
  533. // Clamping reads BOTH inherited bounds, not just the hard one. A deployment
  534. // that inherited a soft rlimit below the configured cap must keep that
  535. // stricter soft: returning the configured value would RAISE the effective
  536. // soft limit, loosening containment. The wrapper lowers only the SOFT CPU
  537. // limit (`ulimit -S -t`) and leaves the hard limit unlimited, so the
  538. // requested soft (`cpuSeconds`) sits above the inherited soft — the case that
  539. // exposed the bug. RLIMIT_CPU is used because macOS ignores `ulimit -v`
  540. // (RLIMIT_AS), which is exactly why the backend skips address space there.
  541. const dir = await mkdtemp(join(tmpdir(), 'dsh-rlimit-soft-'))
  542. const wrapper = join(dir, 'python3-soft-capped')
  543. // Soft CPU 5 s, well below the configured 30 s, hard left unlimited.
  544. await writeFile(wrapper, '#!/bin/sh\nulimit -S -t 5\nexec python3 "$@"\n', { mode: 0o755 })
  545. const { runtime } = await setup({ pythonBin: wrapper, cpuSeconds: 30 })
  546. const result = await runtime.run({
  547. program: 'import resource\nreturn resource.getrlimit(resource.RLIMIT_CPU)[0]',
  548. bindings: [],
  549. })
  550. expect(result.error).toBeUndefined()
  551. // The applied SOFT limit is the inherited 5 s, not the configured 30 s.
  552. expect(result.value).toBe(5)
  553. }, 15_000)
  554. it('reports a CPU overrun under a dual-limit ulimit as a timeout, not a worker-exit', async () => {
  555. // `ulimit -t N` sets BOTH the soft and hard CPU limit to N. The kernel
  556. // checks the hard limit and SIGKILLs a busy loop directly, so with
  557. // soft == hard the SIGXCPU signal is never delivered — and the host
  558. // classifies a CPU overrun ONLY on `signal === 'SIGXCPU'`, so the overrun
  559. // would be misreported as a `worker-exit` instead of a timeout. `_clamped`
  560. // now lowers a clamped soft==hard result by one unit (when hard >= 2), so
  561. // the SIGXCPU signal fires at the softer limit and the run reports a
  562. // timeout. This uses `ulimit -t 2` (hard == 2, so the soft is lowered to 1)
  563. // and leaves SIGXCPU unhandled, so the kernel terminates the busy loop at
  564. // 1 s with SIGXCPU and the host classifies it as a timeout.
  565. const dir = await mkdtemp(join(tmpdir(), 'dsh-rlimit-dual-'))
  566. const wrapper = join(dir, 'python3-dual-capped')
  567. // Both soft and hard CPU 2 s; configured cpuSeconds 30 s.
  568. await writeFile(wrapper, '#!/bin/sh\nulimit -t 2\nexec python3 "$@"\n', { mode: 0o755 })
  569. const { runtime } = await setup({ pythonBin: wrapper, cpuSeconds: 30, maxWallMs: 12_000 })
  570. const result = await runtime.run({
  571. program: [
  572. 'while True:',
  573. ' pass',
  574. 'return "unreachable"',
  575. ].join('\n'),
  576. bindings: [],
  577. })
  578. expect(result.error?.kind).toBe('timeout')
  579. expect(result.error?.message).toContain('CPU time exhausted')
  580. }, 15_000)
  581. it('reports a timeout when a program masks SIGXCPU and returns past the soft limit', async () => {
  582. // A program can mask SIGXCPU (pthread_sigmask SIG_BLOCK), burn past the
  583. // soft CPU limit, and return during the soft-to-hard gap. The settlement
  584. // recheck (`die_if_cpu_exhausted`) must UNBLOCK the signal before
  585. // re-delivering it, or the SIGXCPU stays pending and the child exits
  586. // normally with a success result. With the unblock, the re-delivered
  587. // SIGXCPU (default disposition) terminates the child and the host
  588. // classifies the run as a timeout. Fail-before: without the unblock the
  589. // run reports `value: "escaped"` and no error. The masking is guarded by
  590. // hasattr so the case is a no-op on platforms without pthread_sigmask.
  591. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 12_000 })
  592. const result = await runtime.run({
  593. program: [
  594. 'import signal, time',
  595. 'if hasattr(signal, "pthread_sigmask"):',
  596. ' signal.pthread_sigmask(signal.SIG_BLOCK, {signal.SIGXCPU})',
  597. 'end = time.process_time() + 1.05',
  598. 'while time.process_time() < end:',
  599. ' pass',
  600. 'return "escaped"',
  601. ].join('\n'),
  602. bindings: [],
  603. })
  604. expect(result.error?.kind).toBe('timeout')
  605. expect(result.value).toBeUndefined()
  606. }, 20_000)
  607. it('reports a timeout when a program traps AND masks SIGXCPU and returns past the soft limit', async () => {
  608. // The mask-only case exercises the unblock; the trap+mask combination is
  609. // the harder one: a program that installed a custom handler AND masked the
  610. // signal has that PENDING handler run the moment the signal is unblocked
  611. // (CPython delivers it at the next eval-breaker checkpoint in model code),
  612. // and the handler re-masks — so the settlement recheck must restore the
  613. // default disposition BEFORE unblocking. With SIG_DFL restored first, the
  614. // pending signal kills the process inside the kernel with no bytecode
  615. // window; without it, the handler re-blocks and the child exits normally
  616. // with a success value. Fail-before: the run reports `value: "escaped"`.
  617. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 12_000 })
  618. const result = await runtime.run({
  619. program: [
  620. 'import signal, time',
  621. 'if hasattr(signal, "pthread_sigmask"):',
  622. ' def h(signum, frame):',
  623. ' signal.pthread_sigmask(signal.SIG_BLOCK, {signal.SIGXCPU})',
  624. ' signal.signal(signal.SIGXCPU, h)',
  625. ' signal.pthread_sigmask(signal.SIG_BLOCK, {signal.SIGXCPU})',
  626. ' end = time.process_time() + 1.05',
  627. ' while time.process_time() < end:',
  628. ' pass',
  629. 'return "escaped"',
  630. ].join('\n'),
  631. bindings: [],
  632. })
  633. expect(result.error?.kind).toBe('timeout')
  634. expect(result.value).toBeUndefined()
  635. }, 20_000)
  636. it('rechecks CPU at settlement against the effective inherited soft limit', async () => {
  637. // The settlement-time CPU recheck must compare against the EFFECTIVE soft
  638. // limit (`_clamped` may have lowered it to a stricter inherited value), not
  639. // the configured `cpuSeconds`. A program that traps SIGXCPU, burns past the
  640. // inherited soft, and returns inside the soft-to-hard gap would otherwise be
  641. // compared to the configured value and falsely reported successful, bypassing
  642. // the inherited limit. The wrapper sets a 1 s soft CPU limit; the program
  643. // traps SIGXCPU and busy-loops past it, then returns — the recheck must
  644. // re-deliver SIGXCPU so the host classifies the run as a timeout.
  645. const dir = await mkdtemp(join(tmpdir(), 'dsh-cpu-recheck-'))
  646. const wrapper = join(dir, 'python3-cpu-capped')
  647. await writeFile(wrapper, '#!/bin/sh\nulimit -S -t 1\nexec python3 "$@"\n', { mode: 0o755 })
  648. const { runtime } = await setup({ pythonBin: wrapper, cpuSeconds: 30, maxWallMs: 12_000 })
  649. const result = await runtime.run({
  650. program: [
  651. 'import signal, time',
  652. // Trap SIGXCPU so the soft limit does not terminate the program; burn
  653. // CPU well past the inherited 1 s soft, then return normally.
  654. 'signal.signal(signal.SIGXCPU, lambda *a: None)',
  655. 'end = time.process_time() + 2.5',
  656. 'while time.process_time() < end:',
  657. ' pass',
  658. 'return "returned"',
  659. ].join('\n'),
  660. bindings: [],
  661. })
  662. // The recheck compares spent CPU against the effective 1 s soft, not 30 s, so
  663. // the run is a timeout rather than a false success.
  664. expect(result.error?.kind).toBe('timeout')
  665. }, 20_000)
  666. })
  667. describe('PythonCodeRuntime — programs and bindings', () => {
  668. it('runs a top-level script, captures print output, and returns `result`', async () => {
  669. const { runtime } = await setup()
  670. const result = await runtime.run({
  671. program: [
  672. 'x = 40',
  673. 'y = 2',
  674. 'print("hello", x + y)',
  675. 'return {"answer": x + y}',
  676. ].join('\n'),
  677. bindings: [],
  678. })
  679. expect(result.error).toBeUndefined()
  680. expect(result.value).toEqual({ answer: 42 })
  681. // `print` in Python emits: text, ' ', text, '\n'. Concat the captured
  682. // fragments and assert the model-visible message survives.
  683. expect(result.logs.join('')).toContain('hello 42')
  684. // 15s: this is usually the suite's first real subprocess — a cold python3
  685. // start (interpreter + asyncio import) on a loaded CI runner can exceed
  686. // the 5s default alone; later tests reuse the warm page cache.
  687. }, 15_000)
  688. it('bridges binding calls both ways and rejects the program-side call on a host rejection', async () => {
  689. const { runtime } = await setup()
  690. const calls: unknown[] = []
  691. const result = await runtime.run({
  692. program: [
  693. 'first = await tools.echo({"n": 1})',
  694. 'caught = ""',
  695. 'try:',
  696. ' await tools.fail({})',
  697. 'except RuntimeError as e:',
  698. ' caught = str(e)',
  699. 'return {"first": first, "caught": caught}',
  700. ].join('\n'),
  701. bindings: tools({
  702. echo: async (args) => { calls.push(args); return { echoed: args as CodeJsonValue } },
  703. fail: async () => { throw new Error('nope') },
  704. }),
  705. })
  706. expect(result.error).toBeUndefined()
  707. expect(result.value).toEqual({ first: { echoed: { n: 1 } }, caught: 'nope' })
  708. expect(calls).toEqual([{ n: 1 }])
  709. })
  710. it('keeps decoding binding replies when _decode_json_plain is rebound', async () => {
  711. // read_frame_async resolves _decode_json_plain at call time; a program that
  712. // rebinds __main__._decode_json_plain would otherwise kill the reply pump
  713. // (a broken decode strands every pending Future to the wall clock). The
  714. // decode primitives are def-time captures on the channel methods, so a
  715. // rebind cannot break reply delivery.
  716. const { runtime } = await setup()
  717. const result = await runtime.run({
  718. program: [
  719. 'import __main__',
  720. '__main__._decode_json_plain = None',
  721. 'first = await tools.echo({"n": 1})',
  722. 'return first',
  723. ].join('\n'),
  724. bindings: tools({
  725. echo: async args => ({ echoed: args as CodeJsonValue }),
  726. }),
  727. })
  728. expect(result.error).toBeUndefined()
  729. expect(result.value).toEqual({ echoed: { n: 1 } })
  730. }, 15_000)
  731. it('keeps dispatch working when _lossless_json_violation, asyncio, and send_sync are rebound', async () => {
  732. // dispatch binds _lossless_json_violation, asyncio.get_event_loop, and the
  733. // channel's send method into _run locals before the program runs, so a
  734. // rebind of __main__._lossless_json_violation/__main__.asyncio/
  735. // __main__.ProtocolChannel.send_sync cannot turn a legitimate binding call
  736. // into an exception or a wall-clock timeout.
  737. const { runtime } = await setup()
  738. const result = await runtime.run({
  739. program: [
  740. 'import __main__',
  741. 'def boom(*a, **k):',
  742. ' raise RuntimeError("hijacked")',
  743. '__main__._lossless_json_violation = boom',
  744. '__main__.asyncio = boom',
  745. '__main__.ProtocolChannel.send_sync = boom',
  746. '__main__._encode_json_plain = boom',
  747. '__main__.ProtocolChannel.write_encoded = boom',
  748. 'first = await tools.echo({"n": 1})',
  749. 'return first',
  750. ].join('\n'),
  751. bindings: tools({
  752. echo: async args => ({ echoed: args as CodeJsonValue }),
  753. }),
  754. })
  755. expect(result.error).toBeUndefined()
  756. expect(result.value).toEqual({ echoed: { n: 1 } })
  757. }, 15_000)
  758. it('keeps the reply pump reading when the read_frame_async class attribute is rebound', async () => {
  759. // _pump_replies' frame reader is a bound method captured by _run before the
  760. // program runs and passed in as an explicit argument, so a program rebinding
  761. // `__main__.ProtocolChannel.read_frame_async` cannot redirect the pump (a
  762. // body-local `channel.read_frame_async` lookup would resolve the rebound
  763. // class attribute, since the pump starts after the program's top-level
  764. // statements).
  765. const { runtime } = await setup()
  766. const result = await runtime.run({
  767. program: [
  768. 'import __main__',
  769. 'async def boom(*a, **k):',
  770. ' raise RuntimeError("hijacked reader")',
  771. '__main__.ProtocolChannel.read_frame_async = boom',
  772. 'first = await tools.echo({"n": 1})',
  773. 'return first',
  774. ].join('\n'),
  775. bindings: tools({
  776. echo: async args => ({ echoed: args as CodeJsonValue }),
  777. }),
  778. })
  779. expect(result.error).toBeUndefined()
  780. expect(result.value).toEqual({ echoed: { n: 1 } })
  781. }, 15_000)
  782. it('keeps the rejection contract when _BindingRejection is rebound', async () => {
  783. // `dispatch`'s except clause resolves `_BindingRejection` at call time; a
  784. // program that rebinds `__main__._BindingRejection = ValueError` would
  785. // otherwise let the internal marker type leak into model code (the program
  786. // would catch a `ValueError` for a host rejection). The class is now bound
  787. // into `_run` locals before the program runs, so a host rejection still
  788. // surfaces as the declared `RuntimeError`.
  789. const { runtime } = await setup()
  790. const result = await runtime.run({
  791. program: [
  792. 'import __main__',
  793. '__main__._BindingRejection = ValueError',
  794. 'caught = ""',
  795. 'try:',
  796. ' await tools.fail({})',
  797. 'except RuntimeError as e:',
  798. ' caught = e.args[0] if e.args else ""',
  799. 'except Exception as e:',
  800. ' caught = "WRONG TYPE: " + type(e).__name__',
  801. 'return caught',
  802. ].join('\n'),
  803. bindings: tools({
  804. fail: async () => { throw new Error('nope') },
  805. }),
  806. })
  807. expect(result.error).toBeUndefined()
  808. expect(result.value).toBe('nope')
  809. }, 15_000)
  810. it('still answers the call when the rejection value cannot be converted to a string', async () => {
  811. // `messageOf` calls `String(error)`, which runs the value's own conversion,
  812. // and this call site is a DETACHED async reply callback. A rejection whose
  813. // `Symbol.toPrimitive` throws therefore escaped as an unhandled rejection:
  814. // the reply frame was never written, the program stayed blocked on `await`,
  815. // and the run degraded to a `maxWallMs` timeout (observed) — a host with no
  816. // `unhandledRejection` listener would exit instead. The rejection must reach
  817. // the program as an ordinary error carrying a fixed placeholder.
  818. const { runtime } = await setup({ maxWallMs: 8_000 })
  819. const result = await runtime.run({
  820. program: [
  821. 'try:',
  822. ' await tools.hostile({})',
  823. 'except RuntimeError as e:',
  824. ' return "rejected: " + str(e)',
  825. 'return "no rejection"',
  826. ].join('\n'),
  827. bindings: tools({
  828. hostile: async () => {
  829. throw { [Symbol.toPrimitive]() { throw new Error('toPrimitive blew up') } }
  830. },
  831. }),
  832. })
  833. expect(result.error).toBeUndefined()
  834. expect(result.value).toBe('rejected: <unrenderable rejection value>')
  835. }, 15_000)
  836. it('still answers the call when an Error carries a cyclic value in place of its message', async () => {
  837. // `Error.message` is typed `string` but is a plain writable property, so a
  838. // rejection can carry any value there. Returning it verbatim handed a
  839. // non-string to `sendReply`, breaching `encodeJsonPlain`'s JSON-plain
  840. // precondition: a cyclic object grew the encoder stack until the host threw
  841. // RangeError from the detached reply callback, so no reply frame was written
  842. // and the run degraded to a `maxWallMs` timeout (observed). The conversion
  843. // must contain it — `String()` on a cycle throws inside the guard and lands
  844. // on the placeholder, so the program sees an ordinary error.
  845. const { runtime } = await setup({ maxWallMs: 8_000 })
  846. const result = await runtime.run({
  847. program: [
  848. 'try:',
  849. ' await tools.hostile({})',
  850. 'except RuntimeError as e:',
  851. ' return "rejected: " + str(e)',
  852. 'return "no rejection"',
  853. ].join('\n'),
  854. bindings: tools({
  855. hostile: async () => {
  856. const cyclic: { self?: unknown; [Symbol.toPrimitive]: () => string } = {
  857. // A cycle alone is inert for `String()`; the throwing conversion is
  858. // what proves the guard runs rather than the encoder.
  859. [Symbol.toPrimitive]: () => { throw new Error('cyclic message') },
  860. }
  861. cyclic.self = cyclic
  862. const error = new Error('placeholder')
  863. // Writable per spec, so no cast is needed to install a non-string.
  864. ;(error as unknown as { message: unknown }).message = cyclic
  865. throw error
  866. },
  867. }),
  868. })
  869. expect(result.error).toBeUndefined()
  870. expect(result.value).toBe('rejected: <unrenderable rejection value>')
  871. }, 15_000)
  872. it('renders an Error whose message is a value with no JSON form', async () => {
  873. // The non-cyclic arm. A number would not discriminate: `scalarJson` renders
  874. // it as digits and the child `str()`s the field back, so it survives the
  875. // wire either way. `undefined` is the value that separates the two orders —
  876. // `scalarJson` emits a bare `undefined` token, so the reply line is not JSON
  877. // at all, the child's parse drops the frame, and the program stays blocked
  878. // on `await` until the wall ceiling (observed). Converting first sends the
  879. // string "undefined", which the program receives as an ordinary rejection.
  880. const { runtime } = await setup({ maxWallMs: 8_000 })
  881. const result = await runtime.run({
  882. program: [
  883. 'try:',
  884. ' await tools.absent({})',
  885. 'except RuntimeError as e:',
  886. ' return "rejected: " + str(e)',
  887. 'return "no rejection"',
  888. ].join('\n'),
  889. bindings: tools({
  890. absent: async () => {
  891. const error = new Error('placeholder')
  892. ;(error as unknown as { message: unknown }).message = undefined
  893. throw error
  894. },
  895. }),
  896. })
  897. expect(result.error).toBeUndefined()
  898. expect(result.value).toBe('rejected: undefined')
  899. }, 15_000)
  900. it('runs a program with no await', async () => {
  901. const { runtime } = await setup()
  902. const result = await runtime.run({
  903. program: 'return 2 + 2',
  904. bindings: [],
  905. })
  906. expect(result.error).toBeUndefined()
  907. expect(result.value).toBe(4)
  908. })
  909. it('returns JSON null whether the program returns None or falls off the end', async () => {
  910. // Python has no `undefined`: an async body that returns None and one that
  911. // never returns both yield None, so both complete as an exact JSON null.
  912. // (The worker/TS backend can tell `return undefined` from `return null`;
  913. // Python cannot, and reporting null for both is the honest rendering.)
  914. const { runtime } = await setup()
  915. const explicit = await runtime.run({ program: 'return None', bindings: [] })
  916. expect(explicit.error).toBeUndefined()
  917. expect(explicit.value).toBeNull()
  918. const noReturn = await runtime.run({ program: 'x = 1', bindings: [] })
  919. expect(noReturn.error).toBeUndefined()
  920. expect(noReturn.value).toBeNull()
  921. })
  922. it('settles with no value on a forged valueless done frame', async () => {
  923. // The child always sends a value now (return None → JSON null), so a done
  924. // frame with no value key can only be forged; the host settles it as a
  925. // value-less completion rather than crashing on the absent field.
  926. const { runtime } = await setup()
  927. const result = await runtime.run({
  928. program: [
  929. 'import os',
  930. 'os.write(3, b\'{"type":"done"}\\n\')',
  931. 'import time',
  932. 'time.sleep(5)',
  933. ].join('\n'),
  934. bindings: [],
  935. })
  936. expect(result.error).toBeUndefined()
  937. expect(result.value).toBeUndefined()
  938. })
  939. it('coalesces print arguments into one log line, not per-write fragments', async () => {
  940. // print("a","b") calls write() per arg/sep/newline; the stream must emit
  941. // one logical line "a b" so Code Mode's join(newline) does not insert
  942. // spurious blank lines. Two prints → exactly two entries, no empties.
  943. const { runtime } = await setup()
  944. const result = await runtime.run({
  945. program: ['print("a", "b")', 'print("c")', 'return None'].join('\n'),
  946. bindings: [],
  947. })
  948. expect(result.error).toBeUndefined()
  949. expect(result.logs).toEqual(['a b', 'c'])
  950. })
  951. it('flushes a print with no trailing newline', async () => {
  952. const { runtime } = await setup()
  953. const result = await runtime.run({
  954. program: ['print("partial", end="")', 'return None'].join('\n'),
  955. bindings: [],
  956. })
  957. expect(result.error).toBeUndefined()
  958. expect(result.logs).toEqual(['partial'])
  959. })
  960. it('aggregates a large newline-free native write into one log entry, not one per pipe chunk', async () => {
  961. // A single `os.write` larger than one pipe read arrives as several Node
  962. // `data` chunks. `logs` entries are joined with `\n` downstream, so pushing
  963. // one entry per transport chunk would insert model-visible newlines at
  964. // arbitrary pipe boundaries inside one native write. Stray capture holds a
  965. // per-stream residual and admits only on a real `\n`, so a 200 KiB blast
  966. // with no newline reads back as exactly one entry with no interior breaks.
  967. const { runtime } = await setup({ maxLogBytes: 300_000 })
  968. const size = 200_000
  969. const result = await runtime.run({
  970. program: ['import os', `os.write(1, b"A" * ${size})`, 'return None'].join('\n'),
  971. bindings: [],
  972. })
  973. expect(result.error).toBeUndefined()
  974. expect(result.logs).toEqual(['A'.repeat(size)])
  975. })
  976. it('splits native output on its own newlines, one entry per line', async () => {
  977. // The complement of the aggregation case: real newlines in a native write
  978. // still delimit entries, matching the child's line-granular `log` frames.
  979. const { runtime } = await setup()
  980. const result = await runtime.run({
  981. program: ['import os', 'os.write(1, b"one\\ntwo\\nthree")', 'return None'].join('\n'),
  982. bindings: [],
  983. })
  984. expect(result.error).toBeUndefined()
  985. expect(result.logs).toEqual(['one', 'two', 'three'])
  986. })
  987. it('bounds a newline-free native flood by the ledger instead of buffering it whole', async () => {
  988. // A newline-free write far larger than maxLogBytes must not accumulate in
  989. // the host-side residual: when the pending residual would cross the budget
  990. // it is admitted (and truncated) immediately, and once the ledger has
  991. // truncated, later chunks stop buffering entirely. The run still completes
  992. // and the captured output ends at the truncation marker rather than
  993. // retaining the whole flood.
  994. const { runtime } = await setup({ maxLogBytes: 4096 })
  995. const result = await runtime.run({
  996. program: ['import os', 'os.write(1, b"A" * 2_000_000)', 'return None'].join('\n'),
  997. bindings: [],
  998. })
  999. expect(result.error).toBeUndefined()
  1000. expect(result.logs.at(-1)).toBe(logTruncationMarker(4096))
  1001. // The retained output is bounded by the budget, not the 2 MB flood.
  1002. expect(result.logs.join('').length).toBeLessThan(4096)
  1003. })
  1004. it('bounds a newline-free single-character Python write drip by the fragment cap, not OOM', async () => {
  1005. // The child-side `_LogStream` buffers one fragment per `write` (so
  1006. // `print("x", end="")` does not concatenate quadratically). A newline-free
  1007. // drip of one character per call past a large `maxLogBytes` would otherwise
  1008. // accumulate one list slot (and one str object) per call — 25 M calls =
  1009. // ~25 M slots, which OOMs the host on its own accounting before the byte
  1010. // budget is reached. The stream seals the fragment list past
  1011. // `_PENDING_MAX_CHUNKS` into one joined block (character count unchanged),
  1012. // bounding the live fragment count exactly as the host-side `captureStray`
  1013. // seal does. This drives well past the cap and asserts the run still
  1014. // completes with a truncation marker rather than a MemoryError.
  1015. const { runtime } = await setup({ maxLogBytes: 4096 })
  1016. const result = await runtime.run({
  1017. program: [
  1018. 'import sys',
  1019. 'for _ in range(200_000):',
  1020. ' sys.stdout.write("x")',
  1021. 'return None',
  1022. ].join('\n'),
  1023. bindings: [],
  1024. })
  1025. expect(result.error).toBeUndefined()
  1026. expect(result.logs.at(-1)).toBe(logTruncationMarker(4096))
  1027. })
  1028. it('bounds a control-char-dense native residual by serialized cost, not raw length', async () => {
  1029. // A newline-free NUL flood passes the cheap `length + 3` lower bound at a
  1030. // raw length well under the budget, but each NUL serializes to `\u0000` (6
  1031. // bytes), so the true JSON cost is ~6x. The ledger must charge that
  1032. // serialized cost — and `jsonStringCostUpTo` must measure it WITHOUT
  1033. // allocating the escaped copy, so a near-budget line under a large
  1034. // maxLogBytes cannot momentarily allocate a multi-gigabyte `JSON.stringify`
  1035. // result. Under a small budget the residual is truncated once the serialized
  1036. // cost crosses it.
  1037. const { runtime } = await setup({ maxLogBytes: 4096 })
  1038. const result = await runtime.run({
  1039. program: ['import os', 'os.write(1, b"\\x00" * 4000)', 'return None'].join('\n'),
  1040. bindings: [],
  1041. })
  1042. expect(result.error).toBeUndefined()
  1043. expect(result.logs.at(-1)).toBe(logTruncationMarker(4096))
  1044. })
  1045. it('rejects an output budget that could breach addressSpaceMb during encode at load', async () => {
  1046. // The child builds, charges, and encodes a `maxLogBytes` log entry or a
  1047. // `maxValueBytes` completion value under RLIMIT_AS, and both trigger on
  1048. // character count against a serialized-byte budget — an astral character is
  1049. // one character but ~4 bytes stored and ~4 encoded, and THREE such copies are
  1050. // live at the peak (the caller's write argument, the slice/join handed to
  1051. // push, and the encode copy), so a budget approaching the address space lets a
  1052. // legitimate near-budget output breach it and die as worker-exit. The
  1053. // incompatible pair is rejected at load: each budget times the worst-case
  1054. // multiple (12) must fit the address space LEFT after the fixed interpreter
  1055. // baseline. Against a 256 MiB address space that leaves 192 MiB budgetable
  1056. // (~16 MiB admissible), so a 50 MB cap is far over; the default caps against
  1057. // 512 MiB are not. Both budgets are gated symmetrically — the value case sets
  1058. // a default-fitting maxLogBytes so the maxValueBytes check is what fires.
  1059. const ctxLog = new Context()
  1060. await expect(ctxLog.plugin(PythonCodeRuntime, { maxLogBytes: 50_000_000, addressSpaceMb: 256 }))
  1061. .rejects.toThrow(/maxLogBytes times the 12x worst-case Unicode expansion must fit/)
  1062. const ctxValue = new Context()
  1063. await expect(ctxValue.plugin(PythonCodeRuntime, { maxValueBytes: 50_000_000, addressSpaceMb: 256 }))
  1064. .rejects.toThrow(/maxValueBytes times the 12x worst-case Unicode expansion must fit/)
  1065. // Discriminates 12 from 8: a 48 MiB maxLogBytes against a 512 MiB address
  1066. // space leaves 448 MiB budgetable. 48*8 = 384 MiB fits (the old 8x multiple
  1067. // wrongly ADMITTED this), but 48*12 = 576 MiB does not. The ~12x peak this
  1068. // guards is the NEWLINE path's single near-budget write — the caller's own
  1069. // string, the line slice, and the encode copy live at once. The settlement
  1070. // flush is no longer the binding case: `flush_line` drops the pending chunks
  1071. // before its push, so it holds two copies, not three.
  1072. const ctxTwelve = new Context()
  1073. await expect(ctxTwelve.plugin(PythonCodeRuntime, { maxLogBytes: 48 * 1024 * 1024, addressSpaceMb: 512 }))
  1074. .rejects.toThrow(/maxLogBytes times the 12x worst-case Unicode expansion must fit/)
  1075. // An addressSpaceMb at or below the interpreter baseline leaves nothing
  1076. // budgetable, so no budget value can pass. It is rejected on its own terms:
  1077. // the budget loop would otherwise report "a limit of -1" (or -2796203 at
  1078. // 32 MiB) while naming maxLogBytes, sending the operator to the wrong knob.
  1079. const ctxBaseline = new Context()
  1080. await expect(ctxBaseline.plugin(PythonCodeRuntime, { addressSpaceMb: 64 }))
  1081. .rejects.toThrow(/addressSpaceMb must exceed the 67108864-byte interpreter baseline/)
  1082. const ctxBelow = new Context()
  1083. await expect(ctxBelow.plugin(PythonCodeRuntime, { addressSpaceMb: 32 }))
  1084. .rejects.toThrow(/addressSpaceMb must exceed the 67108864-byte interpreter baseline/)
  1085. // The default caps against the default 512 MiB address space load.
  1086. const ok = new Context()
  1087. const fiber = await ok.plugin(PythonCodeRuntime, { maxLogBytes: 65536, maxValueBytes: 32768, addressSpaceMb: 512 })
  1088. await fiber.dispose()
  1089. })
  1090. it('bounds an illegal-UTF-8 native residual by its U+FFFD-decoded cost', async () => {
  1091. // Every 0xFF byte is illegal in any UTF-8 sequence, so `toString('utf8')`
  1092. // renders each as U+FFFD (3 serialized bytes). `accrueStrayCost` must charge
  1093. // that 3, not the raw 1: otherwise the newline-free residual grows to a full
  1094. // budget's worth of RAW bytes before flushing — a ~3x undercount that near a
  1095. // large maxLogBytes retains hundreds of MiB then expands toward a ~1 GiB peak
  1096. // in flushStray's concat + toString. Paced single-byte writes (each its own
  1097. // `data` chunk, like the sealing case) expose the sub-chunk accrual: charged
  1098. // at 3 the residual crosses a 3072-byte budget after ~1024 bytes and flushes;
  1099. // charged at 1 it would need ~3072 bytes, so the peak residual triples. The
  1100. // largest merged buffer is the discriminator.
  1101. const realConcat = Buffer.concat.bind(Buffer)
  1102. let maxConcat = 0
  1103. Buffer.concat = (list: readonly Uint8Array[], total?: number): Buffer<ArrayBuffer> => {
  1104. const merged = realConcat(list, total)
  1105. if (merged.length > maxConcat) maxConcat = merged.length
  1106. return merged
  1107. }
  1108. let result: CodeRunResult
  1109. try {
  1110. const { runtime } = await setup({ maxLogBytes: 3072, maxWallMs: 30_000 })
  1111. result = await runtime.run({
  1112. program: [
  1113. 'import os',
  1114. 'for _ in range(6000):',
  1115. ' os.write(1, b"\\xff")',
  1116. ' os.sched_yield()',
  1117. 'return None',
  1118. ].join('\n'),
  1119. bindings: [],
  1120. })
  1121. } finally {
  1122. Buffer.concat = realConcat
  1123. }
  1124. expect(result.error).toBeUndefined()
  1125. expect(result.logs.at(-1)).toBe(logTruncationMarker(3072))
  1126. // Charged at 3, the residual flushes around 1024 raw bytes; the largest
  1127. // merged buffer stays well under 2048. A raw-byte undercount would let it
  1128. // reach ~3072 before flushing, so 2048 discriminates.
  1129. expect(maxConcat).toBeLessThan(2048)
  1130. })
  1131. it('charges a structurally-valid but illegal UTF-8 sequence its U+FFFD-decoded cost', async () => {
  1132. // A CESU-8 lone surrogate `ED A0 80` is structurally well-formed (a 3-byte
  1133. // lead plus two 0x80–0xBF continuations) but ILLEGAL: `toString('utf8')`
  1134. // renders each of the three bytes as its own U+FFFD (serialized cost 9), not
  1135. // one width-3 character. The newline-free flush trigger weighs the residual
  1136. // through `accrueStrayCost`, which must validate each lead's
  1137. // first-continuation range (ED excludes A0–BF) and charge the true 9 — else a
  1138. // CESU flood undercounts 3x and the residual grows toward a full budget's raw
  1139. // bytes before flushing, the same peak-memory vector as the 0xFF case. The
  1140. // bytes are written one at a time (each its own `data` chunk, no pipe
  1141. // coalescing) and `Buffer.concat` is wrapped to measure the peak residual.
  1142. const realConcat = Buffer.concat.bind(Buffer)
  1143. let maxConcat = 0
  1144. Buffer.concat = (list: readonly Uint8Array[], total?: number): Buffer<ArrayBuffer> => {
  1145. const merged = realConcat(list, total)
  1146. if (merged.length > maxConcat) maxConcat = merged.length
  1147. return merged
  1148. }
  1149. let result: CodeRunResult
  1150. try {
  1151. const { runtime } = await setup({ maxLogBytes: 3072, maxWallMs: 30_000 })
  1152. result = await runtime.run({
  1153. program: [
  1154. 'import os',
  1155. 'seq = (0xed, 0xa0, 0x80)',
  1156. 'for _ in range(2000):',
  1157. ' for b in seq:',
  1158. ' os.write(1, bytes((b,)))',
  1159. ' os.sched_yield()',
  1160. 'return None',
  1161. ].join('\n'),
  1162. bindings: [],
  1163. })
  1164. } finally {
  1165. Buffer.concat = realConcat
  1166. }
  1167. expect(result.error).toBeUndefined()
  1168. expect(result.logs.at(-1)).toBe(logTruncationMarker(3072))
  1169. // Each 3-byte sequence costs 9 (three U+FFFD), so single-byte-paced the
  1170. // residual crosses the 3072 budget after ~342 raw bytes and flushes; the
  1171. // largest merged buffer stays well under 2048. Charging the structural width
  1172. // 3 would need ~1024 raw bytes, tripling the peak past 2048.
  1173. expect(maxConcat).toBeLessThan(2048)
  1174. })
  1175. it('charges a lone surrogate its full six escaped bytes, not three', async () => {
  1176. // A forged `log` frame carrying `\ud800` escapes materializes lone
  1177. // surrogates after JSON.parse. `Buffer.byteLength` of U+FFFD is 3, but
  1178. // ES2019 well-formed `JSON.stringify` emits `\ud800` at 6 bytes, so charging
  1179. // the raw width would admit ~2x the configured budget of serialized bytes
  1180. // (the same family as the NUL-flood undercount, at 2x rather than 6x). The
  1181. // cost walker charges surrogates the full 6, so a flood truncates at budget.
  1182. // Forged on fd 3 because Python stdout will not emit lone surrogates.
  1183. const { runtime } = await setup({ maxLogBytes: 4096 })
  1184. const result = await runtime.run({
  1185. program: [
  1186. 'import os',
  1187. // 1000 \ud800 escapes: charged at the buggy raw width 1000 * 3 = 3000
  1188. // bytes fits under 4096 (wrongly admitted), but the correct serialized
  1189. // width 1000 * 6 = 6000 bytes is over budget — so the ledger must
  1190. // truncate. The count sits in the 683..1365 window where the two
  1191. // chargings disagree, making the test discriminate.
  1192. String.raw`frame = b'{"type":"log","text":"' + b'\\ud800' * 1000 + b'"}\n'`,
  1193. 'os.write(3, frame)',
  1194. 'return None',
  1195. ].join('\n'),
  1196. bindings: [],
  1197. })
  1198. expect(result.error).toBeUndefined()
  1199. expect(result.logs.at(-1)).toBe(logTruncationMarker(4096))
  1200. })
  1201. it('drops a second stray line in the same chunk once the first truncated the ledger', async () => {
  1202. // One `os.write` carrying two newline-terminated lines where the first
  1203. // exhausts maxLogBytes: the first line's admit truncates and marks the
  1204. // ledger, and the second line's admit — reached in the same `data` callback
  1205. // — must be the post-truncation no-op. Proves that branch is exercised, so
  1206. // it carries no v8-ignore. Kept to 108 bytes (< the smallest PIPE_BUF, 512 on
  1207. // macOS) so the whole payload lands in ONE atomic write and one `data`
  1208. // callback — the two newlines cannot split across callbacks and leave the
  1209. // branch un-exercised, which would be a hard-to-attribute per-file coverage
  1210. // flake. The first line's 100 bytes already exceed the 64-byte budget, so it truncates.
  1211. const { runtime } = await setup({ maxLogBytes: 64 })
  1212. const result = await runtime.run({
  1213. program: ['import os', 'os.write(1, b"A" * 100 + b"\\nSECOND\\n")', 'return None'].join('\n'),
  1214. bindings: [],
  1215. })
  1216. expect(result.error).toBeUndefined()
  1217. expect(result.logs.at(-1)).toBe(logTruncationMarker(64))
  1218. expect(result.logs.join('\n')).not.toContain('SECOND')
  1219. })
  1220. it('charges a broken multibyte sequence its U+FFFD bytes, split across pipe chunks', async () => {
  1221. // A 3-byte lead (0xE4) whose continuation never arrives — the next byte is a
  1222. // fresh ASCII 'A' — must be costed as U+FFFD (3) for the orphaned lead, not
  1223. // folded into a phantom character. Driven byte-by-byte so the lead and the
  1224. // breaking byte land in separate `data` chunks, exercising accrueStrayCost's
  1225. // cross-chunk broken-sequence branch. The run completes and the bytes are
  1226. // captured (rendered U+FFFD by toString), proving the walk resynchronizes.
  1227. const { runtime } = await setup({ maxLogBytes: 1024 })
  1228. const result = await runtime.run({
  1229. program: [
  1230. 'import os',
  1231. 'os.write(1, b"\\xe4")',
  1232. 'os.sched_yield()',
  1233. 'os.write(1, b"A\\n")',
  1234. 'return None',
  1235. ].join('\n'),
  1236. bindings: [],
  1237. })
  1238. expect(result.error).toBeUndefined()
  1239. expect(result.logs.join('')).toContain('A')
  1240. expect(result.logs.join('')).toContain('�')
  1241. })
  1242. it('charges the exact serialized cost of short-escape and quote/backslash characters', async () => {
  1243. // Exercises every branch of jsonStringCostUpTo's per-character cost: a tab
  1244. // and other C0 controls with short JSON forms (\t etc., 2 bytes), a quote
  1245. // and backslash (2 bytes each), a `\uXXXX` control (6 bytes), a multibyte
  1246. // BMP character (raw UTF-8 width), and plain ASCII. Under a budget large
  1247. // enough to admit it, the line survives verbatim — proving the cost walker
  1248. // does not over- or under-charge and the string round-trips unescaped.
  1249. const { runtime } = await setup({ maxLogBytes: 4096 })
  1250. const result = await runtime.run({
  1251. program: ['import os', String.raw`os.write(1, "\ta\"b\\c\x01é\n".encode("utf-8"))`, 'return None'].join('\n'),
  1252. bindings: [],
  1253. })
  1254. expect(result.error).toBeUndefined()
  1255. expect(result.logs).toEqual(['\ta"b\\c\x01é'])
  1256. })
  1257. it('fails a completion dict with a non-string key as invalid-output (no key coercion)', async () => {
  1258. // json.dumps would coerce {1: "a", "1": "b"} to a single "1" key, silently
  1259. // dropping data. The shape validator rejects it before encoding.
  1260. const { runtime } = await setup()
  1261. const result = await runtime.run({
  1262. program: 'return {1: "first", "1": "second"}',
  1263. bindings: [],
  1264. })
  1265. expect(result.value).toBeUndefined()
  1266. expect(result.error?.kind).toBe('invalid-output')
  1267. expect(result.error?.message).toContain('non-string dict key')
  1268. })
  1269. it('rejects a binding argument with a non-string dict key before dispatch', async () => {
  1270. const { runtime } = await setup()
  1271. let called = false
  1272. const result = await runtime.run({
  1273. program: [
  1274. 'caught = ""',
  1275. 'try:',
  1276. ' await tools.sink({1: "x"})',
  1277. 'except RuntimeError as e:',
  1278. ' caught = str(e)',
  1279. 'return caught',
  1280. ].join('\n'),
  1281. bindings: tools({ sink: async () => { called = true; return null } }),
  1282. })
  1283. expect(result.error).toBeUndefined()
  1284. expect(result.value).toContain('lossless JSON')
  1285. expect(called).toBe(false)
  1286. })
  1287. it('fails a non-JSON completion value as invalid-output (no repr substitution)', async () => {
  1288. // A set is not lossless JSON. The old draft substituted repr(); the seam
  1289. // now requires refusing the run instead.
  1290. const { runtime } = await setup()
  1291. const result = await runtime.run({
  1292. program: 'return {1, 2, 3}',
  1293. bindings: [],
  1294. })
  1295. expect(result.value).toBeUndefined()
  1296. expect(result.error?.kind).toBe('invalid-output')
  1297. expect(result.error?.message).toContain('lossless JSON')
  1298. expect(result.error?.message).toContain('set')
  1299. })
  1300. it('fails a negative-zero completion value as invalid-output (sign bit is lossy over JSON)', async () => {
  1301. // JSON serialization turns -0.0 into 0 (or JS -0), silently changing the
  1302. // sign bit; the canonical lossless-JSON boundary rejects it, so the
  1303. // Python side must too — as a completion and as a binding argument.
  1304. const { runtime } = await setup()
  1305. const completion = await runtime.run({
  1306. program: 'return -0.0',
  1307. bindings: [],
  1308. })
  1309. expect(completion.error?.kind).toBe('invalid-output')
  1310. expect(completion.error?.message).toContain('negative zero')
  1311. const argument = await runtime.run({
  1312. program: [
  1313. 'try:',
  1314. ' await tools.echo(-0.0)',
  1315. ' return "accepted"',
  1316. 'except RuntimeError as e:',
  1317. ' return str(e)',
  1318. ].join('\n'),
  1319. bindings: tools({ echo: async args => args as never }),
  1320. })
  1321. expect(argument.error).toBeUndefined()
  1322. expect(argument.value).toContain('negative zero')
  1323. })
  1324. it('fails a NaN completion value as invalid-output (allow_nan=False)', async () => {
  1325. // json.dumps would happily emit NaN by default, but NaN is not JSON; the
  1326. // bootstrap passes allow_nan=False so it fails as invalid-output.
  1327. const { runtime } = await setup()
  1328. const result = await runtime.run({
  1329. program: 'return float("nan")',
  1330. bindings: [],
  1331. })
  1332. expect(result.error?.kind).toBe('invalid-output')
  1333. })
  1334. it('fails an over-budget completion value as output-limit (child-side check)', async () => {
  1335. const { runtime } = await setup({ maxValueBytes: 64 })
  1336. const result = await runtime.run({
  1337. program: 'return "V" * 5000',
  1338. bindings: [],
  1339. })
  1340. expect(result.value).toBeUndefined()
  1341. expect(result.error?.kind).toBe('output-limit')
  1342. expect(result.error?.message).toContain('exceeded 64 bytes')
  1343. })
  1344. it('meters a control-heavy completion value without materializing its escaped form', async () => {
  1345. // The child's lower bound admits a string by CHARACTER count, then the meter
  1346. // charged what `_dump_string(current).encode()` returned -- building the
  1347. // escaped copy plus its encode. Each NUL escapes to six bytes, so metering a
  1348. // value the budget then REJECTS allocated ~6x the original twice over:
  1349. // measured at 228.9 MiB of peak for a 20M-NUL string, against 19.1 MiB for
  1350. // the counting path that returns the identical 120,000,002 bytes. Past
  1351. // RLIMIT_AS the meter died as `exception: MemoryError`, inverting the
  1352. // `output-limit` this seam promises for an over-budget value.
  1353. //
  1354. // 8M NULs is 8,000,002 raw but 48,000,002 escaped: over the 16 MiB budget
  1355. // only when charged the escaped cost, so this also pins that the cheap
  1356. // character bound alone does not decide the verdict.
  1357. const { runtime } = await setup({ maxValueBytes: 16 * 1024 * 1024, maxWallMs: 60_000 })
  1358. const result = await runtime.run({
  1359. program: 'return "\\x00" * 8_000_000',
  1360. bindings: [],
  1361. })
  1362. expect(result.value).toBeUndefined()
  1363. expect(result.error?.kind).toBe('output-limit')
  1364. }, 90_000)
  1365. it('rejects a wide completion as output-limit before materializing its traversal state', async () => {
  1366. // `[0] * 2000000` sits far above maxValueBytes but well below the frame
  1367. // ceiling. The folded checker must reject it via the pre-enqueue bound —
  1368. // BEFORE pushing two million elements onto the walk — so a small
  1369. // addressSpaceMb does not turn the check itself into an RLIMIT_AS death.
  1370. const { runtime } = await setup({ maxValueBytes: 64, addressSpaceMb: 256, maxWallMs: 15_000 })
  1371. const result = await runtime.run({
  1372. program: 'return [0] * 2000000',
  1373. bindings: [],
  1374. })
  1375. expect(result.value).toBeUndefined()
  1376. expect(result.error?.kind).toBe('output-limit')
  1377. expect(result.error?.message).toContain('exceeded 64 bytes')
  1378. }, 20_000)
  1379. it('rejects a wide dict as output-limit without materializing its items list', async () => {
  1380. // Same pre-enqueue bound on the dict branch: `len(current)` replaces
  1381. // `list(current.items())`, which allocated one tuple per member before the
  1382. // bound could reject the value. Two million entries under a 64-byte cap
  1383. // fits the 256 MiB address space as a dict but not as a dict PLUS a
  1384. // two-million-tuple list.
  1385. const { runtime } = await setup({ maxValueBytes: 64, addressSpaceMb: 256, maxWallMs: 15_000 })
  1386. const result = await runtime.run({
  1387. program: 'return {str(i): 0 for i in range(2000000)}',
  1388. bindings: [],
  1389. })
  1390. expect(result.value).toBeUndefined()
  1391. expect(result.error?.kind).toBe('output-limit')
  1392. expect(result.error?.message).toContain('exceeded 64 bytes')
  1393. }, 20_000)
  1394. it('meters a float completion in the host\'s number spelling', async () => {
  1395. // CPython's repr disagrees with the host's String(number): `1.0` is three
  1396. // bytes here and one there, `1e-07` pads the exponent the host writes as
  1397. // `1e-7`. Both sides meter the SAME budget, so the child must count the
  1398. // bytes the host will receive — otherwise a boundary-sized value is
  1399. // falsely reported as output-limit.
  1400. const { runtime } = await setup({ maxValueBytes: 1 })
  1401. const integral = await runtime.run({ program: 'return 1.0', bindings: [] })
  1402. expect(integral.error).toBeUndefined()
  1403. expect(integral.value).toBe(1)
  1404. const exponent = await setup({ maxValueBytes: 4 })
  1405. const small = await exponent.runtime.run({ program: 'return 1e-7', bindings: [] })
  1406. expect(small.error).toBeUndefined()
  1407. expect(small.value).toBe(1e-7)
  1408. // The spelling is a meter input, not a licence to overshoot: `1.5` is three
  1409. // bytes on both sides and still fails a two-byte budget.
  1410. const tight = await setup({ maxValueBytes: 2 })
  1411. const over = await tight.runtime.run({ program: 'return 1.5', bindings: [] })
  1412. expect(over.error?.kind).toBe('output-limit')
  1413. })
  1414. it('carries floats across the wire in the host\'s number spelling', async () => {
  1415. // The child ENCODES with the same speller it meters with, so the frame the
  1416. // host parses must reproduce every double exactly — including the branches
  1417. // where CPython and ECMAScript disagree (integral floats, sub-1e-6
  1418. // exponents, >= 1e21, and beyond-safe-range integral doubles whose exact
  1419. // digits differ from the shortest round-trip form).
  1420. const { runtime } = await setup()
  1421. const result = await runtime.run({
  1422. 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]',
  1423. bindings: [],
  1424. })
  1425. expect(result.error).toBeUndefined()
  1426. 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])
  1427. })
  1428. it('rejects a forged non-lossless done value host-side as invalid-output', async () => {
  1429. // A forged done frame bypasses the child's _check_done_value. JSON.parse
  1430. // turns 1e400 into Infinity; validateChildFrame no longer scans done.value,
  1431. // so the host's own checkDoneValue must catch the non-lossless number.
  1432. const { runtime } = await setup()
  1433. const result = await runtime.run({
  1434. program: [
  1435. 'import os',
  1436. String.raw`os.write(3, b'{"type":"done","value":1e400}' + b'\n')`,
  1437. 'import time',
  1438. 'time.sleep(5)',
  1439. ].join('\n'),
  1440. bindings: [],
  1441. })
  1442. expect(result.value).toBeUndefined()
  1443. expect(result.error?.kind).toBe('invalid-output')
  1444. expect(result.error?.message).toContain('non-lossless number')
  1445. })
  1446. it('reports a syntax error as an exception without settling with a value', async () => {
  1447. const { runtime } = await setup()
  1448. const result = await runtime.run({
  1449. program: '$$invalid python$$',
  1450. bindings: [],
  1451. })
  1452. expect(result.error?.kind).toBe('exception')
  1453. expect(result.error?.message).toContain('SyntaxError')
  1454. // The parse-time diagnostic must carry the same source label as compile and
  1455. // runtime tracebacks (ast.parse passes filename="<model>"); a stale
  1456. // "<unknown>" label would leak an inconsistent origin to the model.
  1457. expect(result.error?.message).toContain('File \"<model>\"')
  1458. expect(result.value).toBeUndefined()
  1459. })
  1460. it('reports a runtime raise as an exception with the traceback', async () => {
  1461. const { runtime } = await setup()
  1462. const result = await runtime.run({
  1463. program: 'raise ValueError("intentional")',
  1464. bindings: [],
  1465. })
  1466. expect(result.error?.kind).toBe('exception')
  1467. expect(result.error?.message).toContain('ValueError')
  1468. expect(result.error?.message).toContain('intentional')
  1469. })
  1470. it('bounds a deep exception cause chain instead of burning the wall budget formatting it', async () => {
  1471. // A chain thousands of links deep would make the rendering walk and
  1472. // format() linear in its length, consuming maxWallMs. Rendering is capped
  1473. // at 100 links with a marker; the run reports the exception well within
  1474. // budget rather than timing out.
  1475. const { runtime } = await setup({ maxValueBytes: 1024 * 1024, maxWallMs: 20_000 })
  1476. const start = Date.now()
  1477. const result = await runtime.run({
  1478. program: [
  1479. 'err = None',
  1480. 'for i in range(3000):',
  1481. ' try:',
  1482. ' raise ValueError(i) from err',
  1483. ' except ValueError as e:',
  1484. ' err = e',
  1485. 'raise err',
  1486. ].join('\n'),
  1487. bindings: [],
  1488. })
  1489. expect(result.error?.kind).toBe('exception')
  1490. expect(result.error?.message).toContain('exception chain truncated at 100 links')
  1491. expect(Date.now() - start).toBeLessThan(15_000)
  1492. }, 25_000)
  1493. it('bounds an over-cap chain without assigning to the live exception', async () => {
  1494. // The cap used to be applied by severing the over-cap link ON the live
  1495. // exception. An exception class overriding __setattr__ to raise turned that
  1496. // assignment into model code running inside the bootstrap's failure
  1497. // handler; the throw skipped the `done` send that sits after the handler,
  1498. // so the host blocked on fd 3 and reported a maxWallMs timeout instead of
  1499. // the model's own exception. Cutting the chain on the TracebackException
  1500. // COPY touches no model hook, so the marker still appears and the run
  1501. // reports `exception`.
  1502. const { runtime } = await setup({ maxValueBytes: 1024 * 1024, maxWallMs: 15_000 })
  1503. const start = Date.now()
  1504. const result = await runtime.run({
  1505. program: [
  1506. 'class Sealed(Exception):',
  1507. ' def __setattr__(self, name, value):',
  1508. ' raise RuntimeError("live mutation refused")',
  1509. 'err = None',
  1510. 'for i in range(150):',
  1511. ' try:',
  1512. ' raise Sealed(i) from err',
  1513. ' except Sealed as e:',
  1514. ' err = e',
  1515. 'raise err',
  1516. ].join('\n'),
  1517. bindings: [],
  1518. })
  1519. expect(result.error?.kind).toBe('exception')
  1520. expect(result.error?.message).toContain('Sealed')
  1521. expect(result.error?.message).toContain('exception chain truncated at 100 links')
  1522. // The sever attempt is what used to leak: its message must not appear, and
  1523. // the run must settle well inside the wall budget rather than timing out.
  1524. expect(result.error?.message).not.toContain('live mutation refused')
  1525. expect(Date.now() - start).toBeLessThan(10_000)
  1526. }, 20_000)
  1527. it('still sends done when rendering the diagnostic itself raises', async () => {
  1528. // format() reaches the exception's own __str__, so a model class whose
  1529. // __str__ raises can throw from inside the failure handler. CPython's
  1530. // _safe_string absorbs a raising __str__ during formatting, but the
  1531. // fallback must hold for any throw on that path (a raising __repr__ of an
  1532. // argument, a MemoryError under RLIMIT_AS), so the assertion is the
  1533. // invariant that matters: a `done` frame carrying `exception`, never a
  1534. // timeout, and never the failing renderer's own message.
  1535. const { runtime } = await setup({ maxWallMs: 10_000 })
  1536. const result = await runtime.run({
  1537. program: [
  1538. 'class Unprintable(Exception):',
  1539. ' def __str__(self):',
  1540. ' raise RuntimeError("str refused")',
  1541. ' def __repr__(self):',
  1542. ' raise RuntimeError("repr refused")',
  1543. 'raise Unprintable()',
  1544. ].join('\n'),
  1545. bindings: [],
  1546. })
  1547. expect(result.error?.kind).toBe('exception')
  1548. expect(result.error?.message).toContain('Unprintable')
  1549. expect(result.error?.message).not.toContain('str refused')
  1550. expect(result.error?.message).not.toContain('repr refused')
  1551. }, 15_000)
  1552. it('sends done with an inert diagnostic when the whole rendering path raises', async () => {
  1553. // Drive the fallback itself. `TracebackException.format` reads the
  1554. // exception class's `__module__` to decide whether to qualify the name, and
  1555. // a metaclass property can raise there — a throw INSIDE the formatter,
  1556. // reached with no rebinding of anything the bootstrap owns. Without the
  1557. // wrapper it escapes the handler, the `done` send never runs, and the host
  1558. // times out at maxWallMs.
  1559. const { runtime } = await setup({ maxWallMs: 10_000 })
  1560. const result = await runtime.run({
  1561. program: [
  1562. 'class Meta(type):',
  1563. ' @property',
  1564. ' def __module__(cls):',
  1565. ' raise RuntimeError("renderer refused")',
  1566. 'class Hostile(ValueError, metaclass=Meta):',
  1567. ' pass',
  1568. 'raise Hostile("original failure")',
  1569. ].join('\n'),
  1570. bindings: [],
  1571. })
  1572. expect(result.error?.kind).toBe('exception')
  1573. // The inert fallback names the class and a fixed literal; it must not carry
  1574. // the renderer's message, and must not have become a timeout. `__name__` is
  1575. // still a plain str here, so the class name survives.
  1576. expect(result.error?.message).toBe('Hostile: <diagnostic rendering failed>')
  1577. }, 15_000)
  1578. it('falls back to a placeholder class name when __name__ itself raises', async () => {
  1579. // The fallback reads type(exc).__name__, which a metaclass property can
  1580. // hijack. It must neither run that override's failure into the handler nor
  1581. // format a non-str __name__ into the message. The hostile `__module__` is
  1582. // what drives execution into the fallback in the first place.
  1583. const { runtime } = await setup({ maxWallMs: 10_000 })
  1584. const result = await runtime.run({
  1585. program: [
  1586. 'class Meta(type):',
  1587. ' @property',
  1588. ' def __module__(cls):',
  1589. ' raise RuntimeError("renderer refused")',
  1590. ' @property',
  1591. ' def __name__(cls):',
  1592. ' raise RuntimeError("name refused")',
  1593. 'class Nameless(Exception, metaclass=Meta):',
  1594. ' pass',
  1595. 'raise Nameless()',
  1596. ].join('\n'),
  1597. bindings: [],
  1598. })
  1599. expect(result.error?.kind).toBe('exception')
  1600. expect(result.error?.message).toBe('<unknown>: <diagnostic rendering failed>')
  1601. }, 15_000)
  1602. it('reports the real exception when the program rebinds every name the failure path uses', async () => {
  1603. // The bootstrap IS __main__, so `import __main__; __main__._X = ...` reaches
  1604. // any module global a call-time lookup would read. The failure path is the
  1605. // worst place for that: the reporter, the byte cap, the traceback formatter,
  1606. // the settlement flush and the `done` send all run AFTER the `except` block,
  1607. // so a replacement that raises skips the send, leaves the host blocked on
  1608. // fd 3, and the run reports a maxWallMs timeout instead of the model's own
  1609. // exception. Rebind all of them at once; the run must still carry the real
  1610. // ValueError.
  1611. const { runtime } = await setup({ maxWallMs: 10_000 })
  1612. const result = await runtime.run({
  1613. program: [
  1614. 'import __main__',
  1615. 'def boom(*a, **k):',
  1616. ' raise RuntimeError("hijacked")',
  1617. '__main__._SAFE_MODEL_TRACEBACK = boom',
  1618. '__main__._cap_message = boom',
  1619. '__main__._model_traceback = boom',
  1620. '__main__._UNRENDERABLE_DIAGNOSTIC = boom',
  1621. '__main__._LogStream.flush_line = boom',
  1622. // `send_done` writes via LOCALLY-BOUND `_encode_json_plain` +
  1623. // `ProtocolChannel.write_encoded`; rebinding these at call time must not
  1624. // redirect the done frame (a late lookup would be `boom` -> worker-exit).
  1625. '__main__.ProtocolChannel.send_sync = boom',
  1626. '__main__.ProtocolChannel.write_encoded = boom',
  1627. '__main__._encode_json_plain = boom',
  1628. 'raise ValueError("real failure")',
  1629. ].join('\n'),
  1630. bindings: [],
  1631. })
  1632. expect(result.error?.kind).toBe('exception')
  1633. expect(result.error?.message).toContain('ValueError: real failure')
  1634. expect(result.error?.message).not.toContain('hijacked')
  1635. }, 15_000)
  1636. it('still delivers a done frame when a transitive encode name is rebound', async () => {
  1637. // `send_done` binds `_encode_json_plain` and `ProtocolChannel.write_encoded`
  1638. // into locals, but those callables' BODIES still resolve transitive module
  1639. // globals at call time: `_encode_json_plain` reaches `_dump_scalar`/`_dump_string`/
  1640. // `json.dumps`, and `write_encoded` reaches `os.write`. This bootstrap is
  1641. // `__main__`, so rebinding `__main__._dump_scalar` to a raising function makes
  1642. // the error-frame encode throw AFTER the `except` block. `send_done` catches
  1643. // that and writes a fixed literal done frame (kind `exception`) with the
  1644. // LOCALLY-BOUND `_os_write`/`_memoryview`/`_FALLBACK_DONE_FRAME` captured
  1645. // before the program runs, so the host still gets a verdict — the run must be an
  1646. // `exception`, never a `worker-exit`. The real message is lost (the literal
  1647. // carries a fixed `<unrenderable>` text), which is acceptable: the verdict
  1648. // outranks the diagnostic detail.
  1649. const { runtime } = await setup({ maxWallMs: 10_000 })
  1650. const result = await runtime.run({
  1651. program: [
  1652. 'import __main__',
  1653. 'def boom(*a, **k):',
  1654. ' raise RuntimeError("hijacked")',
  1655. '__main__._dump_scalar = boom',
  1656. '__main__.os = boom',
  1657. // The fallback must also survive a rebind of its own primitives.
  1658. '__main__._os_write = boom',
  1659. '__main__._memoryview = boom',
  1660. '__main__._FALLBACK_DONE_FRAME = boom',
  1661. 'raise ValueError("real failure")',
  1662. ].join('\n'),
  1663. bindings: [],
  1664. })
  1665. expect(result.error?.kind).toBe('exception')
  1666. expect(result.error?.kind).not.toBe('worker-exit')
  1667. }, 15_000)
  1668. it('still reports a model exception when the program rebinds BaseException', async () => {
  1669. // `_run`'s outer try/except catches the program's failure and builds a
  1670. // `done` frame. The clause previously used the module-global `BaseException`,
  1671. // which the program (running as `__main__`) can rebind: `__main__.BaseException
  1672. // = RuntimeError` makes the `except BaseException` resolve to `RuntimeError`,
  1673. // so a subsequent `ValueError` does not match and escapes `_run` with no
  1674. // `done` frame — misreporting the run as a `worker-exit`. The exception class
  1675. // is now bound into a `_run` LOCAL before the program runs, so the rebind
  1676. // cannot change which class the clause catches; the run must still report an
  1677. // `exception`, not a `worker-exit`.
  1678. const { runtime } = await setup({ maxWallMs: 10_000 })
  1679. const result = await runtime.run({
  1680. program: [
  1681. 'import __main__',
  1682. '__main__.BaseException = RuntimeError',
  1683. 'raise ValueError("real failure")',
  1684. ].join('\n'),
  1685. bindings: [],
  1686. })
  1687. expect(result.error?.kind).toBe('exception')
  1688. expect(result.error?.kind).not.toBe('worker-exit')
  1689. }, 15_000)
  1690. it('still reports the exception when BaseException and the traceback reporter are rebound together', async () => {
  1691. // The two rebind families compose: `__main__.BaseException = ValueError`
  1692. // must not change which class the `_run` catch resolves (it is a pre-program
  1693. // local), and a rebound reporter (`_SAFE_MODEL_TRACEBACK`/`_cap_message`/
  1694. // `_model_traceback`/`_UNRENDERABLE_DIAGNOSTIC`) must not break the done
  1695. // frame — `safe_model_traceback` holds its primitives as import-time closure
  1696. // cells. A `KeyError` (not a `ValueError` subclass) escapes a catch that
  1697. // resolves to the rebound class, so without the local binding the run would
  1698. // misreport as `worker-exit`; with it, the run reports the exception and the
  1699. // fallback reporter still produces the fixed literal.
  1700. const { runtime } = await setup({ maxWallMs: 10_000 })
  1701. const result = await runtime.run({
  1702. program: [
  1703. 'import __main__',
  1704. 'def boom(*a, **k):',
  1705. ' raise RuntimeError("hijacked")',
  1706. '__main__.BaseException = ValueError',
  1707. '__main__._SAFE_MODEL_TRACEBACK = boom',
  1708. '__main__._cap_message = boom',
  1709. '__main__._model_traceback = boom',
  1710. '__main__._UNRENDERABLE_DIAGNOSTIC = boom',
  1711. 'raise KeyError("real failure")',
  1712. ].join('\n'),
  1713. bindings: [],
  1714. })
  1715. expect(result.error?.kind).toBe('exception')
  1716. expect(result.error?.kind).not.toBe('worker-exit')
  1717. }, 15_000)
  1718. it('rejects an fd-3 frame whose raw length exceeds the parse cap before joining it', async () => {
  1719. // The 64 MiB frame parse cap bounds the RAW frame bytes, not the decoded
  1720. // structure; a compact wide frame near that ceiling could decode to far
  1721. // more host memory. The unframed-buffer counter is checked against
  1722. // FRAME_PARSE_CAP_BYTES BEFORE the Buffer.concat join, so an oversized
  1723. // frame is dropped at one copy of its wire bytes instead of being fully
  1724. // joined (a second copy) and only then discarded in the line loop — the
  1725. // peak-memory doubling the pre-join check exists to prevent. Fail-before:
  1726. // without the check the frame is joined whole and parsed (its log text
  1727. // admitted, truncating the ledger), and the run completes normally.
  1728. const { runtime } = await setup({ maxWallMs: 60_000 })
  1729. const result = await runtime.run({
  1730. program: [
  1731. 'import os',
  1732. // One frame just past the 64 MiB parse cap.
  1733. 'os.write(3, b"{\\"type\\":\\"log\\",\\"text\\":\\"" + b"a" * (65 * 1024 * 1024) + b"\\"}\\n")',
  1734. 'return "done"',
  1735. ].join('\n'),
  1736. bindings: [],
  1737. })
  1738. expect(result.error?.kind).toBe('worker-exit')
  1739. expect(result.error?.message).toContain('protocol frame exceeded')
  1740. }, 90_000)
  1741. it('caps an oversized rejection diagnostic so an invalid completion stays invalid-output', async () => {
  1742. // _done_with_value caps its rejection diagnostic through _cap_message: a
  1743. // hostile class name (huge type(value).__name__) would otherwise push the
  1744. // done frame past the host's 64 MiB parse cap, misreporting an
  1745. // invalid-output run as a worker-exit. The diagnostic is capped to the
  1746. // value budget, so the frame always crosses the parser.
  1747. const { runtime } = await setup({ maxWallMs: 60_000 })
  1748. const result = await runtime.run({
  1749. program: [
  1750. 'return type("N" * (70 * 1024 * 1024), (), {})()',
  1751. ].join('\n'),
  1752. bindings: [],
  1753. })
  1754. expect(result.error?.kind).toBe('invalid-output')
  1755. expect(result.error?.kind).not.toBe('worker-exit')
  1756. }, 90_000)
  1757. it('appends a flushed unterminated line to the next entry without a fake newline', async () => {
  1758. // An explicit flush of an unterminated line (print(..., end='', flush=True))
  1759. // used to push a full log frame, so the following print() landed in a
  1760. // SECOND entry and logs.join('\n') rendered 'a\nb' for what the program
  1761. // printed as one line. The flush frame now carries `open: true` and the
  1762. // host appends the next frame to the same entry.
  1763. const { runtime } = await setup()
  1764. const result = await runtime.run({
  1765. program: [
  1766. "print('a', end='', flush=True)",
  1767. "print('b')",
  1768. 'return "done"',
  1769. ].join('\n'),
  1770. bindings: [],
  1771. })
  1772. expect(result.error).toBeUndefined()
  1773. expect(result.logs).toEqual(['ab'])
  1774. }, 15_000)
  1775. it('keeps a flushed unterminated line when the run ends with it still open', async () => {
  1776. // The settlement flush pushes the residual with `open: true`; finish()
  1777. // admits it so a program that commits a partial line and returns does not
  1778. // lose it from logs.
  1779. const { runtime } = await setup()
  1780. const result = await runtime.run({
  1781. program: [
  1782. "print('committed', end='', flush=True)",
  1783. 'return "done"',
  1784. ].join('\n'),
  1785. bindings: [],
  1786. })
  1787. expect(result.error).toBeUndefined()
  1788. expect(result.logs).toEqual(['committed'])
  1789. }, 15_000)
  1790. it('bounds a forged open-frame flood against the log budget', async () => {
  1791. // The open hold must be bounded by the ledger: without the exact-cost check
  1792. // a forged open flood would grow the held fragment without touching
  1793. // logBudget — unbounded host retention under a small budget. The flood now
  1794. // truncates to the marker like any over-budget log traffic.
  1795. const { runtime } = await setup({ maxLogBytes: 64 })
  1796. const result = await runtime.run({
  1797. program: [
  1798. 'import os',
  1799. // 2000 forged open frames, each under the frame parse cap.
  1800. 'for _ in range(2000):',
  1801. " os.write(3, b'{\"type\":\"log\",\"text\":\"a\",\"open\":true}\\n')",
  1802. 'return "done"',
  1803. ].join('\n'),
  1804. bindings: [],
  1805. })
  1806. expect(result.error).toBeUndefined()
  1807. expect(result.logs).toEqual([logTruncationMarker(64)])
  1808. }, 15_000)
  1809. it('no-ops a closing frame once an open flood already truncated the ledger', async () => {
  1810. // The closing-frame branch's post-truncation arm: an open flood exhausts
  1811. // the ledger (logsTruncated set, marker pushed), then a closing frame
  1812. // arrives — it must be a no-op, not append content past the marker.
  1813. const { runtime } = await setup({ maxLogBytes: 64 })
  1814. const result = await runtime.run({
  1815. program: [
  1816. 'import os',
  1817. 'for _ in range(2000):',
  1818. " os.write(3, b'{\"type\":\"log\",\"text\":\"a\",\"open\":true}\\n')",
  1819. "os.write(3, b'{\"type\":\"log\",\"text\":\"b\"}\\n')",
  1820. 'return "done"',
  1821. ].join('\n'),
  1822. bindings: [],
  1823. })
  1824. expect(result.error).toBeUndefined()
  1825. expect(result.logs).toEqual([logTruncationMarker(64)])
  1826. }, 15_000)
  1827. it('bills a merged open entry once, not per fragment', async () => {
  1828. // A merged entry's wire cost is billed ONCE, split across its fragments
  1829. // (first fragment pays quotes+separator, continuations pay only content).
  1830. // Under maxLogBytes: 64, 16 single-character flushes merge to one 16-char
  1831. // entry (2 quotes + 16 content + 1 separator = 19), which fits; per-
  1832. // fragment billing (each charged quotes+separator, ~4 bytes) would truncate
  1833. // at 16 x 4 = 64.
  1834. const { runtime } = await setup({ maxLogBytes: 64 })
  1835. const result = await runtime.run({
  1836. program: [
  1837. 'for _ in range(16):',
  1838. " print('x', end='', flush=True)",
  1839. "print('')",
  1840. 'return "done"',
  1841. ].join('\n'),
  1842. bindings: [],
  1843. })
  1844. expect(result.error).toBeUndefined()
  1845. expect(result.logs).toEqual(['x'.repeat(16)])
  1846. }, 15_000)
  1847. it('admits a compliant merged entry whose closing frame fits the remaining budget', async () => {
  1848. // The review's arithmetic check: print('a'*30, flush); print('b'*25) under
  1849. // maxLogBytes: 64 has a merged wire cost of 2 quotes + 55 content + 1
  1850. // separator = 58 <= 63, so it MUST be admitted as one entry. The earlier
  1851. // cap math (logBudget - openCost) made the closing frame's walk see a
  1852. // negative cap and truncate a compliant entry.
  1853. const { runtime } = await setup({ maxLogBytes: 64 })
  1854. const result = await runtime.run({
  1855. program: [
  1856. "print('a' * 30, end='', flush=True)",
  1857. "print('b' * 25)",
  1858. 'return "done"',
  1859. ].join('\n'),
  1860. bindings: [],
  1861. })
  1862. expect(result.error).toBeUndefined()
  1863. expect(result.logs).toEqual(['a'.repeat(30) + 'b'.repeat(25)])
  1864. }, 15_000)
  1865. it('rejects an open frame that would overflow the ledger by one byte', async () => {
  1866. // The review's arithmetic check: an open frame whose full JSON cost is 63
  1867. // (maxLogBytes: 64 -> ledger 63) must be rejected by the first-fragment
  1868. // cap logBudget - 1 (62), not admitted with a bill of 64 that pushes the
  1869. // ledger negative. The frame is FORGED on fd 3 so the child ledger cannot
  1870. // truncate first: a reverted cap of logBudget (63) would admit the frame,
  1871. // hold it, and flush it at settlement, so the marker assertion fails.
  1872. const { runtime } = await setup({ maxLogBytes: 64 })
  1873. const result = await runtime.run({
  1874. program: [
  1875. 'import os',
  1876. "os.write(3, ('{\"type\":\"log\",\"text\":\"' + 'x' * 61 + '\",\"open\":true}\\n').encode())",
  1877. 'return "done"',
  1878. ].join('\n'),
  1879. bindings: [],
  1880. })
  1881. expect(result.error).toBeUndefined()
  1882. expect(result.logs).toEqual([logTruncationMarker(64)])
  1883. }, 15_000)
  1884. it('bills the closing frame as the merged tail under an exact-fit budget', async () => {
  1885. // The child's split billing: a 30-char open + a 30-char closing frame cost
  1886. // 2 + 60 + 1 = 63 = ledger 63 exactly; the closing frame must be billed as
  1887. // the merged tail (content only), not as a fresh entry (which would
  1888. // double-charge the quotes+separator and truncate an exact-fit entry).
  1889. const { runtime } = await setup({ maxLogBytes: 64 })
  1890. const result = await runtime.run({
  1891. program: [
  1892. "print('a' * 30, end='', flush=True)",
  1893. "print('b' * 30)",
  1894. 'return "done"',
  1895. ].join('\n'),
  1896. bindings: [],
  1897. })
  1898. expect(result.error).toBeUndefined()
  1899. expect(result.logs).toEqual(['a'.repeat(30) + 'b'.repeat(30)])
  1900. }, 15_000)
  1901. it('does not over-reject an exact-fit closing line while an open entry accumulates', async () => {
  1902. // The write-path pre-check's cheap bound used +3 (quotes + separator) even
  1903. // while an open entry was accumulating, so an exact-fit merged TAIL was
  1904. // truncated. The recipe below goes through the SCAN pre-check (the
  1905. // newline-terminated write arrives with an empty pending buffer, so the
  1906. // buffered-chunks branch is skipped): 'a'*29 flush bills 32 (ledger 31
  1907. // left), then one write of 'b'*30 + newline merges 30 more chars whose
  1908. // cheap bound is 30, not 33 — the +3 form saw 30 + 3 = 33 > 31, sliced to
  1909. // a budget prefix, and pushed past the ledger, emitting the marker for a
  1910. // line that fits (merged cost 2 + 59 + 1 = 62 <= 63).
  1911. const { runtime } = await setup({ maxLogBytes: 64 })
  1912. const result = await runtime.run({
  1913. program: [
  1914. 'import sys',
  1915. "sys.stdout.write('a' * 29)",
  1916. 'sys.stdout.flush()',
  1917. "sys.stdout.write('b' * 30 + chr(10))",
  1918. 'return "done"',
  1919. ].join('\n'),
  1920. bindings: [],
  1921. })
  1922. expect(result.error).toBeUndefined()
  1923. expect(result.logs).toEqual(['a'.repeat(29) + 'b'.repeat(30)])
  1924. }, 15_000)
  1925. it('rejects a new open entry once the ledger has only two bytes left', async () => {
  1926. // The jsonStringCostUpTo sub-2-byte guard: forged open frames drive the
  1927. // host ledger down to 1 byte, then a new open entry's first-fragment cap
  1928. // (logBudget - 1 = 0) trips the guard and truncates.
  1929. const { runtime } = await setup({ maxLogBytes: 64 })
  1930. const result = await runtime.run({
  1931. program: [
  1932. 'import os',
  1933. "os.write(3, ('{\"type\":\"log\",\"text\":\"' + 'a' * 28 + '\",\"open\":true}\\n').encode())",
  1934. "os.write(3, ('{\"type\":\"log\",\"text\":\"' + 'a' * 31 + '\"}\\n').encode())",
  1935. "os.write(3, b'{\"type\":\"log\",\"text\":\"x\",\"open\":true}\\n')",
  1936. 'return "done"',
  1937. ].join('\n'),
  1938. bindings: [],
  1939. })
  1940. expect(result.error).toBeUndefined()
  1941. expect(result.logs).toEqual(['a'.repeat(59), logTruncationMarker(64)])
  1942. }, 15_000)
  1943. it('truncates when the closing frame of a merged entry overflows the budget', async () => {
  1944. // The merged entry's billed-once cost: an open fragment that nearly
  1945. // exhausts the budget, then a closing frame whose content no longer fits —
  1946. // the closing frame's exact-cost walk trips and the marker replaces the
  1947. // entry, exactly like any other over-budget log traffic.
  1948. const { runtime } = await setup({ maxLogBytes: 64 })
  1949. const result = await runtime.run({
  1950. program: [
  1951. "print('x' * 40, end='', flush=True)",
  1952. "print('y' * 40)",
  1953. 'return "done"',
  1954. ].join('\n'),
  1955. bindings: [],
  1956. })
  1957. expect(result.error).toBeUndefined()
  1958. expect(result.logs).toEqual([logTruncationMarker(64)])
  1959. }, 15_000)
  1960. it('keeps a float completion exact when the program mutates the decimal context', async () => {
  1961. // The float encoder's Decimal(repr(value)).normalize() used the process
  1962. // GLOBAL decimal context: a legitimate program setting
  1963. // `getcontext().prec = 2` silently rounded the completion value's digits,
  1964. // and `traps[Inexact] = True` made the encode raise, misclassifying a
  1965. // successful run as an exception. A fixed module-level Context(prec=28)
  1966. // makes the spelling decision context-independent.
  1967. const { runtime } = await setup()
  1968. const result = await runtime.run({
  1969. program: [
  1970. 'from decimal import getcontext',
  1971. 'getcontext().prec = 2',
  1972. 'getcontext().traps[__import__("decimal").Inexact] = True',
  1973. 'return 1.2345678901234567',
  1974. ].join('\n'),
  1975. bindings: [],
  1976. })
  1977. expect(result.error).toBeUndefined()
  1978. expect(result.value).toBe(1.2345678901234567)
  1979. }, 15_000)
  1980. it('bounds an over-cap exception-group nesting on the copy', async () => {
  1981. // Exception groups link through `exceptions`, not the cause/context
  1982. // dunders, so the cap has to count that edge too — otherwise a deeply
  1983. // nested group walks past the bound the marker claims to enforce.
  1984. const { runtime } = await setup({ maxValueBytes: 1024 * 1024, maxWallMs: 15_000 })
  1985. const result = await runtime.run({
  1986. program: [
  1987. 'group = ValueError("leaf")',
  1988. 'for i in range(150):',
  1989. ' group = ExceptionGroup(f"g{i}", [group])',
  1990. 'raise group',
  1991. ].join('\n'),
  1992. bindings: [],
  1993. })
  1994. expect(result.error?.kind).toBe('exception')
  1995. expect(result.error?.message).toContain('exception chain truncated at 100 links')
  1996. }, 20_000)
  1997. it('filters every bootstrap frame from the traceback of an uncaught binding rejection', async () => {
  1998. // A rejection re-raised by the bootstrap's dispatch adds bootstrap frames
  1999. // AFTER the model's own; only <model> frames may reach model-visible,
  2000. // durable output — a bootstrap.py path would leak host absolutes and make
  2001. // transcripts machine-dependent.
  2002. const { runtime } = await setup()
  2003. const result = await runtime.run({
  2004. program: 'await tools.boom({})',
  2005. bindings: tools({ boom: async () => { throw new Error('exploded') } }),
  2006. })
  2007. expect(result.error?.kind).toBe('exception')
  2008. expect(result.error?.message).toContain('exploded')
  2009. expect(result.error?.message).toContain('<model>')
  2010. expect(result.error?.message).not.toContain('bootstrap.py')
  2011. })
  2012. it('renders a non-Error thrown value from a host binding as its String form', async () => {
  2013. const { runtime } = await setup()
  2014. const result = await runtime.run({
  2015. program: [
  2016. 'caught = ""',
  2017. 'try:',
  2018. ' await tools.failRaw({})',
  2019. 'except RuntimeError as e:',
  2020. ' caught = str(e)',
  2021. 'return caught',
  2022. ].join('\n'),
  2023. bindings: tools({
  2024. failRaw: async () => { throw 'raw-nope' },
  2025. }),
  2026. })
  2027. expect(result.error).toBeUndefined()
  2028. expect(result.value).toContain('raw-nope')
  2029. })
  2030. it('reassembles a frame split across writes behind a completed one', async () => {
  2031. // One os.write carrying "<frame>\n<partial...>" leaves a non-empty
  2032. // residual after the newline loop; the tail must survive until its own
  2033. // newline arrives and then parse as a normal frame.
  2034. const { runtime } = await setup()
  2035. const result = await runtime.run({
  2036. program: [
  2037. 'import os, json',
  2038. 'head = json.dumps({"type":"log","text":"first"}).encode()',
  2039. 'tail = json.dumps({"type":"log","text":"second"}).encode()',
  2040. 'import time',
  2041. 'os.write(3, head + b"\\n" + tail[:5])',
  2042. 'time.sleep(0.2)',
  2043. 'os.write(3, tail[5:] + b"\\n")',
  2044. 'return "ok"',
  2045. ].join('\n'),
  2046. bindings: [],
  2047. })
  2048. expect(result.error).toBeUndefined()
  2049. expect(result.value).toBe('ok')
  2050. expect(result.logs).toContain('first')
  2051. expect(result.logs).toContain('second')
  2052. })
  2053. it('raises the declared errorClass with the member name on rejection', async () => {
  2054. // Code Mode declares { name: ToolCallError, memberNameProperty: toolName };
  2055. // a host rejection must surface as that class, carrying the failed tool.
  2056. const { runtime } = await setup()
  2057. const result = await runtime.run({
  2058. program: [
  2059. 'caught = ""',
  2060. 'try:',
  2061. ' await tools.fail({})',
  2062. 'except ToolCallError as e:',
  2063. ' caught = f"{type(e).__name__}:{e.toolName}:{e}"',
  2064. 'return caught',
  2065. ].join('\n'),
  2066. bindings: [{
  2067. global: 'tools',
  2068. functions: { fail: async () => { throw new Error('typed-nope') } },
  2069. errorClass: { name: 'ToolCallError', memberNameProperty: 'toolName' },
  2070. }],
  2071. })
  2072. expect(result.error).toBeUndefined()
  2073. expect(result.value).toBe('ToolCallError:fail:typed-nope')
  2074. })
  2075. it('keeps the declared error class catching when Exception and setattr are rebound', async () => {
  2076. // _make_error_class's minted __init__ def-time captures Exception and
  2077. // setattr, so a program rebinding __main__.Exception/__main__.setattr
  2078. // cannot break the rejection constructor: `except ToolCallError` must still
  2079. // catch and read the member property.
  2080. const { runtime } = await setup()
  2081. const result = await runtime.run({
  2082. program: [
  2083. 'import __main__',
  2084. 'def boom(*a, **k):',
  2085. ' raise RuntimeError("hijacked")',
  2086. '__main__.Exception = boom',
  2087. '__main__.setattr = boom',
  2088. 'caught = ""',
  2089. 'try:',
  2090. ' await tools.fail({})',
  2091. 'except ToolCallError as e:',
  2092. ' caught = f"{type(e).__name__}:{e.toolName}"',
  2093. 'return caught',
  2094. ].join('\n'),
  2095. bindings: [{
  2096. global: 'tools',
  2097. functions: { fail: async () => { throw new Error('typed-nope') } },
  2098. errorClass: { name: 'ToolCallError', memberNameProperty: 'toolName' },
  2099. }],
  2100. })
  2101. expect(result.error).toBeUndefined()
  2102. expect(result.value).toBe('ToolCallError:fail')
  2103. }, 15_000)
  2104. it('rejects an errorClass name colliding with its namespace global at the seam', async () => {
  2105. const { runtime } = await setup()
  2106. await expect(runtime.run({
  2107. program: 'return 1',
  2108. bindings: [{ global: 'tools', functions: {}, errorClass: { name: 'tools', memberNameProperty: 'toolName' } }],
  2109. })).rejects.toThrow(/collides with another injected global/)
  2110. })
  2111. it('rejects a namespace global colliding with a runtime-owned name at the seam', async () => {
  2112. // `__dsh_main__` passes the identifier check, but exec()ing the generated
  2113. // wrapper would silently overwrite the binding after injection. `console`
  2114. // is the WORKER backend's slot — refused here too so a namespace list
  2115. // valid on one backend is valid on all.
  2116. const { runtime } = await setup()
  2117. // `__debug__` is refused for a different reason than a collision: CPython
  2118. // compiles a bare `__debug__` reference to the constant True and refuses to
  2119. // assign the name at compile time, so an injected global under it is
  2120. // unreachable from the program — accepted by the seam, unusable here.
  2121. for (const global of ['__dsh_main__', 'console', '__debug__']) {
  2122. await expect(runtime.run({
  2123. program: 'x = 1',
  2124. bindings: [{ global, functions: {} }],
  2125. })).rejects.toThrow(/collides with a runtime-owned global/)
  2126. }
  2127. })
  2128. it('accepts a non-identifier memberNameProperty and rejects only an empty one', async () => {
  2129. // The seam permits any non-empty own property except the reserved
  2130. // members; Python setattr/getattr carry exotic names like `tool-name`,
  2131. // and the worker backend accepts them, so this backend must too.
  2132. const { runtime } = await setup()
  2133. const result = await runtime.run({
  2134. program: [
  2135. 'try:',
  2136. ' await tools.boom({})',
  2137. 'except ToolCallError as e:',
  2138. ' return getattr(e, "tool-name")',
  2139. ].join('\n'),
  2140. bindings: [{
  2141. global: 'tools',
  2142. functions: { boom: async () => { throw new Error('nope') } },
  2143. errorClass: { name: 'ToolCallError', memberNameProperty: 'tool-name' },
  2144. }],
  2145. })
  2146. expect(result.error).toBeUndefined()
  2147. expect(result.value).toBe('boom')
  2148. await expect(runtime.run({
  2149. program: 'return 1',
  2150. bindings: [{ global: 'tools', functions: {}, errorClass: { name: 'ToolCallError', memberNameProperty: '' } }],
  2151. })).rejects.toThrow(/memberNameProperty must be a non-empty attribute name/)
  2152. })
  2153. it('resolves a basename pythonBin to an absolute path (runs a real program)', async () => {
  2154. // A bare `python3` basename must resolve against PATH and actually launch
  2155. // under the empty-env spawn — exercises the accessSync success branch.
  2156. const { runtime } = await setup({ pythonBin: 'python3' })
  2157. const result = await runtime.run({ program: 'return 7', bindings: [] })
  2158. expect(result.error).toBeUndefined()
  2159. expect(result.value).toBe(7)
  2160. })
  2161. it('spawns via an absolute python path resolved from a basename against PATH', async () => {
  2162. // resolvePythonBin turns the default basename into an absolute path before
  2163. // the empty-env spawn; a basename with no PATH match falls through to the
  2164. // normal ENOENT worker-exit rather than throwing.
  2165. const { runtime } = await setup({ pythonBin: 'definitely-no-such-python-xyz' })
  2166. const result = await runtime.run({ program: 'return 1', bindings: [] })
  2167. expect(result.error?.kind).toBe('worker-exit')
  2168. })
  2169. it('rejects a memberNameProperty naming a constrained BaseException attribute', async () => {
  2170. // `__dict__`/`__class__` are constrained descriptors alongside
  2171. // `__traceback__` — setattr of a string raises TypeError while
  2172. // constructing the rejection — so every dunder is refused at the seam.
  2173. const { runtime } = await setup()
  2174. // name/message/stack are the seam's own exclusions (CodeBindingErrorClass
  2175. // forbids replacing them; the worker backend rejects them identically).
  2176. for (const member of ['__traceback__', '__dict__', '__class__', 'args', 'name', 'message', 'stack']) {
  2177. await expect(runtime.run({
  2178. program: 'return 1',
  2179. bindings: [{ global: 'tools', functions: {}, errorClass: { name: 'ToolCallError', memberNameProperty: member } }],
  2180. })).rejects.toThrow(/reserved error member/)
  2181. }
  2182. })
  2183. it('rejects a lossy binding resolution (NaN) instead of coercing it to null', async () => {
  2184. // JSON.stringify would turn NaN into null and drop undefined fields; the
  2185. // seam requires a descriptive rejection so data cannot silently corrupt.
  2186. const { runtime } = await setup()
  2187. const result = await runtime.run({
  2188. program: [
  2189. 'caught = ""',
  2190. 'try:',
  2191. ' await tools.bad({})',
  2192. 'except RuntimeError as e:',
  2193. ' caught = str(e)',
  2194. 'return caught',
  2195. ].join('\n'),
  2196. bindings: tools({ bad: async () => Number.NaN }),
  2197. })
  2198. expect(result.error).toBeUndefined()
  2199. expect(result.value).toContain('lossless JSON')
  2200. })
  2201. it('contains a forged pathological done value without crashing the host', async () => {
  2202. // A ~20k-deep nested array forged onto fd 3 would overflow a recursive
  2203. // JSON.stringify; the host's iterative encoder measures it stack-safely
  2204. // and fails it deterministically on the byte budget (40 kB > 32 KiB).
  2205. const { runtime } = await setup()
  2206. const result = await runtime.run({
  2207. program: [
  2208. 'import os, json',
  2209. 'depth = 20000',
  2210. 'payload = "[" * depth + "]" * depth',
  2211. 'os.write(3, b\'{"type":"done","value":\' + payload.encode() + b\'}\\n\')',
  2212. 'import time',
  2213. 'time.sleep(5)',
  2214. ].join('\n'),
  2215. bindings: [],
  2216. })
  2217. expect(result.value).toBeUndefined()
  2218. expect(result.error?.kind).toBe('output-limit')
  2219. })
  2220. it('preserves a deeply nested completion value below the byte budget', async () => {
  2221. // CodeJsonValue has no depth limit: a 10000-deep nested list is only
  2222. // ~20 kB — under maxValueBytes — and must cross intact. That depth
  2223. // overflows BOTH recursive serializers the pipeline used to rely on
  2224. // (CPython's json.dumps recursion limit ~1000s, V8's JSON.stringify), so
  2225. // it proves the child-side _encode_json_plain and the host-side
  2226. // encodeJsonPlain together. The host JSON.parse of the frame is iterative
  2227. // in V8 for arrays, so only the two encoders were at risk.
  2228. const { runtime } = await setup()
  2229. const result = await runtime.run({
  2230. program: [
  2231. 'v = None',
  2232. 'for _ in range(10000):',
  2233. ' v = [v]',
  2234. 'return v',
  2235. ].join('\n'),
  2236. bindings: [],
  2237. })
  2238. expect(result.error).toBeUndefined()
  2239. // Walk down iteratively (a recursive toEqual would itself overflow).
  2240. let depth = 0
  2241. let cursor: unknown = result.value
  2242. while (Array.isArray(cursor)) {
  2243. expect(cursor).toHaveLength(1)
  2244. cursor = cursor[0]
  2245. depth++
  2246. }
  2247. expect(depth).toBe(10000)
  2248. expect(cursor).toBeNull()
  2249. })
  2250. it('bridges a deeply nested binding resolution back into the program stack-safely', async () => {
  2251. // A binding resolution has no seam-level depth or byte cap; neither the
  2252. // host's reply serialization nor the CHILD's reply decode may die on
  2253. // recursion (json.loads raises RecursionError ~10k levels deep; the
  2254. // bootstrap decodes frames iteratively). 12000 levels sits past that
  2255. // limit while staying tiny in bytes.
  2256. const { runtime } = await setup()
  2257. const deep = ((): unknown => {
  2258. let v: unknown = null
  2259. for (let i = 0; i < 12000; i++) v = [v]
  2260. return v
  2261. })()
  2262. const result = await runtime.run({
  2263. program: [
  2264. 'v = await tools.deep({})',
  2265. 'depth = 0',
  2266. 'while isinstance(v, list):',
  2267. ' v = v[0]',
  2268. ' depth += 1',
  2269. 'return depth',
  2270. ].join('\n'),
  2271. bindings: tools({ deep: async () => deep as never }),
  2272. })
  2273. expect(result.error).toBeUndefined()
  2274. expect(result.value).toBe(12000)
  2275. })
  2276. it('rejects a reserved errorClass name at the seam', async () => {
  2277. const { runtime } = await setup()
  2278. await expect(runtime.run({
  2279. program: 'return 1',
  2280. bindings: [{
  2281. global: 'tools',
  2282. functions: {},
  2283. errorClass: { name: 'class', memberNameProperty: 'toolName' },
  2284. }],
  2285. })).rejects.toThrow(/errorClass.name "class" is not a usable Python identifier/)
  2286. })
  2287. it('routes a declared inherited-attribute name through the bridge via subscript', async () => {
  2288. // __class__ resolves on `object` before any fallback hook; the proxy's
  2289. // __getattribute__ intercepts declared names first, and subscript access
  2290. // is the SDK-advertised route for underscore names.
  2291. const { runtime } = await setup()
  2292. const seen: string[] = []
  2293. const result = await runtime.run({
  2294. program: [
  2295. 'a = await tools["__class__"]({"via": "subscript"})',
  2296. 'b = await tools.__class__({"via": "dot"})',
  2297. 'return [a, b]',
  2298. ].join('\n'),
  2299. bindings: tools({
  2300. '__class__': async () => { seen.push('called'); return 'bridged' },
  2301. }),
  2302. })
  2303. expect(result.error).toBeUndefined()
  2304. expect(result.value).toEqual(['bridged', 'bridged'])
  2305. expect(seen).toEqual(['called', 'called'])
  2306. })
  2307. it('rejects NaN binding arguments immediately instead of hanging', async () => {
  2308. // Default json.dumps would emit a non-standard NaN token that the host
  2309. // JSON.parse drops silently, hanging the call until the wall clock;
  2310. // allow_nan=False raises in-program right away.
  2311. const { runtime } = await setup({ maxWallMs: 8000 })
  2312. const start = Date.now()
  2313. const result = await runtime.run({
  2314. program: [
  2315. 'caught = ""',
  2316. 'try:',
  2317. ' await tools.echo({"x": float("nan")})',
  2318. 'except RuntimeError as e:',
  2319. ' caught = str(e)',
  2320. 'return caught',
  2321. ].join('\n'),
  2322. bindings: tools({ echo: async args => args as CodeJsonValue }),
  2323. })
  2324. expect(result.error).toBeUndefined()
  2325. expect(result.value).toContain('lossless JSON')
  2326. expect(Date.now() - start).toBeLessThan(5000)
  2327. })
  2328. it('carries large binding arguments well past maxValueBytes', async () => {
  2329. // Binding traffic has no seam byte cap: a call frame far larger than the
  2330. // completion budget must reach the host intact (the fd-3 ceiling is a
  2331. // fixed memory-safety bound, not an output budget).
  2332. const maxValueBytes = 4096
  2333. const { runtime } = await setup({ maxValueBytes })
  2334. let receivedLength = 0
  2335. const result = await runtime.run({
  2336. program: [
  2337. `big = "B" * ${maxValueBytes * 50}`,
  2338. 'r = await tools.measure({"payload": big})',
  2339. 'return r',
  2340. ].join('\n'),
  2341. bindings: tools({
  2342. measure: async (args) => {
  2343. receivedLength = ((args as { payload: string }).payload).length
  2344. return receivedLength
  2345. },
  2346. }),
  2347. })
  2348. expect(result.error).toBeUndefined()
  2349. expect(receivedLength).toBe(maxValueBytes * 50)
  2350. expect(result.value).toBe(maxValueBytes * 50)
  2351. })
  2352. it('rejects an unknown binding name inside the program with a matching error', async () => {
  2353. const { runtime } = await setup()
  2354. const result = await runtime.run({
  2355. program: [
  2356. 'caught = ""',
  2357. 'try:',
  2358. ' await tools.nope({})',
  2359. 'except (AttributeError, RuntimeError) as e:',
  2360. ' caught = str(e)',
  2361. 'return caught',
  2362. ].join('\n'),
  2363. bindings: tools({ known: async () => 'ok' }),
  2364. })
  2365. expect(result.error).toBeUndefined()
  2366. expect(result.value).toContain('nope')
  2367. })
  2368. it('bounds an unknown-binding diagnostic built from a forged call frame', async () => {
  2369. // `call.global` and `call.name` carry no byte cap of their own, only the
  2370. // 64 MiB fd-3 frame parse cap, and the reply interpolated them raw: one copy
  2371. // into the template result, one into the `JSON.stringify` escape, one into
  2372. // the `encodeJsonPlain` frame, one into the pipe write. Slicing each field
  2373. // to `maxValueBytes` code units first makes an 8 MiB forged name a
  2374. // 128-byte reply. The observable effect is the reply the child then has to
  2375. // READ: its fd-3 reader is unbuffered, so `readline` consumes an oversized
  2376. // reply one `read(2)` per byte and the run's own legitimate call never gets
  2377. // answered — measured under a 60 s ceiling, the 8 MiB case timed out and a
  2378. // 64 MiB case cost the host 509.9 MiB of heap against 120.3 MiB with the
  2379. // slices in place. The child's address space stays generous enough to BUILD
  2380. // the forgery, which is not what is under test.
  2381. const { runtime } = await setup({ maxValueBytes: 128, addressSpaceMb: 1024, maxWallMs: 20_000 })
  2382. const result = await runtime.run({
  2383. program: [
  2384. 'import os',
  2385. 'frame = b\'{"type":"call","id":9001,"global":"tools","name":"\' + b"n" * (8 * 1024 * 1024) + b\'","args":{}}\\n\'',
  2386. // One os.write returns short past the pipe buffer, and a partial frame
  2387. // would glue itself to the next one and be dropped as malformed, so the
  2388. // forgery goes out through a drain loop.
  2389. 'view = memoryview(frame)',
  2390. 'while view:',
  2391. ' view = view[os.write(3, view):]',
  2392. // A legitimate call after the forgery: its reply can only arrive once
  2393. // the child has read past whatever the forged frame was answered with.
  2394. 'await tools.known({})',
  2395. 'return "settled"',
  2396. ].join('\n'),
  2397. bindings: tools({ known: async () => 'ok' }),
  2398. })
  2399. expect(result.error).toBeUndefined()
  2400. expect(result.value).toBe('settled')
  2401. }, 40_000)
  2402. it('bridges a binding call reached via subscript access (tools["name"])', async () => {
  2403. // The SDK tells the model `await tools["my-tool"](args)` works for exotic
  2404. // names; the proxy's __getitem__ must route it through the bridge.
  2405. const { runtime } = await setup()
  2406. const result = await runtime.run({
  2407. program: [
  2408. 'r = await tools["my-tool"]({"n": 7})',
  2409. 'return r',
  2410. ].join('\n'),
  2411. bindings: tools({ 'my-tool': async args => ({ got: args as CodeJsonValue }) }),
  2412. })
  2413. expect(result.error).toBeUndefined()
  2414. expect(result.value).toEqual({ got: { n: 7 } })
  2415. })
  2416. it('raises KeyError for an undeclared subscript name', async () => {
  2417. const { runtime } = await setup()
  2418. const result = await runtime.run({
  2419. program: [
  2420. 'caught = ""',
  2421. 'try:',
  2422. ' await tools["absent"]({})',
  2423. 'except KeyError as e:',
  2424. ' caught = str(e)',
  2425. 'return caught',
  2426. ].join('\n'),
  2427. bindings: tools({ known: async () => 'ok' }),
  2428. })
  2429. expect(result.error).toBeUndefined()
  2430. expect(result.value).toContain('absent')
  2431. })
  2432. })
  2433. describe('PythonCodeRuntime — budgets, termination, disposal', () => {
  2434. it('kills a wall-clock runaway program via SIGTERM/SIGKILL and reports timeout', async () => {
  2435. const { runtime } = await setup({ maxWallMs: 500, graceMs: 200 })
  2436. const start = Date.now()
  2437. const result = await runtime.run({
  2438. program: 'import time\nwhile True: time.sleep(1)',
  2439. bindings: [],
  2440. })
  2441. const elapsed = Date.now() - start
  2442. // The wall timer may fire first or the exit-after-signal may resolve; both are ok.
  2443. expect(['timeout', 'worker-exit']).toContain(result.error?.kind)
  2444. // We got somewhere in the neighborhood of maxWallMs, not the underlying `sleep(1)`.
  2445. expect(elapsed).toBeLessThan(2000)
  2446. }, 5000)
  2447. it('aborts a run when the outer signal fires mid-flight', async () => {
  2448. const { runtime } = await setup({ maxWallMs: 10_000 })
  2449. const controller = new AbortController()
  2450. const settled: Promise<CodeRunResult> = runtime.run({
  2451. program: 'import time\nwhile True: time.sleep(0.1)',
  2452. bindings: [],
  2453. signal: controller.signal,
  2454. })
  2455. setTimeout(() => { controller.abort('outer-abort') }, 200)
  2456. const result = await settled
  2457. expect(['abort', 'worker-exit']).toContain(result.error?.kind)
  2458. }, 5000)
  2459. it('settles the run when a mid-flight abort reason cannot be converted', async () => {
  2460. // The listener converted the reason before calling `finish()`, so a hostile
  2461. // reason threw from inside an `AbortSignal` listener. Node reports that as an
  2462. // uncaught exception — it can terminate the host — and `finish()` never ran,
  2463. // so the run stayed live until the wall ceiling and misreported as `timeout`
  2464. // (observed) instead of the caller's cancellation. `maxWallMs` is short so
  2465. // that misreport is a fast assertion failure rather than a suite timeout.
  2466. const uncaught: unknown[] = []
  2467. const record = (error: unknown): void => { uncaught.push(error) }
  2468. process.on('uncaughtException', record)
  2469. try {
  2470. const { runtime } = await setup({ maxWallMs: 4_000, graceMs: 200 })
  2471. const controller = new AbortController()
  2472. const settled: Promise<CodeRunResult> = runtime.run({
  2473. program: 'import time\nwhile True: time.sleep(0.1)',
  2474. bindings: [],
  2475. signal: controller.signal,
  2476. })
  2477. setTimeout(() => {
  2478. controller.abort({ [Symbol.toPrimitive]() { throw new Error('reason blew up') } })
  2479. }, 200)
  2480. const result = await settled
  2481. expect(result.error?.kind).toBe('abort')
  2482. expect(result.error?.message).toBe('<unrenderable rejection value>')
  2483. expect(uncaught).toEqual([])
  2484. } finally {
  2485. process.off('uncaughtException', record)
  2486. }
  2487. }, 15_000)
  2488. it('disposes to quiescence: an in-flight run resolves as abort and the child exits', async () => {
  2489. const { fiber, runtime } = await setup({ maxWallMs: 10_000 })
  2490. const pending = runtime.run({
  2491. program: 'import time\nwhile True: time.sleep(0.1)',
  2492. bindings: [],
  2493. })
  2494. // Give the process time to spawn and start running.
  2495. await new Promise(resolve => setTimeout(resolve, 200))
  2496. await fiber.dispose()
  2497. const result = await pending
  2498. expect(['abort', 'worker-exit']).toContain(result.error?.kind)
  2499. }, 5000)
  2500. it('reports a spawn failure via a bogus python binary as worker-exit', async () => {
  2501. const { runtime } = await setup({ pythonBin: '/nonexistent/python-binary', maxWallMs: 3000 })
  2502. const result = await runtime.run({
  2503. program: 'return 1',
  2504. bindings: [],
  2505. })
  2506. expect(result.error?.kind).toBe('worker-exit')
  2507. }, 8000)
  2508. it('applies the strictest of the configured and inherited resource limits', async () => {
  2509. // This case used to drive the bootstrap's `applying resource limits failed`
  2510. // handler with `cpuSeconds: 2 ** 63`, asserting that a cap the child cannot
  2511. // apply fails the run rather than running it uncapped. That premise no longer
  2512. // holds, for two independent reasons, so the test now pins what is actually
  2513. // guaranteed instead of a path no admissible input reaches.
  2514. //
  2515. // First, `2 ** 63` is not a safe integer, so it is now rejected at LOAD as a
  2516. // configuration error — it can never reach the child at all. Second, even the
  2517. // largest admissible values are applied successfully, because `_clamped`
  2518. // bounds every requested pair by the inherited hard limit: an unprivileged
  2519. // process may lower a hard limit but never raise one, so the child keeps the
  2520. // stricter of the two rather than asking for something `setrlimit` refuses.
  2521. // The failure handler remains as a substrate guard (a platform whose kernel
  2522. // refuses the call for its own reasons), but it is no longer reachable from
  2523. // configuration, and a test that pretends otherwise documents a contract the
  2524. // code does not have.
  2525. //
  2526. // What is observable: a very large cap still yields a working run, and the
  2527. // containment it promises is met by the inherited ceiling.
  2528. const { runtime } = await setup({ cpuSeconds: Number.MAX_SAFE_INTEGER - 1, maxWallMs: 10_000 })
  2529. const result = await runtime.run({ program: 'return 1', bindings: [] })
  2530. expect(result.error).toBeUndefined()
  2531. expect(result.value).toBe(1)
  2532. }, 20_000)
  2533. it('settles as worker-exit when the child exits before sending done (no hang)', async () => {
  2534. // Regression: settlement must key off `close` (process reaped AND stdio
  2535. // drained), not `exit`. With `exit`, finish() re-armed a second exit
  2536. // listener that never fired — run() hung forever whenever the exit event
  2537. // beat the final fd-3 data (deterministic on macOS, a lost race elsewhere).
  2538. const { runtime } = await setup({ maxWallMs: 10_000 })
  2539. const result = await runtime.run({
  2540. program: 'import os\nos._exit(7)',
  2541. bindings: [],
  2542. })
  2543. expect(result.error?.kind).toBe('worker-exit')
  2544. expect(result.error?.message).toContain('code=7')
  2545. }, 5000)
  2546. it('classifies RLIMIT_CPU soft-limit expiry (SIGXCPU) as a timeout', async () => {
  2547. // A CPU hot loop burns the soft limit; the kernel delivers SIGXCPU, whose
  2548. // close signal the host maps to `timeout`. macOS re-delivers SIGXCPU
  2549. // differently, so we assert only kind/message here — CI's darwin leg
  2550. // validates real delivery. cpuSeconds must be an integer for setrlimit.
  2551. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 20_000 })
  2552. const result = await runtime.run({
  2553. program: 'while True: pass',
  2554. bindings: [],
  2555. })
  2556. expect(result.error?.kind).toBe('timeout')
  2557. expect(result.error?.message).toContain('CPU time exhausted')
  2558. }, 8000)
  2559. it('keeps an early self-inflicted SIGKILL a worker-exit, not a CPU timeout', async () => {
  2560. // The unsolicited-SIGKILL-as-timeout classification applies only when the
  2561. // CPU budget could have expired (wall time >= cpuSeconds). A SIGKILL
  2562. // seconds before that (cgroup OOM, an operator, os.kill) is substrate
  2563. // death and stays worker-exit per the orthogonal taxonomy.
  2564. const { runtime } = await setup({ cpuSeconds: 60, maxWallMs: 10_000 })
  2565. const result = await runtime.run({
  2566. program: [
  2567. 'import os, signal',
  2568. 'os.kill(os.getpid(), signal.SIGKILL)',
  2569. ].join('\n'),
  2570. bindings: [],
  2571. })
  2572. expect(result.error?.kind).toBe('worker-exit')
  2573. expect(result.error?.message).toContain('SIGKILL')
  2574. })
  2575. it('charges a forked descendant against the run CPU budget', async () => {
  2576. // RLIMIT_CPU is per-process and every child inherits a FRESH budget, so a
  2577. // program that shells out multiplies `cpuSeconds` by the number of
  2578. // descendants it starts. Measured before the aggregate meter existed: with
  2579. // cpuSeconds 1, two sequential busy children burned 2.0 CPU-seconds
  2580. // (RUSAGE_CHILDREN) and the run still returned a SUCCESS completion. The
  2581. // settle-time check meters RUSAGE_SELF + RUSAGE_CHILDREN and converts the
  2582. // overrun into the same SIGXCPU the untrapped soft limit sends.
  2583. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 30_000 })
  2584. const result = await runtime.run({
  2585. program: [
  2586. 'import subprocess, sys',
  2587. 'for _ in range(2):',
  2588. ' subprocess.run([sys.executable, "-c", "import time\\nt=time.time()\\nwhile time.time()-t<1.2: pass"])',
  2589. 'return "escaped the cpu budget"',
  2590. ].join('\n'),
  2591. bindings: [],
  2592. })
  2593. // Darwin's SIGXCPU re-delivery differs, so accept either terminal
  2594. // classification; what must NOT happen is the completion crossing.
  2595. expect(['timeout', 'worker-exit']).toContain(result.error?.kind)
  2596. expect(result.value).toBeUndefined()
  2597. }, 40_000)
  2598. it('does not charge wall time or a cheap descendant against the CPU budget', async () => {
  2599. // The meter is CPU, not wall clock, and it must not fire on a child that
  2600. // burns almost nothing: a sleeping program and a trivial subprocess both
  2601. // have to complete normally, or the check would reject every program that
  2602. // shells out.
  2603. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 30_000 })
  2604. const slept = await runtime.run({
  2605. program: 'import time\ntime.sleep(1.5)\nreturn "slept"',
  2606. bindings: [],
  2607. })
  2608. expect(slept.error).toBeUndefined()
  2609. expect(slept.value).toBe('slept')
  2610. const cheap = await runtime.run({
  2611. program: [
  2612. 'import subprocess, sys',
  2613. 'subprocess.run([sys.executable, "-c", "pass"])',
  2614. 'return "cheap child"',
  2615. ].join('\n'),
  2616. bindings: [],
  2617. })
  2618. expect(cheap.error).toBeUndefined()
  2619. expect(cheap.value).toBe('cheap child')
  2620. }, 40_000)
  2621. it('spends no part of addressSpaceMb on bootstrap machinery', async () => {
  2622. // RLIMIT_AS counts RESERVED address space, so anything the bootstrap maps
  2623. // for its own accounting is subtracted from the program's `addressSpaceMb`.
  2624. // A sampling thread for the descendant-CPU meter cost 72 MiB here (an 8 MiB
  2625. // stack plus a 64 MiB glibc per-thread malloc arena reservation) and turned
  2626. // the 2-million-entry dict rejection below into a MemoryError under a
  2627. // 256 MiB cap on a slower runner. Assert the child's own mappings directly
  2628. // rather than inferring the budget from a near-cap allocation, so the bound
  2629. // is read from /proc instead of from how much headroom one machine happens
  2630. // to have; 48 MiB is well above the ~30 MiB a bare interpreter maps and
  2631. // well below the 102 MiB the thread produced. `addressSpaceMb` itself is
  2632. // skipped on darwin (the dyld shared cache makes any practical cap
  2633. // unsettable) and /proc/self/maps does not exist there, so the mapping
  2634. // assertion is Linux-only; the completion path is checked everywhere.
  2635. const { runtime } = await setup({ maxValueBytes: 4096, addressSpaceMb: 256 })
  2636. const mapped = await runtime.run({
  2637. program: [
  2638. 'import sys',
  2639. 'if sys.platform != "linux":',
  2640. ' return 0',
  2641. 'total = 0',
  2642. 'with open("/proc/self/maps") as handle:',
  2643. ' for line in handle:',
  2644. ' low, high = (int(part, 16) for part in line.split(" ", 1)[0].split("-"))',
  2645. ' total += high - low',
  2646. 'return total // (1024 * 1024)',
  2647. ].join('\n'),
  2648. bindings: [],
  2649. })
  2650. expect(mapped.error).toBeUndefined()
  2651. expect(mapped.value).toBeLessThan(48)
  2652. }, 20_000)
  2653. it('spends no part of addressSpaceMb on the reply pump, across a binding await', async () => {
  2654. // The test above measures BEFORE the program yields, so it could not see the
  2655. // reply pump's cost: `loop.run_in_executor(None, read_frame)` created the
  2656. // default executor's first thread on the first `await tools.*`, and that
  2657. // thread's 8 MiB stack plus a 64 MiB glibc per-thread malloc arena are
  2658. // charged to RLIMIT_AS while the limit is already in force — measured, the
  2659. // child went from 30.34 MiB to 102.39 MiB across one binding call. Under a
  2660. // small `addressSpaceMb` the thread cannot start and a legitimate call hangs
  2661. // to `maxWallMs`; under a larger one an allocation that should have fit dies
  2662. // as MemoryError. `loop.add_reader` watches the fd with no thread at all.
  2663. //
  2664. // Measuring both sides inside one run is what discriminates: a single
  2665. // after-the-fact number cannot separate the pump's cost from the
  2666. // interpreter's own footprint. Linux-only for the same reason as above.
  2667. const { runtime } = await setup({ addressSpaceMb: 256, maxWallMs: 20_000 })
  2668. const result = await runtime.run({
  2669. program: [
  2670. 'import sys',
  2671. 'def mapped():',
  2672. ' if sys.platform != "linux":',
  2673. ' return 0',
  2674. ' total = 0',
  2675. ' with open("/proc/self/maps") as handle:',
  2676. ' for line in handle:',
  2677. ' low, high = (int(part, 16) for part in line.split(" ", 1)[0].split("-"))',
  2678. ' total += high - low',
  2679. ' return total // (1024 * 1024)',
  2680. 'before = mapped()',
  2681. 'echoed = await tools.echo({"ping": True})',
  2682. 'return {"before": before, "after": mapped(), "echoed": echoed}',
  2683. ].join('\n'),
  2684. bindings: tools({ echo: async args => args as CodeJsonValue }),
  2685. })
  2686. expect(result.error).toBeUndefined()
  2687. const value = result.value as { before: number; after: number; echoed: unknown }
  2688. // The binding call really happened, so the pump really ran.
  2689. expect(value.echoed).toEqual({ ping: true })
  2690. // Awaiting a binding maps nothing extra. The 8 MiB allowance absorbs ordinary
  2691. // heap growth while staying far below the 72 MiB a pump thread cost.
  2692. expect(value.after - value.before).toBeLessThan(8)
  2693. }, 30_000)
  2694. it('still terminates a program that ignores SIGXCPU (hard-limit backstop)', async () => {
  2695. // A hot loop under SIG_IGN burns through the soft limit; the kernel's
  2696. // hard limit (cpuSeconds + 1) SIGKILLs it. Only a kernel-authoritative
  2697. // SIGXCPU close classifies as the CPU timeout — a bare SIGKILL is
  2698. // indistinguishable from a cgroup OOM kill, so it reports worker-exit
  2699. // (Darwin re-delivers SIGXCPU instead, where the wall clock settles it
  2700. // as timeout). Either way the run TERMINATES within the budget — the
  2701. // backstop holds even when the classification is the opaque one.
  2702. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 6_000 })
  2703. const result = await runtime.run({
  2704. program: [
  2705. 'import signal',
  2706. 'signal.signal(signal.SIGXCPU, signal.SIG_IGN)',
  2707. 'while True: pass',
  2708. ].join('\n'),
  2709. bindings: [],
  2710. })
  2711. expect(['timeout', 'worker-exit']).toContain(result.error?.kind)
  2712. }, 12_000)
  2713. it('enforces the CPU budget even when the program monkeypatches the enforcement primitives', async () => {
  2714. // The check uses import-time-captured references, so replacing
  2715. // resource.getrusage / signal.signal / os.kill on the modules cannot
  2716. // defang it: a trapping program that also swaps the callables and burns
  2717. // past the budget still dies by the authoritative SIGXCPU.
  2718. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 15_000 })
  2719. const result = await runtime.run({
  2720. program: [
  2721. 'import signal, os, resource, time',
  2722. 'signal.signal(signal.SIGXCPU, lambda *a: None)',
  2723. 'resource.getrusage = lambda *a: (_ for _ in ()).throw(RuntimeError("nope"))',
  2724. 'os.kill = lambda *a: None',
  2725. 'signal.signal = lambda *a: None',
  2726. 'deadline = time.process_time() + 1.05',
  2727. 'while time.process_time() < deadline: pass',
  2728. 'return "escaped"',
  2729. ].join('\n'),
  2730. bindings: [],
  2731. })
  2732. expect(result.error?.kind).toBe('timeout')
  2733. expect(result.value).toBeUndefined()
  2734. }, 15_000)
  2735. it('re-delivers SIGXCPU when a trapping program returns inside the soft-to-hard gap', async () => {
  2736. // A program can trap SIGXCPU and settle during the one-second gap; the
  2737. // bootstrap re-checks the kernel CPU meter (getrusage) after settlement
  2738. // and dies by SIGXCPU with the default disposition restored, so the host
  2739. // still classifies the exhausted budget as a timeout instead of success.
  2740. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 15_000 })
  2741. const result = await runtime.run({
  2742. program: [
  2743. 'import signal, time',
  2744. 'fired = []',
  2745. 'signal.signal(signal.SIGXCPU, lambda *a: fired.append(1))',
  2746. 'deadline = time.process_time() + 1.05',
  2747. 'while time.process_time() < deadline: pass',
  2748. 'return "escaped"',
  2749. ].join('\n'),
  2750. bindings: [],
  2751. })
  2752. expect(result.error?.kind).toBe('timeout')
  2753. expect(result.error?.message).toContain('CPU time exhausted')
  2754. expect(result.value).toBeUndefined()
  2755. }, 15_000)
  2756. it('enforces the CPU budget when the program rebinds the enforcer on __main__', async () => {
  2757. // The bootstrap IS `__main__`, so `import __main__` reaches its globals.
  2758. // The enforcement callable holds its primitives in closure cells (not
  2759. // module attributes) and `_run` reads the callable into a frame local
  2760. // before the program starts, so neither replacing the global nor swapping
  2761. // the module's captured names changes what runs after settlement.
  2762. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 15_000 })
  2763. const result = await runtime.run({
  2764. program: [
  2765. 'import signal, time, __main__',
  2766. 'signal.signal(signal.SIGXCPU, lambda *a: None)',
  2767. '__main__._DIE_IF_CPU_EXHAUSTED = lambda *_: None',
  2768. 'deadline = time.process_time() + 1.05',
  2769. 'while time.process_time() < deadline: pass',
  2770. 'return "escaped"',
  2771. ].join('\n'),
  2772. bindings: [],
  2773. })
  2774. expect(result.error?.kind).toBe('timeout')
  2775. expect(result.value).toBeUndefined()
  2776. }, 15_000)
  2777. it('bounds a program that defeats the post-check by writing its closure cell', async () => {
  2778. // The closure-cell capture raises the cost of defeating the post-check; it
  2779. // does NOT make it unreachable, and nothing in-process could: a cell is
  2780. // writable through `fn.__closure__[i].cell_contents`, and `sys._getframe`
  2781. // reads _run's frame locals. This program does exactly that — walks to
  2782. // _run's frame, takes the enforcement callable, and replaces its captured
  2783. // `getrusage` with one reporting zero CPU used — then burns past cpuSeconds
  2784. // with SIGXCPU trapped. The run must still fail, because the bound that
  2785. // model code cannot forge is outside the interpreter: the RLIMIT_CPU HARD
  2786. // limit at cpuSeconds + 1, whose SIGKILL admits no handler. No success is
  2787. // reportable either way.
  2788. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 20_000 })
  2789. const start = Date.now()
  2790. const result = await runtime.run({
  2791. program: [
  2792. 'import signal, sys, time',
  2793. 'signal.signal(signal.SIGXCPU, lambda *a: None)',
  2794. // Walk out of __dsh_main__ to _run's frame and take its local.
  2795. 'die = None',
  2796. 'depth = 1',
  2797. 'while depth < 12:',
  2798. ' frame = sys._getframe(depth)',
  2799. ' if "die_if_cpu_exhausted" in frame.f_locals:',
  2800. ' die = frame.f_locals["die_if_cpu_exhausted"]',
  2801. ' break',
  2802. ' depth += 1',
  2803. 'assert die is not None, "enforcer not reachable from the frame chain"',
  2804. 'class Zero:',
  2805. ' ru_utime = 0.0',
  2806. ' ru_stime = 0.0',
  2807. 'names = die.__code__.co_freevars',
  2808. 'die.__closure__[names.index("getrusage")].cell_contents = lambda *a: Zero()',
  2809. // Burn well past the soft limit into the hard limit's SIGKILL.
  2810. 'while True: pass',
  2811. ].join('\n'),
  2812. bindings: [],
  2813. })
  2814. // What holds on EVERY platform: the tampering bought no success. The run
  2815. // failed, carried no value, and the reported kind is one of the two
  2816. // kernel-level outcomes — never a completion.
  2817. expect(result.value).toBeUndefined()
  2818. expect(result.error?.kind === 'worker-exit' || result.error?.kind === 'timeout').toBe(true)
  2819. if (process.platform === 'linux') {
  2820. // Linux enforces the RLIMIT_CPU HARD limit at cpuSeconds + 1 promptly, so
  2821. // the CPU bound — not the 20 s wall ceiling — is what stops the program.
  2822. // Its SIGKILL is not SIGXCPU, so the orthogonal-failure taxonomy reports
  2823. // `worker-exit`: a bare SIGKILL is not evidence of CPU burn.
  2824. expect(result.error?.kind).toBe('worker-exit')
  2825. expect(Date.now() - start).toBeLessThan(15_000)
  2826. } else {
  2827. // Darwin does not deliver the hard limit's SIGKILL on the same schedule;
  2828. // observed on the macOS lane, a program that patches the post-check runs
  2829. // to the WALL ceiling instead. The CPU budget is therefore not the
  2830. // binding constraint against a tampering program there — the wall clock
  2831. // is. Asserted rather than skipped so the difference stays visible.
  2832. expect(result.error?.kind).toBe('timeout')
  2833. }
  2834. }, 30_000)
  2835. it('keeps a finished-but-not-closed run live so dispose awaits the child\'s death', async () => {
  2836. // finish() no longer drops the run from `live`; settle() (at close) does.
  2837. // A SIGTERM-trapping program with a small graceMs sits in the grace window
  2838. // after finish() fires — dispose() must not resolve until the SIGKILL
  2839. // backstop actually reaps the child. The program prints its pid (captured
  2840. // as a log even on abort); once dispose() resolves, that pid must be dead
  2841. // (process.kill(pid, 0) throws ESRCH).
  2842. const { fiber, runtime } = await setup({ maxWallMs: 10_000, graceMs: 400 })
  2843. // Deterministic readiness: the program reports its pid through a binding
  2844. // AFTER installing the trap, so dispose cannot race the spawn (a fixed
  2845. // sleep lost that race on slow CI runners — SIGTERM landed pre-trap).
  2846. let reportedPid!: (pid: number) => void
  2847. const trapReady = new Promise<number>((resolve) => { reportedPid = resolve })
  2848. const pending = runtime.run({
  2849. program: [
  2850. 'import signal, time, os',
  2851. 'signal.signal(signal.SIGTERM, lambda *a: None)',
  2852. 'await tools.ready({"pid": os.getpid()})',
  2853. 'while True: time.sleep(0.05)',
  2854. ].join('\n'),
  2855. bindings: tools({
  2856. ready: async (args) => {
  2857. reportedPid((args as { pid: number }).pid)
  2858. return 'ok'
  2859. },
  2860. }),
  2861. })
  2862. const pid = await trapReady
  2863. const start = Date.now()
  2864. await fiber.dispose()
  2865. const elapsed = Date.now() - start
  2866. const result = await pending
  2867. expect(['abort', 'worker-exit', 'timeout']).toContain(result.error?.kind)
  2868. // dispose() returned only after the grace window elapsed (the SIGTERM trap
  2869. // forces the SIGKILL backstop path), proving the run stayed live past finish().
  2870. expect(elapsed).toBeGreaterThanOrEqual(300)
  2871. expect(Number.isInteger(pid) && pid > 0).toBe(true)
  2872. // The child is fully reaped by the time dispose() resolved.
  2873. expect(() => process.kill(pid, 0)).toThrow(/ESRCH/)
  2874. }, 8000)
  2875. it('settles on the decided result even when a setsid-escaped orphan holds stdio open past close', async () => {
  2876. // `close` only fires once every inherited stdio stream drains. A descendant
  2877. // started with start_new_session=True escapes the child's process group, so
  2878. // the SIGTERM/SIGKILL aimed at that group never reaches it; if it inherited
  2879. // our stdout/stderr/fd 3 and outlives the run, `close` would never fire and
  2880. // run() would hang forever. The close-deadline backstop (graceMs + margin)
  2881. // must force settlement on the value the `done` frame already decided.
  2882. const { runtime } = await setup({ graceMs: 100 })
  2883. const start = Date.now()
  2884. const result = await runtime.run({
  2885. program: [
  2886. 'import subprocess, sys',
  2887. // Orphan in a fresh session, inheriting our stdout/stderr/fd 3, alive
  2888. // past the close-deadline so `close` cannot fire on its own. Its own
  2889. // 5 s self-exit is the leak ceiling AND the discriminator: it must stay
  2890. // ABOVE the < 4000 ms upper-bound assertion below, so if the deadline
  2891. // backstop failed to settle, settlement could only come from this
  2892. // self-exit at ~5 s and blow the bound — a sharper signal than the wall
  2893. // ceiling would give.
  2894. 'subprocess.Popen([sys.executable, "-c", "import time; time.sleep(5)"],',
  2895. ' start_new_session=True)',
  2896. 'return "escaped"',
  2897. ].join('\n'),
  2898. bindings: [],
  2899. })
  2900. const elapsed = Date.now() - start
  2901. // The done frame decided the value; the deadline settled it despite the
  2902. // orphan pinning the pipes open.
  2903. expect(result.error).toBeUndefined()
  2904. expect(result.value).toBe('escaped')
  2905. // Settlement waited for the backstop (graceMs + CLOSE_REAP_MARGIN_MS ≈ 2.1s),
  2906. // not the orphan's 5 s self-exit — proving the deadline, not a fallback, fired.
  2907. expect(elapsed).toBeGreaterThanOrEqual(1_500)
  2908. expect(elapsed).toBeLessThan(4_000)
  2909. }, 8000)
  2910. it('flushes a newline-free diagnostic when the closeDeadline forces settlement', async () => {
  2911. // A leader that writes an unterminated diagnostic via `os.write(1, ...)` and
  2912. // then exits, leaving a setsid orphan holding the pipes open, settles through
  2913. // the closeDeadline destroy() path — which fires no `end`. The residual must
  2914. // be flushed before destroy() drops it, or the diagnostic is lost from
  2915. // `logs`. The value is decided by the done frame; the diagnostic must survive.
  2916. const { runtime } = await setup({ graceMs: 100 })
  2917. const result = await runtime.run({
  2918. program: [
  2919. 'import os, subprocess, sys',
  2920. 'os.write(1, b"leader-diagnostic-no-newline")',
  2921. 'subprocess.Popen([sys.executable, "-c", "import time; time.sleep(5)"],',
  2922. ' start_new_session=True)',
  2923. 'return "escaped"',
  2924. ].join('\n'),
  2925. bindings: [],
  2926. })
  2927. expect(result.error).toBeUndefined()
  2928. expect(result.value).toBe('escaped')
  2929. expect(result.logs).toContain('leader-diagnostic-no-newline')
  2930. }, 8000)
  2931. it('closes the child stdin so a program read sees EOF instead of blocking', async () => {
  2932. // The host closes the child's stdin write handle immediately after spawn
  2933. // (the program is an async body that reads nothing from fd 0; a live pipe
  2934. // would hold a host-side handle open past the run). A program that DOES
  2935. // read fd 0 therefore sees EOF at once. Fail-before: with the handle left
  2936. // open and no data written, `sys.stdin.read()` blocks and the run would
  2937. // hang to maxWallMs as a timeout.
  2938. const { runtime } = await setup({ maxWallMs: 8_000 })
  2939. const result = await runtime.run({
  2940. program: [
  2941. 'import sys',
  2942. 'data = sys.stdin.read()',
  2943. 'return "read: " + repr(data)',
  2944. ].join('\n'),
  2945. bindings: [],
  2946. })
  2947. expect(result.error).toBeUndefined()
  2948. expect(result.value).toBe("read: ''")
  2949. }, 15_000)
  2950. it('keeps runtime type annotations as live classes, not PEP 563 strings, when the program reads them', async () => {
  2951. // bootstrap.py imports `from __future__ import annotations`; without
  2952. // dont_inherit=True on compile(), that PEP 563 flag leaks into the program's
  2953. // compiled code and stringifies its type annotations, changing the semantics
  2954. // of a legal program that reads `f.__annotations__` at runtime.
  2955. const { runtime } = await setup()
  2956. const result = await runtime.run({
  2957. program: [
  2958. 'def f(x: int) -> int:',
  2959. ' return x',
  2960. 'return f.__annotations__["x"].__name__',
  2961. ].join('\n'),
  2962. bindings: [],
  2963. })
  2964. expect(result.error).toBeUndefined()
  2965. expect(result.value).toBe('int')
  2966. }, 15_000)
  2967. it('reaps a same-group child that ignores SIGTERM and releases the pipes before close', async () => {
  2968. // The same-group counterpart to the setsid-orphan case above. A descendant
  2969. // left in the child's OWN process group (no setsid, so `kill(-pid)` reaches
  2970. // it) can ignore SIGTERM yet still release the inherited stdout/stderr/fd 3
  2971. // it does not hold — here by giving the Popen child DEVNULL streams and
  2972. // letting close_fds drop fd 3. The leader then writes `done` and exits, its
  2973. // `close` fires because the pipes drained, and settle() runs while that
  2974. // descendant is still alive. settle() then keeps a REF'd poll alive until the
  2975. // grace-window SIGKILL has emptied the whole process group, so the host cannot
  2976. // exit and reparent the survivor to init: no subprocess outlives the fiber.
  2977. //
  2978. // The descendant must have SIG_IGN installed BEFORE the host sends SIGTERM,
  2979. // or it dies from the default SIGTERM whether the fix is present or not — so
  2980. // it writes a readiness marker after trapping and the leader waits for that
  2981. // marker before returning. While alive it bumps a heartbeat file every 50 ms;
  2982. // the test asserts the heartbeat STOPS, which is what "no longer executing"
  2983. // means whether the killed descendant is reaped or lingers as a zombie (a
  2984. // SIGKILL'd process runs no more code either way). It sleeps 30 s as a safety
  2985. // net so a broken fix cannot leak it forever.
  2986. const handoff = await mkdtemp(join(tmpdir(), 'dsh-samegroup-'))
  2987. const readyMarker = join(handoff, 'ready')
  2988. const heartbeat = join(handoff, 'heartbeat')
  2989. const { runtime } = await setup({ maxWallMs: 10_000, graceMs: 300 })
  2990. const result = await runtime.run({
  2991. program: [
  2992. 'import subprocess, sys, os, time',
  2993. `marker = ${JSON.stringify(readyMarker)}`,
  2994. `heartbeat = ${JSON.stringify(heartbeat)}`,
  2995. // Same group (no start_new_session); ignores SIGTERM; holds none of the
  2996. // leader's pipes (DEVNULL std streams, close_fds drops fd 3). It writes
  2997. // the marker (argv[1]) only AFTER the trap is installed — so the leader
  2998. // cannot return, and the host cannot send SIGTERM, before it is ignored —
  2999. // then rewrites the heartbeat (argv[2]) every 50 ms for up to 30 s.
  3000. 'code = ("import signal, sys, time\\n"',
  3001. ' "signal.signal(signal.SIGTERM, signal.SIG_IGN)\\n"',
  3002. ' "open(sys.argv[1], \'w\').close()\\n"',
  3003. ' "end = time.time() + 30\\n"',
  3004. ' "while time.time() < end:\\n"',
  3005. ' " open(sys.argv[2], \'w\').close()\\n"',
  3006. ' " time.sleep(0.05)\\n")',
  3007. 'child = subprocess.Popen([sys.executable, "-c", code, marker, heartbeat],',
  3008. ' stdin=subprocess.DEVNULL,',
  3009. ' stdout=subprocess.DEVNULL,',
  3010. ' stderr=subprocess.DEVNULL)',
  3011. 'deadline = time.time() + 5',
  3012. 'while not os.path.exists(marker) and time.time() < deadline:',
  3013. ' time.sleep(0.02)',
  3014. 'return "spawned"',
  3015. ].join('\n'),
  3016. bindings: [],
  3017. })
  3018. expect(result.error).toBeUndefined()
  3019. expect(result.value).toBe('spawned')
  3020. // The trap really installed before the leader returned, so this is the
  3021. // SIGTERM-ignoring descendant, not one that would have died to the default.
  3022. expect(existsSync(readyMarker)).toBe(true)
  3023. // The grace-window SIGKILL (graceMs 300 + reap margin) empties the group. Once
  3024. // it has, the descendant stops bumping the heartbeat. Poll the heartbeat's
  3025. // mtime: two consecutive reads far enough apart with no change means it is no
  3026. // longer executing — true whether it was reaped or lingers as a zombie, so
  3027. // the assertion holds in a container whose init does not wait() orphans. The
  3028. // window (well under the 30 s self-timeout) proves the SIGKILL did the work.
  3029. const mtime = (): number => { try { return statSync(heartbeat).mtimeMs } catch { return 0 } }
  3030. const stopDeadline = Date.now() + 8_000
  3031. let last = mtime()
  3032. let still = false
  3033. while (Date.now() < stopDeadline) {
  3034. await new Promise(resolve => setTimeout(resolve, 400))
  3035. const now = mtime()
  3036. if (now === last && now !== 0) { still = true; break }
  3037. last = now
  3038. }
  3039. expect(still).toBe(true)
  3040. }, 20_000)
  3041. it('dispose awaits reaping of a same-group survivor from a completed run', async () => {
  3042. // The quiescence contract also holds for a run that ALREADY resolved: the run
  3043. // stays tracked in `live` until its process group is reaped, so a `dispose()`
  3044. // that races a just-returned run() still awaits the survivor rather than
  3045. // snapshotting an empty `live` and returning while it lives. Here the run
  3046. // completes (leaving a SIGTERM-ignoring same-group descendant), then dispose()
  3047. // is called; the heartbeat must be stale BY THE TIME dispose() resolves —
  3048. // proving teardown waited for the reap, not merely that the reap eventually
  3049. // happened.
  3050. const handoff = await mkdtemp(join(tmpdir(), 'dsh-dispose-quiesce-'))
  3051. const readyMarker = join(handoff, 'ready')
  3052. const heartbeat = join(handoff, 'heartbeat')
  3053. const { runtime, fiber } = await setup({ maxWallMs: 10_000, graceMs: 300 })
  3054. const result = await runtime.run({
  3055. program: [
  3056. 'import subprocess, sys, os, time',
  3057. `marker = ${JSON.stringify(readyMarker)}`,
  3058. `heartbeat = ${JSON.stringify(heartbeat)}`,
  3059. 'code = ("import signal, sys, time\\n"',
  3060. ' "signal.signal(signal.SIGTERM, signal.SIG_IGN)\\n"',
  3061. ' "open(sys.argv[1], \'w\').close()\\n"',
  3062. ' "end = time.time() + 30\\n"',
  3063. ' "while time.time() < end:\\n"',
  3064. ' " open(sys.argv[2], \'w\').close()\\n"',
  3065. ' " time.sleep(0.05)\\n")',
  3066. 'child = subprocess.Popen([sys.executable, "-c", code, marker, heartbeat],',
  3067. ' stdin=subprocess.DEVNULL,',
  3068. ' stdout=subprocess.DEVNULL,',
  3069. ' stderr=subprocess.DEVNULL)',
  3070. 'deadline = time.time() + 5',
  3071. 'while not os.path.exists(marker) and time.time() < deadline:',
  3072. ' time.sleep(0.02)',
  3073. 'return "spawned"',
  3074. ].join('\n'),
  3075. bindings: [],
  3076. })
  3077. expect(result.error).toBeUndefined()
  3078. expect(existsSync(readyMarker)).toBe(true)
  3079. // dispose() must not return until the group is reaped. After it resolves, the
  3080. // heartbeat must already be stale: read its mtime, wait past the heartbeat
  3081. // interval, and confirm it did not advance — the descendant is no longer
  3082. // executing (reaped or zombie), so teardown was genuinely quiescent.
  3083. await fiber.dispose()
  3084. const mtime = (): number => { try { return statSync(heartbeat).mtimeMs } catch { return 0 } }
  3085. const afterDispose = mtime()
  3086. // Pin the assertion to a heartbeat that actually ran: mtime() returns 0 when
  3087. // the file never existed, so without this the `toBe` below would pass
  3088. // vacuously (0 === 0) if the survivor never wrote a heartbeat at all.
  3089. expect(afterDispose).toBeGreaterThan(0)
  3090. await new Promise(resolve => setTimeout(resolve, 500))
  3091. expect(mtime()).toBe(afterDispose)
  3092. }, 20_000)
  3093. it('sends SIGKILL at the poll deadline when the event loop was blocked past both timers', async () => {
  3094. // If the host event loop is blocked (a big synchronous computation) from
  3095. // before the group-reap poll was scheduled until after the deadline, both the
  3096. // poll timer and the grace-window SIGKILL timer are overdue when the loop
  3097. // resumes. Node runs the earlier-scheduled poll first, so the SIGKILL timer
  3098. // may not have fired yet. The deadline arm must then send SIGKILL ITSELF
  3099. // rather than cancel the unfired escalation — otherwise a SIGTERM-ignoring
  3100. // same-group survivor is released for good. A synchronous busy-loop after
  3101. // run() resolves reproduces the block deterministically.
  3102. const handoff = await mkdtemp(join(tmpdir(), 'dsh-deadline-'))
  3103. const readyMarker = join(handoff, 'ready')
  3104. const heartbeat = join(handoff, 'heartbeat')
  3105. const graceMs = 300
  3106. const { runtime } = await setup({ maxWallMs: 10_000, graceMs })
  3107. const result = await runtime.run({
  3108. program: [
  3109. 'import subprocess, sys, os, time',
  3110. `marker = ${JSON.stringify(readyMarker)}`,
  3111. `heartbeat = ${JSON.stringify(heartbeat)}`,
  3112. 'code = ("import signal, sys, time\\n"',
  3113. ' "signal.signal(signal.SIGTERM, signal.SIG_IGN)\\n"',
  3114. ' "open(sys.argv[1], \'w\').close()\\n"',
  3115. ' "end = time.time() + 30\\n"',
  3116. ' "while time.time() < end:\\n"',
  3117. ' " open(sys.argv[2], \'w\').close()\\n"',
  3118. ' " time.sleep(0.05)\\n")',
  3119. 'child = subprocess.Popen([sys.executable, "-c", code, marker, heartbeat],',
  3120. ' stdin=subprocess.DEVNULL,',
  3121. ' stdout=subprocess.DEVNULL,',
  3122. ' stderr=subprocess.DEVNULL)',
  3123. 'deadline = time.time() + 5',
  3124. 'while not os.path.exists(marker) and time.time() < deadline:',
  3125. ' time.sleep(0.02)',
  3126. 'return "spawned"',
  3127. ].join('\n'),
  3128. bindings: [],
  3129. })
  3130. expect(result.error).toBeUndefined()
  3131. expect(existsSync(readyMarker)).toBe(true)
  3132. // Block the event loop synchronously past graceMs + CLOSE_REAP_MARGIN_MS
  3133. // (2000) with margin, so both timers are overdue when the loop resumes.
  3134. const blockUntil = Date.now() + graceMs + 2_000 + 800
  3135. while (Date.now() < blockUntil) { /* busy-wait, no yield */ }
  3136. // Yield: the overdue poll runs (group still non-empty, deadline passed) and
  3137. // must send SIGKILL itself. The survivor then stops bumping the heartbeat.
  3138. const mtime = (): number => { try { return statSync(heartbeat).mtimeMs } catch { return 0 } }
  3139. const stopDeadline = Date.now() + 5_000
  3140. let last = mtime()
  3141. let stopped = false
  3142. while (Date.now() < stopDeadline) {
  3143. await new Promise(resolve => setTimeout(resolve, 400))
  3144. const now = mtime()
  3145. if (now === last && now !== 0) { stopped = true; break }
  3146. last = now
  3147. }
  3148. expect(stopped).toBe(true)
  3149. }, 20_000)
  3150. })
  3151. describe('PythonCodeRuntime — hostile peer', () => {
  3152. it('drops garbage bytes and unknown-shape frames posted directly to fd 3', async () => {
  3153. // The model program can reach fd 3 and write anything. We inject a
  3154. // non-JSON line, a valid JSON but unknown-shape frame, and a broken done
  3155. // frame; the host must not crash, and the real `done` still settles the run.
  3156. const { runtime } = await setup()
  3157. const result = await runtime.run({
  3158. program: [
  3159. 'import os',
  3160. 'os.write(3, b"not-json\\n")',
  3161. 'os.write(3, b\'{"type":"unknown"}\\n\')',
  3162. 'os.write(3, b\'{"type":"done","error":{"message":42}}\\n\')',
  3163. 'return "survived"',
  3164. ].join('\n'),
  3165. bindings: [],
  3166. })
  3167. expect(result.error).toBeUndefined()
  3168. expect(result.value).toBe('survived')
  3169. })
  3170. it('answers a forged call frame for an unknown binding and never crashes', async () => {
  3171. // The unknown-binding reply path, driven through the id the host expects:
  3172. // the program lets its own first call claim id 0 and forges id 1, which the
  3173. // host answers with the `unknown binding` rejection the honest call would
  3174. // have received. A forged id out of sequence is dropped instead — that is
  3175. // the id-bound test below, not this one.
  3176. const { runtime } = await setup({ maxWallMs: 8_000 })
  3177. let seenLegitCall = false
  3178. const result = await runtime.run({
  3179. program: [
  3180. 'import os, json',
  3181. 'x = await tools.echo({"ping": True})',
  3182. 'os.write(3, json.dumps({"type":"call","id":1,"global":"tools","name":"forged","args":{}}).encode() + b"\\n")',
  3183. // The forged frame is answered, but nothing in the child awaits id 1, so
  3184. // the reply is ignored and the run completes on its own value.
  3185. 'return x',
  3186. ].join('\n'),
  3187. bindings: tools({
  3188. echo: async (args) => { seenLegitCall = true; return args as CodeJsonValue },
  3189. }),
  3190. })
  3191. expect(result.error).toBeUndefined()
  3192. expect(result.value).toEqual({ ping: true })
  3193. expect(seenLegitCall).toBe(true)
  3194. }, 15_000)
  3195. it('drops forged call frames whose ids are not the next in sequence, retaining no per-id state', async () => {
  3196. // The host used to remember every answered id in a Set, so a program could
  3197. // write an unbounded run of unique forged ids — each frame far below the
  3198. // 64 MiB cap, so nothing rejected them — and grow host memory for the
  3199. // whole run. Ids are consecutive from 0, so one counter replaces the set.
  3200. //
  3201. // The discriminator is that the forgeries must not be answered. Each names a
  3202. // binding that does exist, so a host answering them would run `echo` once
  3203. // per forgery; the count proves only the legitimate call was dispatched.
  3204. // Ids also run DESCENDING, so a high-water-mark test would drop the honest
  3205. // call that follows rather than the forgeries.
  3206. const { runtime } = await setup()
  3207. let echoCalls = 0
  3208. const result = await runtime.run({
  3209. program: [
  3210. 'import os, json',
  3211. 'for i in range(2000, 0, -1):',
  3212. ' os.write(3, json.dumps({"type":"call","id":i,"global":"tools","name":"echo","args":{"forged":i}}).encode() + b"\\n")',
  3213. 'x = await tools.echo({"ping": True})',
  3214. 'return x',
  3215. ].join('\n'),
  3216. bindings: tools({
  3217. echo: async (args) => { echoCalls += 1; return args as CodeJsonValue },
  3218. }),
  3219. })
  3220. expect(result.error).toBeUndefined()
  3221. expect(result.value).toEqual({ ping: true })
  3222. expect(echoCalls).toBe(1)
  3223. }, 15_000)
  3224. it('keeps answering calls a program makes after one with unserializable arguments', async () => {
  3225. // The child claims an id only once its write succeeds, so a call rejected
  3226. // child-side for non-lossless arguments leaves no gap. Were a gap possible,
  3227. // the host's exact-successor test would drop every later call and the run
  3228. // would hang to the wall ceiling instead of completing.
  3229. const { runtime } = await setup({ maxWallMs: 8_000 })
  3230. const seen: unknown[] = []
  3231. const result = await runtime.run({
  3232. program: [
  3233. 'caught = ""',
  3234. 'try:',
  3235. ' await tools.echo({"bad": float("inf")})',
  3236. 'except RuntimeError as e:',
  3237. ' caught = str(e)',
  3238. 'after = await tools.echo({"ok": True})',
  3239. 'return {"caught": caught, "after": after}',
  3240. ].join('\n'),
  3241. bindings: tools({
  3242. echo: async (args) => { seen.push(args); return args as CodeJsonValue },
  3243. }),
  3244. })
  3245. expect(result.error).toBeUndefined()
  3246. const value = result.value as { caught: string; after: unknown }
  3247. expect(value.caught).toContain('lossless JSON')
  3248. expect(value.after).toEqual({ ok: true })
  3249. // The rejected call never reached the host; the one after it did.
  3250. expect(seen).toEqual([{ ok: true }])
  3251. }, 15_000)
  3252. it('drops a forged frame carrying an integer outside JavaScript safe range', async () => {
  3253. // JSON.parse would silently round 9007199254740993 to ...992 BEFORE any
  3254. // validation, corrupting a dispatched argument or completion. The host
  3255. // scans the raw line and drops such frames as hostile traffic; the honest
  3256. // child cannot produce one (its validator rejects unsafe ints).
  3257. const { runtime } = await setup()
  3258. let dispatched: unknown
  3259. const result = await runtime.run({
  3260. program: [
  3261. 'import os',
  3262. // Forged call frame with an unsafe int argument, then a forged done
  3263. // frame with an unsafe int value — both must be dropped whole.
  3264. 'os.write(3, b\'{"type":"call","id":7,"global":"tools","name":"echo","args":9007199254740993}\\n\')',
  3265. 'os.write(3, b\'{"type":"done","value":9007199254740993}\\n\')',
  3266. 'x = await tools.echo({"ok": True})',
  3267. 'return x',
  3268. ].join('\n'),
  3269. bindings: tools({
  3270. echo: async (args) => { dispatched = args; return args as CodeJsonValue },
  3271. }),
  3272. })
  3273. expect(result.error).toBeUndefined()
  3274. // The forged done did not settle the run; the legit call and completion did.
  3275. expect(result.value).toEqual({ ok: true })
  3276. expect(dispatched).toEqual({ ok: true })
  3277. })
  3278. it('truncates host-side logs once the budget is exhausted and emits the marker', async () => {
  3279. // Set a tiny host-side budget; the Python side has a much larger one, so
  3280. // its LogBuffer will not truncate — the host ledger fires first.
  3281. const { runtime } = await setup({ maxLogBytes: 128 })
  3282. const result = await runtime.run({
  3283. program: [
  3284. 'for _ in range(50):',
  3285. ' print("aaaaaaaaaa")',
  3286. 'return "done"',
  3287. ].join('\n'),
  3288. bindings: [],
  3289. })
  3290. expect(result.error).toBeUndefined()
  3291. const markers = result.logs.filter(line => line.includes('log capture truncated at 128 bytes'))
  3292. expect(markers.length).toBeGreaterThanOrEqual(1)
  3293. })
  3294. it('reports an exception whose message holds an unpaired surrogate instead of stranding to the wall clock', async () => {
  3295. // A strict UTF-8 encode of "\ud800" throws while BUILDING the failure
  3296. // frame; the run would then hang to maxWallMs and misreport as timeout.
  3297. const { runtime } = await setup({ maxWallMs: 8_000 })
  3298. const result = await runtime.run({
  3299. program: String.raw`raise Exception("bad \ud800 surrogate")`,
  3300. bindings: [],
  3301. })
  3302. expect(result.error?.kind).toBe('exception')
  3303. expect(result.error?.message).toContain('bad')
  3304. expect(result.error?.message).toContain('surrogate')
  3305. })
  3306. it('carries a lone-surrogate completion string across the wire as its JSON escape', async () => {
  3307. // UTF-8 has no encoding for a lone surrogate, but JSON does: the ASCII
  3308. // `\ud800` escape, which JSON.parse reads back as the same UTF-16 code
  3309. // unit. `CodeJsonValue`, `snapshotJsonValue`, and the worker backend all
  3310. // accept such a string, so this backend must not narrow the shared seam.
  3311. const { runtime } = await setup()
  3312. const result = await runtime.run({
  3313. program: String.raw`return {"lone": "a\ud800b", "spelled": "😀"}`,
  3314. bindings: [],
  3315. })
  3316. expect(result.error).toBeUndefined()
  3317. // The lone half survives as the code unit itself; a spelled-out high-low
  3318. // PAIR folds into the astral character the host would hold for it.
  3319. expect(result.value).toEqual({ lone: 'a\ud800b', spelled: '\u{1f600}' })
  3320. })
  3321. it('meters a lone surrogate at its six escaped bytes, matching the host', async () => {
  3322. // The child and the host share maxValueBytes, so the child must charge the
  3323. // escape's six ASCII bytes (plus two quotes): eight fits, nine does not.
  3324. const { runtime } = await setup({ maxValueBytes: 8 })
  3325. const ok = await runtime.run({ program: String.raw`return "\ud800"`, bindings: [] })
  3326. expect(ok.error).toBeUndefined()
  3327. expect(ok.value).toBe('\ud800')
  3328. const over = await setup({ maxValueBytes: 7 })
  3329. const result = await over.runtime.run({ program: String.raw`return "\ud800"`, bindings: [] })
  3330. expect(result.error?.kind).toBe('output-limit')
  3331. })
  3332. it('passes a lone-surrogate binding argument through instead of failing the call', async () => {
  3333. // The argument validator shared the same over-narrow rejection; a host
  3334. // binding must receive the code unit the program passed.
  3335. const seen: unknown[] = []
  3336. const { runtime } = await setup()
  3337. const result = await runtime.run({
  3338. program: String.raw`return await tools.echo({"text": "x\udfff"})`,
  3339. bindings: tools({ echo: async (args: unknown) => { seen.push(args); return args as CodeJsonValue } }),
  3340. })
  3341. expect(result.error).toBeUndefined()
  3342. expect(seen).toEqual([{ text: 'x\udfff' }])
  3343. expect(result.value).toEqual({ text: 'x\udfff' })
  3344. })
  3345. it('meters a non-ASCII completion in UTF-8 JSON bytes, matching the host', async () => {
  3346. // json.dumps' default \uXXXX escaping would count "é" as 8 bytes while
  3347. // the host meter counts its UTF-8 JSON form (4); the shared budget must
  3348. // agree, so a 4-byte-fitting value passes a maxValueBytes of 4.
  3349. const { runtime } = await setup({ maxValueBytes: 4 })
  3350. const ok = await runtime.run({ program: 'return "é"', bindings: [] })
  3351. expect(ok.error).toBeUndefined()
  3352. expect(ok.value).toBe('é')
  3353. const over = await runtime.run({ program: 'return "éx"', bindings: [] })
  3354. expect(over.error?.kind).toBe('output-limit')
  3355. })
  3356. it('filters bootstrap frames from exception-group members (TaskGroup)', async () => {
  3357. // Python 3.11+ stores member stacks under TracebackException.exceptions;
  3358. // the <model>-frame filter must recurse into them too.
  3359. const { runtime } = await setup()
  3360. const result = await runtime.run({
  3361. program: [
  3362. 'import asyncio, sys',
  3363. 'if sys.version_info < (3, 11):',
  3364. ' raise ValueError("skip-old <model>")',
  3365. 'async def boom():',
  3366. ' raise ValueError("group-member")',
  3367. 'async with asyncio.TaskGroup() as tg:',
  3368. ' tg.create_task(boom())',
  3369. ].join('\n'),
  3370. bindings: [],
  3371. })
  3372. expect(result.error?.kind).toBe('exception')
  3373. expect(result.error?.message).toContain('<model>')
  3374. expect(result.error?.message).not.toContain('bootstrap.py')
  3375. })
  3376. it('keeps frames intact when a model thread floods logs while a large frame drains', async () => {
  3377. // os.write releases the GIL and a frame beyond PIPE_BUF is not atomic:
  3378. // without the writer lock + full-write loop, the printing thread could
  3379. // interleave bytes mid-frame and the host would drop the malformed JSON,
  3380. // hanging the run to the wall clock (or losing the completion).
  3381. const { runtime } = await setup({ maxValueBytes: 1024 * 1024, maxLogBytes: 4 * 1024 * 1024, maxWallMs: 15_000 })
  3382. const result = await runtime.run({
  3383. program: [
  3384. 'import threading',
  3385. 'stop = False',
  3386. 'def spam():',
  3387. ' while not stop:',
  3388. ' print("spam-line-" + "y" * 100)',
  3389. 't = threading.Thread(target=spam)',
  3390. 't.start()',
  3391. // A ~300 KiB completion — several PIPE_BUF units — while spam runs.
  3392. 'big = "x" * (300 * 1024)',
  3393. 'stop = True',
  3394. 't.join()',
  3395. 'return big',
  3396. ].join('\n'),
  3397. bindings: [],
  3398. })
  3399. expect(result.error).toBeUndefined()
  3400. expect(result.value).toBe('x'.repeat(300 * 1024))
  3401. }, 20_000)
  3402. it('settles cleanly while a daemon thread keeps writing unterminated log text', async () => {
  3403. // WARNING regression: the settlement `flush_out()/flush_err()` on the main
  3404. // coroutine read and clear `_LogStream._pending` and the shared LogBuffer
  3405. // ledger with NO lock, while a model daemon thread's `print`/`write` mutate
  3406. // the same state. Capturing the bound method (`out_stream.flush_line`) only
  3407. // fixes WHICH callable runs, not what it reads mid-flight: the flush could
  3408. // interleave with a concurrent write and join a `_pending` list being
  3409. // mutated under it, corrupting the ledger and costing the `done` frame — the
  3410. // run would then strand to the wall clock instead of completing. The shared
  3411. // re-entrant lock serializes them.
  3412. //
  3413. // A pure data race has no single bad input to reject deterministically, so
  3414. // this maximizes overlap: daemon threads emit UNTERMINATED writes (which
  3415. // pile into `_pending` rather than flushing per line) right up to the moment
  3416. // the body returns and settlement flushes. Repeated so the interleave lands.
  3417. for (let attempt = 0; attempt < 5; attempt++) {
  3418. const { runtime, fiber } = await setup({ maxLogBytes: 4 * 1024 * 1024, maxWallMs: 15_000 })
  3419. const result = await runtime.run({
  3420. program: [
  3421. 'import sys, threading',
  3422. 'stop = False',
  3423. 'def spam():',
  3424. ' while not stop:',
  3425. // No newline: the text accumulates in the stream's `_pending`, which is
  3426. // exactly the state the settlement flush also touches.
  3427. ' sys.stdout.write("tail-fragment-" + "z" * 64)',
  3428. 'workers = [threading.Thread(target=spam, daemon=True) for _ in range(4)]',
  3429. 'for t in workers: t.start()',
  3430. // Let the daemons build up pending writes, then return so settlement
  3431. // flushes while they are still mid-write.
  3432. 'import time; time.sleep(0.05)',
  3433. 'return "settled"',
  3434. ].join('\n'),
  3435. bindings: [],
  3436. })
  3437. expect(result.error).toBeUndefined()
  3438. expect(result.value).toBe('settled')
  3439. await fiber.dispose()
  3440. }
  3441. }, 30_000)
  3442. it('completes a binding called from a worker thread on its own event loop', async () => {
  3443. // A binding reply Future is created on the loop that ran `dispatch`. When the
  3444. // model calls a binding from a worker THREAD via `asyncio.run(tools.x(...))`,
  3445. // that Future belongs to the thread's loop, not the main loop where
  3446. // `_pump_replies` reads the reply. `asyncio.Future` is not thread-safe:
  3447. // completing it from another thread does not wake its own loop, so a direct
  3448. // `set_result` would strand the awaiting thread and the run would degrade to a
  3449. // wall-clock timeout. The pump must schedule completion on the Future's own
  3450. // loop via `call_soon_threadsafe`. The tight maxWallMs makes the pre-fix
  3451. // failure a fast timeout rather than a hang.
  3452. //
  3453. // The main coroutine yields with `await asyncio.sleep` while the worker runs,
  3454. // rather than a synchronous `t.join()`: joining would block the main thread,
  3455. // so the main loop could not run `_pump_replies` and the call would deadlock
  3456. // regardless of the fix — that blocks the pump, not the cross-loop delivery
  3457. // this test pins.
  3458. const { runtime } = await setup({ maxWallMs: 8_000 })
  3459. const seen: unknown[] = []
  3460. const result = await runtime.run({
  3461. program: [
  3462. 'import asyncio, threading',
  3463. 'result = {}',
  3464. 'def worker():',
  3465. // A fresh loop in this thread; the binding Future is created here.
  3466. ' result["value"] = asyncio.run(tools.echo({"from": "thread"}))',
  3467. 't = threading.Thread(target=worker)',
  3468. 't.start()',
  3469. 'while t.is_alive():',
  3470. ' await asyncio.sleep(0.02)',
  3471. 'return result["value"]',
  3472. ].join('\n'),
  3473. bindings: tools({
  3474. echo: async (args) => { seen.push(args); return args as CodeJsonValue },
  3475. }),
  3476. })
  3477. expect(result.error).toBeUndefined()
  3478. expect(result.value).toEqual({ from: 'thread' })
  3479. // The host binding actually ran (the reply round-tripped), not a timeout.
  3480. expect(seen).toEqual([{ from: 'thread' }])
  3481. }, 15_000)
  3482. it('keeps the reply pump alive when a late reply targets a closed thread loop', async () => {
  3483. // A binding called from a worker thread that ABANDONS the call (its
  3484. // `asyncio.run` is cancelled) leaves the pending entry holding that thread's
  3485. // loop, which `asyncio.run` closes on return. When the host later answers
  3486. // that call, `_pump_replies` schedules the completion onto the closed loop —
  3487. // `call_soon_threadsafe` raises `RuntimeError('Event loop is closed')`.
  3488. // Unguarded, that RuntimeError ends the pump task and strands every later
  3489. // reply; the guard drops the moot reply and keeps the pump serving.
  3490. //
  3491. // The ordering is a STRUCTURAL guarantee, not a timing window: the worker
  3492. // closes its loop before the main coroutine signals `closed`; the host
  3493. // answers the abandoned `slow` call (hitting the closed loop) before it
  3494. // answers `release`, because `release`'s handler only resolves `slow` first
  3495. // and then yields a microtask. So the pump provably meets the closed loop on
  3496. // `slow`'s reply before it must deliver `release`'s. Fail-before: the pump
  3497. // dies on `slow`, `release`'s reply is never read, and `await tools.release`
  3498. // hangs to the (small) maxWallMs as a timeout.
  3499. let releaseSlow!: () => void
  3500. const slowGate = new Promise<void>((resolve) => { releaseSlow = resolve })
  3501. const { runtime } = await setup({ maxWallMs: 6_000 })
  3502. const result = await runtime.run({
  3503. program: [
  3504. 'import asyncio, threading',
  3505. 'closed = threading.Event()',
  3506. 'def worker():',
  3507. ' async def body():',
  3508. // Abandon the call: wait_for cancels it, but the pending host-side entry
  3509. // survives (dispatch does not pop on cancellation), holding this loop.
  3510. ' try:',
  3511. ' await asyncio.wait_for(tools.slow({}), timeout=0.1)',
  3512. ' except asyncio.TimeoutError:',
  3513. ' pass',
  3514. ' asyncio.run(body())', // closes the thread's loop on return
  3515. ' closed.set()',
  3516. 't = threading.Thread(target=worker)',
  3517. 't.start()',
  3518. 'while not closed.is_set():',
  3519. ' await asyncio.sleep(0.02)',
  3520. // The loop is closed. Now the host answers slow (dead-loop reply) then
  3521. // release; the pump must survive the first to deliver the second.
  3522. 'after = await tools.release({})',
  3523. 'return after',
  3524. ].join('\n'),
  3525. bindings: tools({
  3526. slow: async () => {
  3527. // Answer only once the worker has closed its loop AND the main
  3528. // coroutine is awaiting release, so this reply reaches the pump against
  3529. // the closed loop.
  3530. await slowGate
  3531. return 'late'
  3532. },
  3533. release: async () => {
  3534. // Let slow's reply be written first, then yield a microtask so the
  3535. // pump processes the dead-loop reply before release's own reply lands.
  3536. releaseSlow()
  3537. await new Promise(resolve => setImmediate(resolve))
  3538. return 'released'
  3539. },
  3540. }),
  3541. })
  3542. expect(result.error).toBeUndefined()
  3543. // The pump survived the closed-loop reply and delivered the later binding.
  3544. expect(result.value).toBe('released')
  3545. }, 15_000)
  3546. it('keeps the reply pump alive when RuntimeError is rebound before the program runs', async () => {
  3547. // `_pump_replies` catches a closed-loop scheduling failure with `except
  3548. // _RuntimeError`. If that name were bound as a pump BODY local, it would be
  3549. // captured at pump-start — but `_run` reaches the model's top-level
  3550. // statements (which run before the pump's first step, since there is no
  3551. // suspension point between `create_task` and `await __dsh_main__`) with the
  3552. // rebind already applied, so `_RuntimeError` would capture the REBOUND class
  3553. // and the closed-loop `RuntimeError` would escape, killing the pump. Binding
  3554. // it as a DEF-TIME default argument captures the original before any model
  3555. // code runs. This rebinds `__main__.RuntimeError` as the very first program
  3556. // statement and drives the closed-loop worker pattern: the pump must survive
  3557. // the dead-loop reply and deliver the later binding.
  3558. let releaseSlow!: () => void
  3559. const slowGate = new Promise<void>((resolve) => { releaseSlow = resolve })
  3560. const { runtime } = await setup({ maxWallMs: 6_000 })
  3561. const result = await runtime.run({
  3562. program: [
  3563. 'import __main__',
  3564. '__main__.RuntimeError = ValueError',
  3565. 'import asyncio, threading',
  3566. 'closed = threading.Event()',
  3567. 'def worker():',
  3568. ' async def body():',
  3569. ' try:',
  3570. ' await asyncio.wait_for(tools.slow({}), timeout=0.1)',
  3571. ' except asyncio.TimeoutError:',
  3572. ' pass',
  3573. ' asyncio.run(body())',
  3574. ' closed.set()',
  3575. 't = threading.Thread(target=worker)',
  3576. 't.start()',
  3577. 'while not closed.is_set():',
  3578. ' await asyncio.sleep(0.02)',
  3579. 'after = await tools.release({})',
  3580. 'return after',
  3581. ].join('\n'),
  3582. bindings: tools({
  3583. slow: async () => { await slowGate; return 'late' },
  3584. release: async () => { releaseSlow(); await new Promise(resolve => setImmediate(resolve)); return 'released' },
  3585. }),
  3586. })
  3587. expect(result.error).toBeUndefined()
  3588. expect(result.value).toBe('released')
  3589. }, 15_000)
  3590. it('keeps a successful completion when _done_with_value is rebound', async () => {
  3591. // `_run` calls `_done_with_value(value, max_value_bytes)` after the program
  3592. // returns. The name is a module global, and this bootstrap IS `__main__`, so
  3593. // `__main__._done_with_value = boom` as a program statement would otherwise
  3594. // be resolved at call time and a legitimate success would be rewritten into
  3595. // an `exception`. `_run` now binds `done_with_value_bound = _done_with_value`
  3596. // before the program runs, so the entry name is immune; the run must still
  3597. // report the success value.
  3598. const { runtime } = await setup({ maxWallMs: 10_000 })
  3599. const result = await runtime.run({
  3600. program: [
  3601. 'import __main__',
  3602. 'def boom(*a, **k):',
  3603. ' raise RuntimeError("hijacked")',
  3604. '__main__._done_with_value = boom',
  3605. 'return 1',
  3606. ].join('\n'),
  3607. bindings: [],
  3608. })
  3609. expect(result.error).toBeUndefined()
  3610. expect(result.value).toBe(1)
  3611. }, 15_000)
  3612. it('round-trips an exactly representable large integer through a binding echo', async () => {
  3613. // The reply serializer must print BigInt digits for a beyond-safe
  3614. // integral double: String(2**60) emits a rounded form, and the child
  3615. // would receive a DIFFERENT integer than the binding resolved.
  3616. const { runtime } = await setup()
  3617. const result = await runtime.run({
  3618. program: [
  3619. 'v = await tools.echo(2**60)',
  3620. 'return v == 2**60',
  3621. ].join('\n'),
  3622. bindings: tools({ echo: async args => args as never }),
  3623. })
  3624. expect(result.error).toBeUndefined()
  3625. expect(result.value).toBe(true)
  3626. })
  3627. it('preserves an exactly representable large integer and rejects a rounding one', async () => {
  3628. // The canonical boundary accepts every JS-double-exact value: 2**53 and
  3629. // 2**60 round-trip exactly and must cross (matching the worker backend);
  3630. // 2**53+1 rounds and must fail as invalid-output.
  3631. const { runtime } = await setup()
  3632. const exact = await runtime.run({ program: 'return [2**53, 2**60]', bindings: [] })
  3633. expect(exact.error).toBeUndefined()
  3634. expect(exact.value).toEqual([2 ** 53, 2 ** 60])
  3635. const lossy = await runtime.run({ program: 'return 2**53 + 1', bindings: [] })
  3636. expect(lossy.error?.kind).toBe('invalid-output')
  3637. expect(lossy.error?.message).toContain('not exactly representable')
  3638. })
  3639. it('rejects a container subclass whose overridden methods hide its contents', async () => {
  3640. // A dict subclass returning [] from items() passes an isinstance check but
  3641. // serializes as {}, so the host would receive a value the program did not
  3642. // compute. Exact-type matching fails it as invalid-output instead. The
  3643. // worker backend rejects the prototype-equivalent shapes the same way.
  3644. const { runtime } = await setup()
  3645. const hidden = await runtime.run({
  3646. program: [
  3647. 'class Sneaky(dict):',
  3648. ' def items(self): return []',
  3649. ' def keys(self): return []',
  3650. ' def __iter__(self): return iter([])',
  3651. ' def __len__(self): return 0',
  3652. 'return Sneaky(secret="kept")',
  3653. ].join('\n'),
  3654. bindings: [],
  3655. })
  3656. expect(hidden.error?.kind).toBe('invalid-output')
  3657. expect(hidden.error?.message).toContain('unsupported type (Sneaky)')
  3658. // A list subclass is refused on the same rule.
  3659. const listish = await runtime.run({
  3660. program: ['class L(list):', ' def __iter__(self): return iter([])', 'return L([1, 2, 3])'].join('\n'),
  3661. bindings: [],
  3662. })
  3663. expect(listish.error?.kind).toBe('invalid-output')
  3664. expect(listish.error?.message).toContain('unsupported type (L)')
  3665. // The exact built-in containers still cross unchanged.
  3666. const plain = await runtime.run({ program: 'return {"secret": [1, 2]}', bindings: [] })
  3667. expect(plain.error).toBeUndefined()
  3668. expect(plain.value).toEqual({ secret: [1, 2] })
  3669. })
  3670. it('rejects a scalar subclass whose overrides disagree with what gets serialized', async () => {
  3671. // The validators checked scalars with isinstance, so a subclass passed
  3672. // every check by its real value while the ENCODER read an override — the
  3673. // host then received a value the walk never approved. Each case below is a
  3674. // distinct override reaching a distinct reader.
  3675. const { runtime } = await setup()
  3676. // _dump_float spells a float from repr(value), so an overridden __repr__
  3677. // decides the digits: F(2.5) serialized as 1.
  3678. const floated = await runtime.run({
  3679. program: [
  3680. 'class F(float):',
  3681. ' def __repr__(self): return "1.0"',
  3682. 'return F(2.5)',
  3683. ].join('\n'),
  3684. bindings: [],
  3685. })
  3686. expect(floated.error?.kind).toBe('invalid-output')
  3687. expect(floated.error?.message).toContain('unsupported type (F)')
  3688. // The JS-safe-range bound is two comparisons, so overriding them admits an
  3689. // int whose true digits (json.dumps reads the C-level value) the host's
  3690. // JSON.parse rounds: 9007199254740993 arrives as ...992.
  3691. const inted = await runtime.run({
  3692. program: [
  3693. 'class I(int):',
  3694. ' def __gt__(self, other): return False',
  3695. ' def __lt__(self, other): return False',
  3696. 'return I(2 ** 53 + 1)',
  3697. ].join('\n'),
  3698. bindings: [],
  3699. })
  3700. expect(inted.error?.kind).toBe('invalid-output')
  3701. expect(inted.error?.message).toContain('unsupported type (I)')
  3702. // The pre-encode size bound reads len(), so overriding it to 0 admits a
  3703. // string of any length past maxValueBytes.
  3704. const stringed = await runtime.run({
  3705. program: [
  3706. 'class S(str):',
  3707. ' def __len__(self): return 0',
  3708. 'return S("Q" * 100000)',
  3709. ].join('\n'),
  3710. bindings: [],
  3711. })
  3712. expect(stringed.error?.kind).toBe('invalid-output')
  3713. expect(stringed.error?.message).toContain('unsupported type (S)')
  3714. // A str-subclass dict KEY reaches the same len() bound.
  3715. const keyed = await runtime.run({
  3716. program: [
  3717. 'class S(str):',
  3718. ' def __len__(self): return 0',
  3719. 'return {S("Q" * 100000): 1}',
  3720. ].join('\n'),
  3721. bindings: [],
  3722. })
  3723. expect(keyed.error?.kind).toBe('invalid-output')
  3724. expect(keyed.error?.message).toContain('non-string dict key (S)')
  3725. // bool is an int subclass that IS lossless JSON, and the exact scalars all
  3726. // still cross unchanged.
  3727. const plain = await runtime.run({
  3728. program: 'return {"t": True, "f": False, "n": None, "i": 7, "d": 2.5, "s": "ok"}',
  3729. bindings: [],
  3730. })
  3731. expect(plain.error).toBeUndefined()
  3732. expect(plain.value).toEqual({ t: true, f: false, n: null, i: 7, d: 2.5, s: 'ok' })
  3733. })
  3734. it('rejects a scalar subclass passed as a binding argument', async () => {
  3735. // The uncapped binding-argument validator shares the exact-type rule, so
  3736. // the call fails through its rejection contract instead of dispatching a
  3737. // float whose digits come from an override.
  3738. const { runtime } = await setup()
  3739. const seen: CodeJsonValue[] = []
  3740. const result = await runtime.run({
  3741. program: [
  3742. 'class F(float):',
  3743. ' def __repr__(self): return "1.0"',
  3744. 'try:',
  3745. ' await tools.echo({"v": F(2.5)})',
  3746. 'except Exception as exc:',
  3747. ' return str(exc)',
  3748. ].join('\n'),
  3749. bindings: tools({ echo: async (args) => {
  3750. seen.push(args as CodeJsonValue)
  3751. return null
  3752. } }),
  3753. })
  3754. expect(result.error).toBeUndefined()
  3755. expect(result.value).toContain('unsupported type (F)')
  3756. expect(seen).toEqual([])
  3757. })
  3758. it('rejects a container subclass passed as a binding argument', async () => {
  3759. // Binding arguments run the uncapped validator, which must apply the same
  3760. // exact-type rule: the call fails descriptively instead of dispatching a
  3761. // value whose serialization disagrees with what was validated.
  3762. const { runtime } = await setup()
  3763. const seen: CodeJsonValue[] = []
  3764. const result = await runtime.run({
  3765. program: [
  3766. 'class Sneaky(dict):',
  3767. ' def items(self): return []',
  3768. 'try:',
  3769. ' await tools.echo(Sneaky(secret="kept"))',
  3770. 'except Exception as exc:',
  3771. ' return str(exc)',
  3772. ].join('\n'),
  3773. bindings: tools({ echo: async (args) => {
  3774. seen.push(args as CodeJsonValue)
  3775. return null
  3776. } }),
  3777. })
  3778. expect(result.error).toBeUndefined()
  3779. expect(result.value).toContain('unsupported type (Sneaky)')
  3780. expect(seen).toEqual([])
  3781. })
  3782. it('fails an oversized completion as output-limit without materializing its encoding', async () => {
  3783. // A 100 MiB string under maxValueBytes: 1024 must fail as output-limit.
  3784. // The address-space cap leaves room for the program to BUILD the string
  3785. // (one copy + interpreter) but not for the old full pre-check encode,
  3786. // which materialized chunk fragments plus the joined copy (~2 more
  3787. // copies) and died on RLIMIT_AS as MemoryError/worker-exit.
  3788. const { runtime } = await setup({ maxValueBytes: 1024, addressSpaceMb: 384, maxWallMs: 15_000 })
  3789. const result = await runtime.run({
  3790. program: 'return "x" * (100 * 1024 * 1024)',
  3791. bindings: [],
  3792. })
  3793. expect(result.error?.kind).toBe('output-limit')
  3794. expect(result.error?.message).toContain('exceeded 1024 bytes')
  3795. }, 20_000)
  3796. it('rejects a control-heavy oversized completion on its length, not its escaped copy', async () => {
  3797. // Every "\x00" escapes to the six bytes "\u0000", so the escaped form of a
  3798. // 40 MB string is ~240 MB. The walk must refuse on the cheap
  3799. // `len(current) + 2` lower bound; the 384 MiB address space holds the raw
  3800. // string but not its escaped expansion, so a pre-escape check dies on
  3801. // RLIMIT_AS instead of returning output-limit.
  3802. const { runtime } = await setup({ maxValueBytes: 1024, addressSpaceMb: 384, maxWallMs: 15_000 })
  3803. const result = await runtime.run({
  3804. program: 'return "\\x00" * (40 * 1024 * 1024)',
  3805. bindings: [],
  3806. })
  3807. expect(result.error?.kind).toBe('output-limit')
  3808. expect(result.error?.message).toContain('exceeded 1024 bytes')
  3809. }, 20_000)
  3810. it('truncates a single print far above maxLogBytes instead of dying on the encode', async () => {
  3811. // LogBuffer must reject via the cheap char-count lower bound BEFORE
  3812. // UTF-8-encoding the whole string: the full encode of a ~100 MB line
  3813. // would double the allocation and can breach RLIMIT_AS. 256 MiB
  3814. // address space comfortably holds one copy of the 100 MB string but
  3815. // not the pre-fix double allocation plus interpreter overhead spikes.
  3816. const { runtime } = await setup({ maxLogBytes: 1024, addressSpaceMb: 256, maxWallMs: 15_000 })
  3817. const result = await runtime.run({
  3818. program: [
  3819. 'print("x" * (100 * 1024 * 1024))',
  3820. 'return "done"',
  3821. ].join('\n'),
  3822. bindings: [],
  3823. })
  3824. expect(result.error).toBeUndefined()
  3825. expect(result.value).toBe('done')
  3826. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  3827. }, 20_000)
  3828. it('stops host capture at the child ledger truncation, keeping exactly one marker', async () => {
  3829. // The two ledgers exhaust independently. One child entry larger than
  3830. // `maxLogBytes` sends ONLY the marker, so the host budget is still nearly
  3831. // untouched — and the marker used to arrive as an ordinary `log` frame the
  3832. // host could not tell from program output. Text written afterwards was
  3833. // therefore retained AFTER the marker, contradicting the stop-after-
  3834. // truncation contract, and a later host-side exhaustion could append a
  3835. // second marker. The frame now carries `truncated: true`.
  3836. //
  3837. // `os.write(1, ...)` bypasses the child's own stream, so those bytes reach
  3838. // the host as stray stdout and take the host ledger path rather than the
  3839. // child's — which is exactly the route that leaked past the marker.
  3840. const { runtime } = await setup({ maxLogBytes: 64, maxWallMs: 10_000 })
  3841. const result = await runtime.run({
  3842. program: [
  3843. 'import os',
  3844. 'print("y" * 70000)',
  3845. 'os.write(1, b"AFTER")',
  3846. 'return "done"',
  3847. ].join('\n'),
  3848. bindings: [],
  3849. })
  3850. expect(result.error).toBeUndefined()
  3851. expect(result.value).toBe('done')
  3852. const markers = result.logs.filter(line => line.includes('log capture truncated'))
  3853. expect(markers).toHaveLength(1)
  3854. // The marker is the LAST entry: nothing was retained after truncation.
  3855. expect(result.logs.at(-1)).toBe(markers[0])
  3856. expect(result.logs.join('\n')).not.toContain('AFTER')
  3857. }, 20_000)
  3858. it('keeps one marker when a program forges repeated truncation frames', async () => {
  3859. // `truncated` is attacker-reachable: the program owns fd 3 and can write the
  3860. // flag itself, so the field is a hostile input rather than a trusted signal.
  3861. // Repeats must collapse to the single marker the contract promises, and only
  3862. // the literal `true` counts — a forged `"yes"` is rebuilt away by
  3863. // validateChildFrame, so that frame stays ordinary text.
  3864. const { runtime } = await setup({ maxLogBytes: 4096, maxWallMs: 10_000 })
  3865. const result = await runtime.run({
  3866. program: [
  3867. 'import os, json',
  3868. 'os.write(3, json.dumps({"type":"log","text":"first","truncated":"yes"}).encode() + b"\\n")',
  3869. 'os.write(3, json.dumps({"type":"log","text":"MARK-A","truncated":True}).encode() + b"\\n")',
  3870. 'os.write(3, json.dumps({"type":"log","text":"MARK-B","truncated":True}).encode() + b"\\n")',
  3871. 'return "done"',
  3872. ].join('\n'),
  3873. bindings: [],
  3874. })
  3875. expect(result.error).toBeUndefined()
  3876. expect(result.value).toBe('done')
  3877. // The non-boolean flag did not truncate, so its text was captured normally.
  3878. expect(result.logs).toContain('first')
  3879. // The first genuine flag stopped capture and emitted the HOST's own marker;
  3880. // the frame's own text is discarded, so neither payload appears.
  3881. expect(result.logs).not.toContain('MARK-A')
  3882. expect(result.logs).not.toContain('MARK-B')
  3883. expect(result.logs.at(-1)).toBe(logTruncationMarker(4096))
  3884. expect(result.logs.filter(line => line.includes('log capture truncated'))).toHaveLength(1)
  3885. }, 20_000)
  3886. it('discards the text of a forged truncation frame instead of retaining it', async () => {
  3887. // The marker branch bypasses `admit`, so retaining the frame's own text put
  3888. // attacker-controlled bytes into `logs` with no cap at all: measured, a 1 MiB
  3889. // forged text was retained whole under `maxLogBytes: 64`, and the only bound
  3890. // left was the 64 MiB frame parse cap. The host emits its own marker instead,
  3891. // so the retained size is fixed regardless of what the program sent.
  3892. const forgedBytes = 1024 * 1024
  3893. const { runtime } = await setup({ maxLogBytes: 64, maxWallMs: 20_000 })
  3894. const result = await runtime.run({
  3895. program: [
  3896. 'import os, json',
  3897. `big = "A" * ${forgedBytes}`,
  3898. 'os.write(3, json.dumps({"type":"log","truncated":True,"text":big}).encode() + b"\\n")',
  3899. 'return "done"',
  3900. ].join('\n'),
  3901. bindings: [],
  3902. })
  3903. expect(result.error).toBeUndefined()
  3904. expect(result.value).toBe('done')
  3905. // Only the host marker is kept, so the total stays orders of magnitude below
  3906. // what the forgery carried — and below the cap it was trying to escape.
  3907. expect(result.logs).toEqual([logTruncationMarker(64)])
  3908. expect(result.logs.join('').length).toBeLessThan(forgedBytes / 1000)
  3909. }, 30_000)
  3910. it('coalesces unframed fd-3 fragments without recopying the sealed prefix', async () => {
  3911. // The frame ceiling meters payload BYTES, but each retained chunk is its own
  3912. // Buffer with object and backing-store overhead the byte count cannot see:
  3913. // 5000 single-byte newline-free writes produced 5000 chunks holding 5031
  3914. // bytes, so a program pacing such writes could accumulate millions of objects
  3915. // inside the wall budget and exhaust the host heap far below 256 MiB.
  3916. //
  3917. // The observable behavior is that the run still completes normally: the
  3918. // fragments are coalesced rather than rejected, since a slow trickle of bytes
  3919. // is not itself a protocol violation.
  3920. //
  3921. // `Buffer.concat` is wrapped for the duration so the cumulative copy volume
  3922. // is measured rather than inferred: that total is what separates sealing into
  3923. // blocks from re-merging the whole buffer, and both shapes pass every
  3924. // behavioral assertion below.
  3925. //
  3926. // The trickle is terminated with its own newline before the real frame is
  3927. // written. Without that, those 5000 bytes prefix the frame on the SAME line,
  3928. // which then parses as junk and is dropped — correct framing behavior, but it
  3929. // would leave this test asserting the wrong thing.
  3930. // Bound at capture: `Buffer.concat` is a static method, and taking a bare
  3931. // reference to one trips no-unbound-method.
  3932. const realConcat = Buffer.concat.bind(Buffer)
  3933. let copied = 0
  3934. Buffer.concat = (list: readonly Uint8Array[], total?: number): Buffer<ArrayBuffer> => {
  3935. for (const part of list) copied += part.length
  3936. return realConcat(list, total)
  3937. }
  3938. const program = [
  3939. 'import os',
  3940. // Newline-free single-byte writes, spaced so each lands as its own read.
  3941. // 60000 rather than 5000: the trickle has to cross the seal threshold
  3942. // enough times for the two shapes to separate. At 5000 writes there are
  3943. // only four seals, so even the quadratic form copies well under a
  3944. // megabyte and the budget below could not tell them apart.
  3945. 'for _ in range(60000):',
  3946. ' os.write(3, b"x")',
  3947. ' os.sched_yield()',
  3948. 'os.write(3, b"\\n")',
  3949. // A real frame after the trickle proves framing still works on the
  3950. // coalesced residual.
  3951. 'print("after-trickle")',
  3952. 'return "done"',
  3953. ].join('\n')
  3954. let result: CodeRunResult
  3955. try {
  3956. const { runtime } = await setup({ maxWallMs: 30_000 })
  3957. result = await runtime.run({ program, bindings: [] })
  3958. } finally {
  3959. Buffer.concat = realConcat
  3960. }
  3961. expect(result.error).toBeUndefined()
  3962. expect(result.value).toBe('done')
  3963. expect(result.logs).toContain('after-trickle')
  3964. // Sealing appends a finished block rather than re-merging everything held, so
  3965. // each byte is copied once. Re-concatenating the whole buffer at every
  3966. // threshold made the cumulative copy volume quadratic — 10 MiB trickled a
  3967. // byte at a time copies 53.7 GB that way. A per-byte-copied budget is the
  3968. // discriminator, and it is measured rather than reasoned about: this shape
  3969. // copies about 119 KB for 60000 trickled bytes, the re-merging shape about
  3970. // 540 KB. 256 KiB sits between them with margin on both sides — most writes
  3971. // are coalesced by the pipe before they reach us, so the observed ratio is
  3972. // smaller than the asymptotic one, and the threshold has to sit where a real
  3973. // measurement lands rather than where the asymptote suggests.
  3974. expect(copied).toBeLessThan(256 * 1024)
  3975. }, 40_000)
  3976. it('seals trickled stray fragments into blocks without recopying the sealed prefix', async () => {
  3977. // The stray-capture buffer has the same object-overhead exposure as the fd-3
  3978. // reader above: each newline-free `data` chunk is its own Buffer, so a
  3979. // program pacing single-byte `os.write(1, ...)` accumulates one object per
  3980. // write, which the serialized-cost counter cannot see. Past MAX_PENDING_CHUNKS
  3981. // the fragments seal into a finished block; re-merging the whole residual at
  3982. // each threshold instead would copy the sealed prefix again and again, making
  3983. // the cumulative copy volume quadratic. `Buffer.concat` is wrapped to measure
  3984. // that volume — both shapes admit the same final log entry, so the copy total
  3985. // is the discriminator. maxLogBytes is raised so the trickle is retained,
  3986. // not truncated, which is what forces the fragments to accumulate and seal.
  3987. const realConcat = Buffer.concat.bind(Buffer)
  3988. let copied = 0
  3989. Buffer.concat = (list: readonly Uint8Array[], total?: number): Buffer<ArrayBuffer> => {
  3990. for (const part of list) copied += part.length
  3991. return realConcat(list, total)
  3992. }
  3993. let result: CodeRunResult
  3994. try {
  3995. const { runtime } = await setup({ maxLogBytes: 200_000, maxWallMs: 30_000 })
  3996. result = await runtime.run({
  3997. program: [
  3998. 'import os',
  3999. 'for _ in range(60000):',
  4000. ' os.write(1, b"x")',
  4001. ' os.sched_yield()',
  4002. 'os.write(1, b"\\n")',
  4003. 'return "done"',
  4004. ].join('\n'),
  4005. bindings: [],
  4006. })
  4007. } finally {
  4008. Buffer.concat = realConcat
  4009. }
  4010. expect(result.error).toBeUndefined()
  4011. expect(result.value).toBe('done')
  4012. // The trickle coalesces into one log line (no interior newlines). Its exact
  4013. // length depends on pipe coalescing, but it is one entry and non-empty.
  4014. expect(result.logs.length).toBe(1)
  4015. expect((result.logs[0] as string).length).toBeGreaterThan(0)
  4016. // Sealing appends a finished block rather than re-merging everything held, so
  4017. // each byte is copied a bounded number of times. Re-merging the whole
  4018. // residual at every seal threshold instead makes the cumulative copy volume
  4019. // quadratic. Measured like the fd-3 sibling above rather than reasoned about:
  4020. // this sealed shape copies about 120 KB for 60000 trickled bytes, the
  4021. // re-merging shape about 538 KB (the stray path adds one whole-residual
  4022. // concat at the terminating newline over the fd-3 sibling's 119/540, landing
  4023. // at the same order). 256 KiB sits between them with margin on both sides, so
  4024. // reverting the seal to a re-merge turns this assertion red.
  4025. expect(copied).toBeLessThan(256 * 1024)
  4026. }, 40_000)
  4027. it('caps a huge exception diagnostic child-side before it crosses the wire', async () => {
  4028. // A program can raise with a multi-megabyte message; the child must cap
  4029. // it at maxValueBytes before formatting/sending, not ship the whole
  4030. // payload for the host to truncate after parsing.
  4031. const { runtime } = await setup({ maxValueBytes: 1024 })
  4032. const result = await runtime.run({
  4033. program: 'raise ValueError("boom-" + "x" * (8 * 1024 * 1024))',
  4034. bindings: [],
  4035. })
  4036. expect(result.error?.kind).toBe('exception')
  4037. expect(result.error?.message).toContain('boom-')
  4038. expect(result.error?.message.endsWith('… [truncated]')).toBe(true)
  4039. expect(Buffer.byteLength(result.error?.message ?? '', 'utf8')).toBeLessThan(2048)
  4040. })
  4041. it('caps a control-heavy exception diagnostic by its serialized cost, not raw bytes', async () => {
  4042. // The diagnostic crosses fd 3 inside a JSON frame where a control character
  4043. // escapes sixfold (a NUL is one raw byte, six as `\u0000`). Capping by raw
  4044. // UTF-8 length would let a NUL-heavy message near maxValueBytes serialize to
  4045. // ~6x that and breach the frame ceiling — the silent worker-exit inversion
  4046. // the load-time cap check exists to prevent. The child meters the diagnostic
  4047. // by its serialized cost, so a NUL flood is truncated to fit the frame and
  4048. // the run still reports the exception rather than a worker-exit.
  4049. const { runtime } = await setup({ maxValueBytes: 4096 })
  4050. const result = await runtime.run({
  4051. // 512 KiB of NUL: ~3 MiB once escaped, far past the 4 KiB cap.
  4052. program: 'raise ValueError("\\x00" * (512 * 1024))',
  4053. bindings: [],
  4054. })
  4055. expect(result.error?.kind).toBe('exception')
  4056. expect(result.error?.message.endsWith('… [truncated]')).toBe(true)
  4057. // The SERIALIZED form (what the frame carried) fits the budget, so its raw
  4058. // length is well under it too — a raw-byte cap would have admitted ~4 KiB of
  4059. // NULs that serialize to ~24 KiB.
  4060. const serialized = JSON.stringify(result.error?.message ?? '')
  4061. expect(Buffer.byteLength(serialized, 'utf8')).toBeLessThanOrEqual(4096 + 8)
  4062. })
  4063. it('bounds a newline-free partial-line flood while the program is still running', async () => {
  4064. // print("x", end="") never completes a line, so nothing reaches the
  4065. // Python LogBuffer until settlement — the buffered tail must still hit
  4066. // the budget mid-run instead of growing without bound to RLIMIT/timeout.
  4067. const { runtime } = await setup({ maxLogBytes: 1024, maxWallMs: 15_000 })
  4068. const result = await runtime.run({
  4069. program: [
  4070. 'for _ in range(100000):',
  4071. ' print("xxxxxxxxxx", end="")',
  4072. 'return "done"',
  4073. ].join('\n'),
  4074. bindings: [],
  4075. })
  4076. expect(result.error).toBeUndefined()
  4077. expect(result.value).toBe('done')
  4078. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  4079. // The retained text is bounded by the budget, not the 1 MB the program wrote.
  4080. expect(result.logs.join('\n').length).toBeLessThan(4096)
  4081. }, 20_000)
  4082. it('discards empty writes instead of buffering one list slot each', async () => {
  4083. // An empty chunk adds no character, so the mid-run budget check (which
  4084. // compares buffered CHARS against the remaining ledger) can never fire on
  4085. // it. Buffering empty strings therefore grew `_pending` without bound —
  4086. // millions of slots per CPU second — until RLIMIT_AS turned an append into
  4087. // a MemoryError, long after the log ledger was exhausted. Two million
  4088. // empty writes must instead settle normally and contribute NO log entry,
  4089. // proving the chunk was dropped rather than joined at flush_line.
  4090. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 256, maxWallMs: 20_000 })
  4091. const result = await runtime.run({
  4092. program: [
  4093. 'import sys',
  4094. 'for _ in range(2000000):',
  4095. ' sys.stdout.write("")',
  4096. 'return "done"',
  4097. ].join('\n'),
  4098. bindings: [],
  4099. })
  4100. expect(result.error).toBeUndefined()
  4101. expect(result.value).toBe('done')
  4102. expect(result.logs).toEqual([])
  4103. }, 30_000)
  4104. it('stops scanning a single-write newline flood once the log ledger truncates', async () => {
  4105. // One write carrying half a million newlines: the offset scan must exit the
  4106. // instant LogBuffer truncates rather than re-slicing and pushing every
  4107. // remaining line. If it kept scanning it would exhaust the CPU/wall budget;
  4108. // the run instead settles quickly with exactly one truncation marker.
  4109. const { runtime } = await setup({ maxLogBytes: 256, maxWallMs: 10_000 })
  4110. const start = Date.now()
  4111. const result = await runtime.run({
  4112. program: ['print("x\\n" * 500000, end="")', 'return "done"'].join('\n'),
  4113. bindings: [],
  4114. })
  4115. expect(result.error).toBeUndefined()
  4116. expect(result.value).toBe('done')
  4117. expect(result.logs.filter(line => line.includes('log capture truncated'))).toHaveLength(1)
  4118. expect(Date.now() - start).toBeLessThan(8_000)
  4119. }, 15_000)
  4120. it('bounds an oversized newline-terminated write before joining and slicing it', async () => {
  4121. // The newline branch slices the first line out of the write before
  4122. // `LogBuffer.push` can apply its cheap budget rejection, so a single
  4123. // over-budget write cost a full extra copy of itself in peak address space —
  4124. // the amplification that bound exists to avoid, applied one layer too late.
  4125. // Measured under a 400 MiB addressSpaceMb with the slice unbounded: writes
  4126. // of 200 MiB and up died on MemoryError inside `sys.stdout.write`, reported
  4127. // as the PROGRAM's own exception rather than the promised truncation marker.
  4128. // `"\\n".rjust(n, "A")` is a single allocation ending in the newline, so the
  4129. // payload itself fits and the only remaining allocation is the stream's own
  4130. // slice; 340 MiB of a 400 MiB cap cannot survive one more copy of it.
  4131. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 400, maxWallMs: 30_000 })
  4132. const result = await runtime.run({
  4133. program: [
  4134. 'import sys',
  4135. 'payload = "\\n".rjust(340 * 1024 * 1024, "A")',
  4136. 'sys.stdout.write(payload)',
  4137. 'return "done"',
  4138. ].join('\n'),
  4139. bindings: [],
  4140. })
  4141. expect(result.error).toBeUndefined()
  4142. expect(result.value).toBe('done')
  4143. expect(result.logs).toEqual([logTruncationMarker(256)])
  4144. }, 40_000)
  4145. it('bounds a newline-free write against the already-buffered chunks before joining them', async () => {
  4146. // The newline-free arm buffers the write and then compared the buffered
  4147. // CHARACTER COUNT against the ledger — correct — but paid for the comparison
  4148. // with `"".join(self._pending)`, a second full copy of everything held. One
  4149. // buffered character is enough to make that join a copy of the whole
  4150. // following write. Measured under a 400 MiB addressSpaceMb with a 340 MiB
  4151. // second write: the join raised MemoryError inside `sys.stdout.write`, and
  4152. // because the oversized chunks stayed in `_pending` the settlement
  4153. // `flush_line` raised it again — that throw sits after the `except
  4154. // BaseException` block, so it costs the `done` frame and the run came back
  4155. // `timeout: wall-clock ceiling reached (30000ms)` with no logs at all. The
  4156. // bound must be applied BEFORE the join and the chunks dropped on that path,
  4157. // so the run settles with the truncation marker it promises.
  4158. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 400, maxWallMs: 30_000 })
  4159. const result = await runtime.run({
  4160. program: [
  4161. 'import sys',
  4162. // One unterminated character first, so `_pending` is non-empty and the
  4163. // large write cannot take the "buffered text IS the write" shortcut.
  4164. 'sys.stdout.write("x")',
  4165. 'payload = "A" * (340 * 1024 * 1024)',
  4166. 'sys.stdout.write(payload)',
  4167. 'return "done"',
  4168. ].join('\n'),
  4169. bindings: [],
  4170. })
  4171. expect(result.error).toBeUndefined()
  4172. expect(result.value).toBe('done')
  4173. expect(result.logs).toEqual([logTruncationMarker(256)])
  4174. }, 40_000)
  4175. it('bounds a newline-terminated write against the already-buffered chunks before joining them', async () => {
  4176. // Same allocation, reached through the newline arm: with chunks pending, the
  4177. // whole write used to be appended and joined so the offset scan could run
  4178. // over one string. Only the FIRST line needs those chunks, so a pending
  4179. // chunk plus a 340 MiB newline-terminated write under a 400 MiB
  4180. // addressSpaceMb died on MemoryError in the join before the per-line bound
  4181. // could reject anything, and the retained chunks made the settlement flush
  4182. // die the same way: measured, `timeout: wall-clock ceiling reached
  4183. // (30000ms)`. The reconstructed first line is now checked against the ledger
  4184. // and only a budget-sized prefix of it is copied; the rest of the write is
  4185. // scanned in place.
  4186. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 400, maxWallMs: 30_000 })
  4187. const result = await runtime.run({
  4188. program: [
  4189. 'import sys',
  4190. 'sys.stdout.write("x")',
  4191. 'payload = "\\n".rjust(340 * 1024 * 1024, "A")',
  4192. 'sys.stdout.write(payload)',
  4193. 'return "done"',
  4194. ].join('\n'),
  4195. bindings: [],
  4196. })
  4197. expect(result.error).toBeUndefined()
  4198. expect(result.value).toBe('done')
  4199. expect(result.logs).toEqual([logTruncationMarker(256)])
  4200. }, 40_000)
  4201. it('emits pending text on an explicit flush, before the run can be killed', async () => {
  4202. // `_LogStream` inherits TextIOBase's no-op `flush()`, so an explicit
  4203. // `print(..., flush=True)` or `sys.stdout.flush()` left the text in
  4204. // `_pending` with nothing to drain it but `flush_line` after settlement — a
  4205. // call a hanging or killed run never reaches. Measured: printing
  4206. // "before hang" with flush=True ahead of an infinite loop returned
  4207. // `logs: []`, losing the one diagnostic the program deliberately committed.
  4208. const { runtime } = await setup({ maxWallMs: 4_000 })
  4209. const result = await runtime.run({
  4210. program: [
  4211. 'import sys',
  4212. 'print("before hang", end="", flush=True)',
  4213. 'while True: pass',
  4214. ].join('\n'),
  4215. bindings: [],
  4216. })
  4217. expect(result.error?.kind).toBe('timeout')
  4218. expect(result.logs).toContain('before hang')
  4219. }, 15_000)
  4220. it('marks a dropped tail when the ledger lands on exactly zero remaining', async () => {
  4221. // One 100-character line costs 103 serialized bytes (quotes + separator),
  4222. // consuming a 104-byte budget minus the 1-byte array-envelope reservation
  4223. // (104 - 1 = 103) EXACTLY. Landing on zero never trips
  4224. // LogBuffer's "cost > remaining" branch, so `_truncated` stays unset and the
  4225. // stream's own `remaining > 0` guard silently discarded the unscanned tail —
  4226. // the run reported a complete log while dropping text. The tail must be
  4227. // pushed so the marker is emitted. (This surfaced only after empty writes
  4228. // stopped being buffered: `print` issues a trailing `write("")` whose
  4229. // buffered-empty path used to force the marker out incidentally.) A single
  4230. // wide line is used rather than many narrow ones so the CHILD ledger is the
  4231. // one that lands on zero: the host's identical ledger truncates first when
  4232. // many small entries precede the long marker text. `["y"*100]` serializes to
  4233. // exactly 104 bytes (103 payload + 1 envelope), so 104 is the smallest
  4234. // budget that admits the entry.
  4235. const { runtime } = await setup({ maxLogBytes: 104, maxWallMs: 10_000 })
  4236. const result = await runtime.run({
  4237. program: ['print("y" * 100 + "\\n" + "z" * 10, end="")', 'return "done"'].join('\n'),
  4238. bindings: [],
  4239. })
  4240. expect(result.error).toBeUndefined()
  4241. expect(result.value).toBe('done')
  4242. expect(result.logs).toContain('y'.repeat(100))
  4243. expect(result.logs.filter(line => line.includes('log capture truncated'))).toHaveLength(1)
  4244. // The dropped tail is not retained, but its loss is now reported.
  4245. expect(result.logs.some(line => line.includes('z'))).toBe(false)
  4246. }, 15_000)
  4247. it('keeps an admitted log within the serialized array envelope at the exact limit', async () => {
  4248. // Each entry is charged its JSON-string cost plus one separator byte, and
  4249. // the serialized outer logs array adds one more byte of envelope (two
  4250. // brackets and n-1 commas). The ledgers reserve that byte, so a result that
  4251. // exactly exhausts the ledger still serializes within the configured cap.
  4252. // At the 64-byte floor (the smallest admissible maxLogBytes): ledger 63,
  4253. // a 60-character line serializes as `"aaa...a"` (62 bytes) + 1 separator
  4254. // = 63, exactly exhausting the ledger and serializing as `["aaa...a"]`
  4255. // = 64 = the cap; a 61-character line costs 64 > 63 and truncates. The
  4256. // marker rides envelope, so the serialized logs run to cap + marker.
  4257. const { runtime } = await setup({ maxLogBytes: 64, maxWallMs: 10_000 })
  4258. const result = await runtime.run({
  4259. program: ['print("a" * 60 + "\\n" + "b" * 61, end="")', 'return "done"'].join('\n'),
  4260. bindings: [],
  4261. })
  4262. expect(result.error).toBeUndefined()
  4263. expect(result.value).toBe('done')
  4264. // The 60-character line was admitted; the 61-character line was not (a
  4265. // single 'b' would also match the marker's "bytes", so check for the line).
  4266. expect(result.logs).toContain('a'.repeat(60))
  4267. expect(result.logs.some(line => line.includes('b'.repeat(61)))).toBe(false)
  4268. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  4269. }, 15_000)
  4270. it('rejects a log budget too small to serialize the truncation marker', async () => {
  4271. // A maxLogBytes below 64 cannot serialize the truncation marker itself;
  4272. // it is rejected at construction so a marker-only truncated run cannot
  4273. // report more than the public cap. maxValueBytes keeps no floor beyond the
  4274. // positive-integer requirement (a completion can be 1 byte).
  4275. await expect(setup({ maxLogBytes: 63, maxWallMs: 10_000 })).rejects.toThrow(/must be at least 64/)
  4276. }, 15_000)
  4277. it('charges the JSON-escaped cost of control characters against the log ledger', async () => {
  4278. // A NUL renders as \u0000 (6 bytes) in the serialized outer logs; the
  4279. // ledger must charge that expansion, or a control-character flood admits
  4280. // 6x the configured cap.
  4281. const { runtime } = await setup({ maxLogBytes: 256 })
  4282. const result = await runtime.run({
  4283. program: [
  4284. 'for _ in range(500):',
  4285. ' print("\\x00" * 10)',
  4286. 'return "done"',
  4287. ].join('\n'),
  4288. bindings: [],
  4289. })
  4290. expect(result.error).toBeUndefined()
  4291. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  4292. // Serialized (escaped) size of retained entries stays in the budget's
  4293. // neighborhood: well under the ~30 kB an uncharged flood would retain.
  4294. const serialized = Buffer.byteLength(JSON.stringify(result.logs), 'utf8')
  4295. expect(serialized).toBeLessThan(1024)
  4296. })
  4297. it('charges the serialized cost child-side, so a control-heavy line truncates instead of being admitted whole', async () => {
  4298. // The child's ledger must charge what the entry costs on the wire, not its
  4299. // raw UTF-8 length: a NUL is one raw byte but six as its escape. A 24 MiB NUL
  4300. // line clears the cheap char-count lower bound (24 MiB < 32 MiB budget), so
  4301. // charging raw bytes would ADMIT it and emit a ~144 MiB escaped entry;
  4302. // charging the serialized cost (~144 MiB > the 32 MiB budget) rejects it
  4303. // before any encode and emits the marker instead. The address space (512 MiB,
  4304. // clearing the 12x load gate for a 32 MiB budget) is sized so the run loads;
  4305. // the gate separately guarantees a correctly-charged near-budget entry fits.
  4306. const { runtime } = await setup({ maxLogBytes: 32 * 1024 * 1024, addressSpaceMb: 512, maxWallMs: 20_000 })
  4307. const result = await runtime.run({
  4308. program: [
  4309. 'print("\\x00" * (24 * 1024 * 1024))',
  4310. 'return "done"',
  4311. ].join('\n'),
  4312. bindings: [],
  4313. })
  4314. expect(result.error).toBeUndefined()
  4315. expect(result.value).toBe('done')
  4316. expect(result.logs.filter(line => line.includes('log capture truncated'))).toHaveLength(1)
  4317. // Nothing of the line itself was retained: the ledger refused the whole entry.
  4318. expect(result.logs.every(line => !line.includes(String.fromCharCode(0)))).toBe(true)
  4319. }, 30_000)
  4320. it('bounds a huge unterminated tail after an early newline without copying it whole', async () => {
  4321. // The newline branch of _LogStream.write buffered the whole unterminated
  4322. // tail after the last newline into `_pending` before the flush trigger could
  4323. // bound it, so an early newline followed by a huge tail made a second full
  4324. // copy of the model's own string — a MemoryError the config gate cannot
  4325. // catch (the tail far exceeds maxLogBytes). The tail is now sliced to a
  4326. // budget-sized prefix, so the run truncates and completes. Linux-only RLIMIT_AS
  4327. // repro (Darwin skips the limit); on macOS this asserts the happy path.
  4328. //
  4329. // Sizing: the model builds `tail` (N) then the `"\n" + tail` write argument
  4330. // (another ~N), so construction peaks at ~2N — kept under the 384 MiB address
  4331. // space at N = 150 MiB (~300 MiB). The pre-fix code then buffered the whole
  4332. // ~150 MiB tail again, pushing past 384 MiB; the sliced prefix does not.
  4333. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 384, maxWallMs: 20_000 })
  4334. const result = await runtime.run({
  4335. program: [
  4336. 'import sys',
  4337. 'tail = "A" * (150 * 1024 * 1024)',
  4338. 'sys.stdout.write("\\n" + tail)',
  4339. 'return "done"',
  4340. ].join('\n'),
  4341. bindings: [],
  4342. })
  4343. expect(result.error).toBeUndefined()
  4344. expect(result.value).toBe('done')
  4345. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  4346. }, 30_000)
  4347. it('flushes logs before framing the value so their peaks do not add against RLIMIT_AS', async () => {
  4348. // The load gate bounds maxLogBytes and maxValueBytes INDEPENDENTLY against the
  4349. // address space, each at the 12x worst case. But the child framed the
  4350. // completion value (materializing its escaped form to meter it, then encoding
  4351. // the frame) while a newline-free log tail still sat unflushed in _pending.
  4352. // Those two peaks added: two budgets each admitted alone could together breach
  4353. // RLIMIT_AS, dying as worker-exit instead of settling. The flush now runs
  4354. // before the value is framed, so the log pending is freed first.
  4355. //
  4356. // Config: 32 MiB each against 512 MiB (each 32*12 = 384 MiB < 448 MiB
  4357. // budgetable, so both load). The program writes ~33M astral chars with no
  4358. // newline (buffered ~132 MB, under the char-count flush trigger) then returns
  4359. // ~33M astral chars — a ~132 MB serialized value that is itself OVER the 32 MiB
  4360. // maxValueBytes, so the correct outcome is `output-limit`. Pre-fix the
  4361. // unflushed 132 MB plus the value's build-and-encode (~396 MB) exceeded 512 MiB
  4362. // and OOM'd (reported as exception/worker-exit); flushing first lets the value
  4363. // check complete (~460 MB alone) and report output-limit. On Darwin (no
  4364. // RLIMIT_AS) the value is over budget too, so output-limit holds either way;
  4365. // the OOM the reorder prevents is the Linux-only failure.
  4366. const { runtime } = await setup({
  4367. maxLogBytes: 32 * 1024 * 1024,
  4368. maxValueBytes: 32 * 1024 * 1024,
  4369. addressSpaceMb: 512,
  4370. maxWallMs: 20_000,
  4371. })
  4372. const result = await runtime.run({
  4373. program: [
  4374. 'import sys',
  4375. 'sys.stdout.write("\\U0001F600" * 33_000_000)',
  4376. 'return "\\U0001F600" * 33_000_000',
  4377. ].join('\n'),
  4378. bindings: [],
  4379. })
  4380. expect(result.error?.kind).toBe('output-limit')
  4381. }, 30_000)
  4382. it('checks and encodes a wide completion value in O(depth), not O(width)', async () => {
  4383. // A wide flat list serializes to ~2 bytes per element but the pre-fix walk
  4384. // enqueued one traversal tuple per element (_check_done_value) and one stack
  4385. // entry plus a separator marker per element (_encode_json_plain) — ~56 bytes
  4386. // per element, ~28x the serialized size. A value the byte meter admits could
  4387. // therefore OOM on the checker's or encoder's own bookkeeping, the inversion
  4388. // the load gate exists to prevent (the gate reserves 12x, not 28x). Both now
  4389. // walk with an O(depth) cursor that pulls one child at a time, so the only
  4390. // width-proportional allocation is the output string the meter bounded.
  4391. //
  4392. // Config: maxValueBytes 20 MiB against 384 MiB (20*12 = 240 MiB < 320 MiB
  4393. // budgetable, so it loads). `[0] * 6_000_000` is ~12 MB of JSON, under the
  4394. // 20 MiB budget, so it must round-trip. Pre-fix the ~400 MB of per-element
  4395. // frames plus the interpreter exceeded 384 MiB and returned MemoryError as an
  4396. // exception. Linux-only RLIMIT_AS repro; on macOS the value round-trips
  4397. // either way, but the fixture stays within the address space so it is honest.
  4398. //
  4399. // `maxWallMs` is 60s, not the 20s the memory assertion alone needs: the O(depth)
  4400. // cursor pulls 6M elements one at a time through Python-level frames, which costs
  4401. // ~11s on an idle machine and more under the coverage lane's V8 instrumentation
  4402. // with several workers sharing a box. This budget bounds the run without letting a
  4403. // loaded runner's scheduling latency read as a `timeout` — what this test asserts
  4404. // is the O(depth) memory shape, not a speed claim.
  4405. const { runtime } = await setup({ maxValueBytes: 20 * 1024 * 1024, addressSpaceMb: 384, maxWallMs: 60_000 })
  4406. const result = await runtime.run({ program: 'return [0] * 6_000_000', bindings: [] })
  4407. expect(result.error).toBeUndefined()
  4408. expect(Array.isArray(result.value)).toBe(true)
  4409. expect((result.value as number[]).length).toBe(6_000_000)
  4410. }, 90_000)
  4411. it('validates wide binding arguments in O(depth), not O(width)', async () => {
  4412. // The completion-value walks are budgeted; this one is not. `dispatch` runs
  4413. // `_lossless_json_violation` on the arguments the MODEL built, and no
  4414. // child-side byte budget bounds them first: the frame ceiling is the host's
  4415. // and applies only after this validation returns. A per-member traversal
  4416. // frame therefore turned a legitimate call into the program's own
  4417. // MemoryError. Measured with tracemalloc on the two walk shapes over this
  4418. // exact argument (JSON ~17 MB): the cursor peaks at 0.0 MiB of auxiliary
  4419. // state, the pre-fix `stack.extend` at 459.1 MiB -- past the 384 MiB
  4420. // configured below, so the discriminating failure is real. It is Linux-only:
  4421. // Darwin skips RLIMIT_AS, so this case round-trips there either way.
  4422. //
  4423. // The binding echoes its argument's length back, so the assertion proves the
  4424. // call actually round-tripped rather than merely avoiding a crash.
  4425. const { runtime } = await setup({ addressSpaceMb: 384, maxWallMs: 60_000 })
  4426. const result = await runtime.run({
  4427. program: 'return await tools.width([0] * 6_000_000)',
  4428. bindings: [{
  4429. global: 'tools',
  4430. functions: { width: async (items: unknown) => (items as number[]).length },
  4431. }],
  4432. })
  4433. expect(result.error).toBeUndefined()
  4434. expect(result.value).toBe(6_000_000)
  4435. }, 90_000)
  4436. it('decodes a multi-megabyte binding reply without regex backtracking state', async () => {
  4437. // The child parses every host reply with `_decode_json_plain`. Its scalar
  4438. // regex matched strings with a `(?:[^"\\]|\\.)*` repetition, which makes
  4439. // CPython's backtracking engine retain state proportional to the string's
  4440. // WIDTH -- measured at ~146 MiB of engine state for a 1 MiB string and
  4441. // ~558 MiB for 4 MiB. A legitimate multi-megabyte reply therefore raised
  4442. // MemoryError inside `_pump_replies`; because that pump is the only settler
  4443. // of the call's future, the run stranded until the wall clock reported a
  4444. // `timeout` instead of returning the value the binding produced.
  4445. //
  4446. // Strings now scan chunk-to-chunk over a character class (no backtracking
  4447. // state). Measured on this exact 4 MiB reply: the pre-fix regex peaks at
  4448. // 557.8 MiB, past the default 512 MiB address space, while the scanner peaks
  4449. // at the 4.0 MiB result itself. Linux-only, like the other RLIMIT_AS repros:
  4450. // Darwin does not apply the limit, so the spike is merely allocated there.
  4451. const reply = 'A'.repeat(4 * 1024 * 1024)
  4452. const { runtime } = await setup({ maxWallMs: 60_000 })
  4453. const result = await runtime.run({
  4454. program: 'value = await tools.big({})\nreturn len(value)',
  4455. bindings: [{ global: 'tools', functions: { big: async () => reply } }],
  4456. })
  4457. expect(result.error).toBeUndefined()
  4458. expect(result.value).toBe(reply.length)
  4459. }, 90_000)
  4460. it('drops a late binding resolution before snapshotting it', async () => {
  4461. // `sendReply` checks `settled`, but only after the resolution has been walked
  4462. // and copied by `snapshotJsonValue`. Binding resolution carries no seam-level
  4463. // byte cap, so a binding that resolves a wide value AFTER the run already
  4464. // settled (here on `maxWallMs`) spent host heap building a frame that is then
  4465. // discarded. The check now runs before the snapshot.
  4466. //
  4467. // The binding resolves well after the 1s wall clock with a 2M-element array;
  4468. // the run must still report `timeout`, and the late value must not appear.
  4469. let resolvedLate = false
  4470. const { runtime } = await setup({ maxWallMs: 1_000 })
  4471. const result = await runtime.run({
  4472. program: 'return await tools.slow({})',
  4473. bindings: [{
  4474. global: 'tools',
  4475. functions: {
  4476. slow: async () => {
  4477. await new Promise(resolve => setTimeout(resolve, 2_500))
  4478. resolvedLate = true
  4479. return Array.from({ length: 2_000_000 }, () => 0)
  4480. },
  4481. },
  4482. }],
  4483. })
  4484. expect(result.error?.kind).toBe('timeout')
  4485. expect(result.value).toBeUndefined()
  4486. // Pin that the late path actually ran, so the assertion above is not vacuous.
  4487. await new Promise(resolve => setTimeout(resolve, 2_000))
  4488. expect(resolvedLate).toBe(true)
  4489. }, 90_000)
  4490. it('paces concurrent binding replies instead of queueing every frame at once', async () => {
  4491. // Binding resolution carries no seam-level byte cap. Before pacing, a program
  4492. // resolving several large values in one `asyncio.gather` round encoded them
  4493. // all in the same turn and queued every frame in fd 3's writable buffer,
  4494. // which exhausted the host heap and killed the whole process rather than
  4495. // failing the run. Replies are now encoded one at a time, waiting for
  4496. // `drain` when the pipe is full.
  4497. //
  4498. // Eight concurrent 4 MiB replies (32 MiB of frames) must all round-trip. The
  4499. // program sums the lengths, so the assertion proves every reply arrived and
  4500. // was matched to its own call -- pacing must not drop or misroute any. What
  4501. // this case cannot show is the peak itself, which lives in the stream's
  4502. // buffer: measured directly on a 64 KiB-highWaterMark pipe with this same
  4503. // 8x4 MiB shape, the unpaced writes buffered 32.0 MiB while the paced ones
  4504. // peaked at 0.0 MiB.
  4505. const chunk = 'A'.repeat(4 * 1024 * 1024)
  4506. const { runtime } = await setup({ maxWallMs: 60_000 })
  4507. const result = await runtime.run({
  4508. program: [
  4509. 'import asyncio',
  4510. 'parts = await asyncio.gather(*[tools.chunk({}) for _ in range(8)])',
  4511. 'return sum(len(p) for p in parts)',
  4512. ].join('\n'),
  4513. bindings: [{ global: 'tools', functions: { chunk: async () => chunk } }],
  4514. })
  4515. expect(result.error).toBeUndefined()
  4516. expect(result.value).toBe(8 * chunk.length)
  4517. }, 90_000)
  4518. it('bounds a flood of zero-byte log lines through the per-entry separator charge', async () => {
  4519. // Blank print() lines carry zero content bytes; without the +1 separator
  4520. // charge they would bypass maxLogBytes entirely and grow the retained
  4521. // array without bound. Each empty entry costs one byte, so a 64-byte
  4522. // budget retains at most 64 entries before the marker.
  4523. const { runtime } = await setup({ maxLogBytes: 64, maxWallMs: 10_000 })
  4524. const result = await runtime.run({
  4525. program: [
  4526. 'for _ in range(10000):',
  4527. ' print()',
  4528. 'return "done"',
  4529. ].join('\n'),
  4530. bindings: [],
  4531. })
  4532. expect(result.error).toBeUndefined()
  4533. expect(result.logs.length).toBeLessThanOrEqual(65)
  4534. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  4535. })
  4536. it('reassembles multibyte UTF-8 split across stray-output pipe chunks', async () => {
  4537. // A single os.write far past the 64 KiB pipe buffer forces multiple
  4538. // 'data' chunks; when the boundary lands inside a multibyte sequence,
  4539. // per-chunk decoding would corrupt it into replacement characters. Raw bytes
  4540. // are buffered and only decoded once a complete line (or the whole tail at
  4541. // flush) is assembled, so the split sequence is whole by the time it is
  4542. // decoded. The payload spans every valid multibyte lead class so
  4543. // accrueStrayCost's per-lead continuation ranges are all exercised: U+0900
  4544. // (E0 A4 80, the range-restricted E0 lead), U+4F60 and U+597D (E4/E5, plain
  4545. // 3-byte), U+1F600 (F0, the range-restricted F0 lead), and U+10FFFF (F4 8F
  4546. // BF BF, the range-restricted F4 lead).
  4547. const { runtime } = await setup({ maxLogBytes: 1024 * 1024 })
  4548. const result = await runtime.run({
  4549. program: [
  4550. 'import os',
  4551. // os.write is one syscall and returns a partial count on a full
  4552. // pipe, so loop until the whole payload (odd prefix -> a chunk
  4553. // boundary lands inside a multibyte sequence) is out.
  4554. String.raw`payload = b"a" * 65535 + "\u0900\u4f60\u597d\U0001f600\U0010ffff".encode("utf-8")`,
  4555. 'view = memoryview(payload)',
  4556. 'while view:',
  4557. ' view = view[os.write(1, view):]',
  4558. 'return "done"',
  4559. ].join('\n'),
  4560. bindings: [],
  4561. })
  4562. expect(result.error).toBeUndefined()
  4563. const text = result.logs.join('')
  4564. expect(text).toContain('\u0900\u4f60\u597d\u{1f600}\u{10ffff}')
  4565. expect(text).not.toContain('\ufffd')
  4566. })
  4567. it('flushes a stray-output byte sequence left incomplete when the pipe ends', async () => {
  4568. // The child writes the first two bytes of a 3-byte UTF-8 character to fd 1
  4569. // and exits, so the pipe closes with the sequence unfinished in the raw
  4570. // residual. The 'end' flush decodes the residual with `toString('utf8')`,
  4571. // which renders the stranded bytes as U+FFFD instead of dropping them.
  4572. const { runtime } = await setup({ maxLogBytes: 1024 * 1024 })
  4573. const result = await runtime.run({
  4574. program: [
  4575. 'import os',
  4576. // b"\xe4\xbd" is the leading two bytes of U+4F60; no continuation byte
  4577. // follows before exit.
  4578. String.raw`os.write(1, b"\xe4\xbd")`,
  4579. 'return "done"',
  4580. ].join('\n'),
  4581. bindings: [],
  4582. })
  4583. expect(result.error).toBeUndefined()
  4584. expect(result.logs.join('')).toContain('�')
  4585. })
  4586. it('rejects reserved words of EITHER backend language as binding globals', async () => {
  4587. // The seam's portable contract: `lambda` (Python keyword, legal JS name)
  4588. // and `typeof` (JS keyword, legal Python name) are both refused, so a
  4589. // namespace list valid on one backend is valid on every backend.
  4590. const { runtime } = await setup()
  4591. for (const global of ['lambda', 'typeof']) {
  4592. await expect(runtime.run({
  4593. program: 'return 1',
  4594. bindings: [{ global, functions: {} }],
  4595. })).rejects.toThrow(/is not a usable Python identifier/)
  4596. }
  4597. })
  4598. it('captures stray stdout bytes the child writes bypassing sys.stdout', async () => {
  4599. // Model code that writes to fd 1 via os.write() bypasses the Python-side
  4600. // LogBuffer, so the host's stray-byte capture on child.stdout is what
  4601. // records it.
  4602. const { runtime } = await setup()
  4603. const result = await runtime.run({
  4604. program: [
  4605. 'import os',
  4606. 'os.write(1, b"stray stdout\\n")',
  4607. 'os.write(2, b"stray stderr\\n")',
  4608. 'return "done"',
  4609. ].join('\n'),
  4610. bindings: [],
  4611. })
  4612. expect(result.error).toBeUndefined()
  4613. expect(result.logs.join('')).toContain('stray stdout')
  4614. expect(result.logs.join('')).toContain('stray stderr')
  4615. })
  4616. it('flushes bytes written through sys.__stdout__/sys.__stderr__ before the done frame', async () => {
  4617. // The bootstrap only replaces sys.stdout/sys.stderr with the _LogStream;
  4618. // sys.__stdout__/sys.__stderr__ are the original block-buffered wrappers
  4619. // over fd 1/2. A program that writes through them without an explicit flush
  4620. // would lose those bytes when the host SIGTERMs the child right after the
  4621. // done frame (the default SIGTERM disposition terminates without
  4622. // interpreter finalization). The settlement flush now drains the original
  4623. // std streams before sending the done frame, so the bytes land in the
  4624. // kernel pipe buffer and the host's stray capture records them.
  4625. const { runtime } = await setup()
  4626. const result = await runtime.run({
  4627. program: [
  4628. 'import sys',
  4629. // `-u` makes the streams write-through; re-enable block buffering so
  4630. // the bytes sit in the wrapper until the SETTLEMENT drain flushes them
  4631. // — the drain path, not the -u immediate write, is what this case pins.
  4632. 'if hasattr(sys.__stdout__, "reconfigure"):',
  4633. ' sys.__stdout__.reconfigure(write_through=False)',
  4634. ' sys.__stderr__.reconfigure(write_through=False)',
  4635. 'sys.__stdout__.write("orig stdout\\n")',
  4636. 'sys.__stderr__.write("orig stderr\\n")',
  4637. 'return "done"',
  4638. ].join('\n'),
  4639. bindings: [],
  4640. })
  4641. expect(result.error).toBeUndefined()
  4642. expect(result.value).toBe('done')
  4643. expect(result.logs.join('')).toContain('orig stdout')
  4644. expect(result.logs.join('')).toContain('orig stderr')
  4645. }, 15_000)
  4646. it('escalates to SIGKILL when the program traps SIGTERM and ignores the grace period', async () => {
  4647. // A program that traps SIGTERM should still die: the kill() escalation
  4648. // fires SIGKILL after graceMs. The full run reports either timeout (wall)
  4649. // or worker-exit depending on which finish reason wins the race.
  4650. const { runtime } = await setup({ maxWallMs: 400, graceMs: 200 })
  4651. const result = await runtime.run({
  4652. program: [
  4653. 'import signal, time',
  4654. 'signal.signal(signal.SIGTERM, lambda *a: None)',
  4655. 'while True: time.sleep(1)',
  4656. ].join('\n'),
  4657. bindings: [],
  4658. })
  4659. expect(['timeout', 'worker-exit']).toContain(result.error?.kind)
  4660. }, 6000)
  4661. it('bounds the fd-3 receive buffer against a newline-free flood', async () => {
  4662. // A program looping os.write(3, ...) with no newline would grow the host
  4663. // accumulator unbounded (the child's RLIMIT_AS does not cover the host
  4664. // string). The frame cap is a fixed 64 MiB memory-safety invariant —
  4665. // deliberately NOT derived from maxValueBytes, because legitimate binding
  4666. // call frames may be large. We flood slightly past it in 8 MiB writes so
  4667. // the test terminates promptly once the guard trips.
  4668. const ceiling = 64 * 1024 * 1024
  4669. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  4670. const start = Date.now()
  4671. const result = await runtime.run({
  4672. program: [
  4673. 'import os',
  4674. `for _ in range(${Math.ceil((ceiling * 1.1) / (8 * 1024 * 1024))}):`,
  4675. ' os.write(3, b"A" * (8 * 1024 * 1024))',
  4676. 'return "never"',
  4677. ].join('\n'),
  4678. bindings: [],
  4679. })
  4680. const elapsed = Date.now() - start
  4681. expect(result.error?.kind).toBe('worker-exit')
  4682. expect(result.error?.message).toContain(`protocol frame exceeded ${ceiling} bytes`)
  4683. // The breach ends the run before the wall ceiling (the run did not idle
  4684. // out); absolute pipe throughput varies too much under parallel suites
  4685. // for a tight bound.
  4686. expect(elapsed).toBeLessThan(30_000)
  4687. }, 45_000)
  4688. it('fails a forged oversized done value host-side as output-limit', async () => {
  4689. // The Python-side _done_with_value check is bypassable by writing a done
  4690. // frame straight to fd 3. The host re-enforces maxValueBytes; the seam
  4691. // forbids substituting a truncated value, so the run FAILS as output-limit
  4692. // instead of returning a lie.
  4693. const maxValueBytes = 64
  4694. const { runtime } = await setup({ maxValueBytes })
  4695. const result = await runtime.run({
  4696. program: [
  4697. 'import os, json',
  4698. 'big = "B" * 5000',
  4699. 'os.write(3, json.dumps({"type":"done","value":big}).encode() + b"\\n")',
  4700. // The real done never sends; the forged one settles the run.
  4701. 'import time',
  4702. 'time.sleep(5)',
  4703. ].join('\n'),
  4704. bindings: [],
  4705. })
  4706. expect(result.value).toBeUndefined()
  4707. expect(result.error?.kind).toBe('output-limit')
  4708. expect(result.error?.message).toContain('exceeded 64 bytes')
  4709. }, 8000)
  4710. it('drops a forged oversized log frame on its code-unit lower bound, before escaping it', async () => {
  4711. // A forged `log` frame carrying a control-heavy string: NULs escape
  4712. // several-fold (one NUL -> six bytes `\u0000`). The raw frame stays under
  4713. // the host's 64 MiB parse cap (4 MiB of `\u0000` text = 24 MiB raw) while
  4714. // the escaped form would be ~24 MiB. Charging it required building that
  4715. // escaped copy first, so a 32-byte maxLogBytes could still force a large
  4716. // host allocation. The cheap `length + 3` lower bound truncates it instead.
  4717. // The host's own heap is what is under test, so keep the child's address
  4718. // space generous enough to BUILD the frame.
  4719. const { runtime } = await setup({ maxLogBytes: 128, addressSpaceMb: 1024, maxWallMs: 60_000 })
  4720. const before = process.memoryUsage().heapUsed
  4721. const result = await runtime.run({
  4722. program: [
  4723. 'import os',
  4724. // Written as a raw frame so the child's own ledger never sees it.
  4725. 'os.write(3, b\'{"type":"log","text":"\' + b"\\\\u0000" * (4 * 1024 * 1024) + b\'"}\\n\')',
  4726. 'return "settled"',
  4727. ].join('\n'),
  4728. bindings: [],
  4729. })
  4730. expect(result.error).toBeUndefined()
  4731. expect(result.value).toBe('settled')
  4732. // The frame was dropped as one truncation marker, not retained.
  4733. expect(result.logs).toEqual([logTruncationMarker(128)])
  4734. // The escaped copy (~144 MiB) was never materialized.
  4735. expect(process.memoryUsage().heapUsed - before).toBeLessThan(256 * 1024 * 1024)
  4736. }, 90_000)
  4737. it('charges a forged log frame its escaped cost once past the code-unit lower bound', async () => {
  4738. // The cheap lower bound only rejects what cannot possibly fit; a SHORT
  4739. // control-heavy frame clears it and must still be charged what it costs on
  4740. // the wire. Eleven NULs are 14 against the 64-byte ledger's cheap bound
  4741. // (six bytes each, two quotes, one separator), so the full charge truncates.
  4742. const { runtime } = await setup({ maxLogBytes: 64 })
  4743. const result = await runtime.run({
  4744. program: [
  4745. 'import os',
  4746. 'os.write(3, b\'{"type":"log","text":"\' + b"\\\\u0000" * 11 + b\'"}\\n\')',
  4747. 'return "settled"',
  4748. ].join('\n'),
  4749. bindings: [],
  4750. })
  4751. expect(result.error).toBeUndefined()
  4752. expect(result.value).toBe('settled')
  4753. expect(result.logs).toEqual([logTruncationMarker(64)])
  4754. }, 8000)
  4755. it('caps a forged done error.message from its code-unit prefix, never encoding the whole message', async () => {
  4756. // `Buffer.from(message)` on a message near the frame ceiling allocates a
  4757. // full UTF-8 copy before maxValueBytes applies. Only the first
  4758. // maxValueBytes code units can fit the cap, so only that prefix is encoded
  4759. // — at most 3x the cap in bytes. The message here is 48 MiB of ASCII: its
  4760. // full encode would be another 48 MiB in the host.
  4761. const maxValueBytes = 64
  4762. const { runtime } = await setup({ maxValueBytes, addressSpaceMb: 1024, maxWallMs: 60_000 })
  4763. const before = process.memoryUsage().heapUsed
  4764. const result = await runtime.run({
  4765. program: [
  4766. 'import os',
  4767. 'os.write(3, b\'{"type":"done","error":{"kind":"exception","message":"\' + b"E" * (48 * 1024 * 1024) + b\'"}}\\n\')',
  4768. 'import time',
  4769. 'time.sleep(30)',
  4770. ].join('\n'),
  4771. bindings: [],
  4772. })
  4773. expect(result.error?.kind).toBe('exception')
  4774. const message = result.error?.message ?? ''
  4775. // The marker's 15 bytes come OUT of the 64-byte cap, so 49 E's precede it
  4776. // and the whole string is exactly 64 bytes — not 64 plus the marker.
  4777. expect(message).toBe(`${'E'.repeat(maxValueBytes - 15)}… [truncated]`)
  4778. expect(Buffer.byteLength(message, 'utf8')).toBe(maxValueBytes)
  4779. // JSON.parse already holds the 48 MiB string; the cap must not add a
  4780. // second full-length copy on top of it.
  4781. expect(process.memoryUsage().heapUsed - before).toBeLessThan(256 * 1024 * 1024)
  4782. }, 90_000)
  4783. it('keeps a capped diagnostic within maxValueBytes, marker included', async () => {
  4784. // The marker is part of the emitted diagnostic, so its bytes are reserved
  4785. // from the cap rather than appended past it — the host meters this same
  4786. // field downstream. Checked on BOTH producers: the child's own _cap_message
  4787. // (a raised exception) and the host's capMessage (a forged done frame).
  4788. const maxValueBytes = 40
  4789. const { runtime } = await setup({ maxValueBytes })
  4790. const raised = await runtime.run({
  4791. program: 'raise ValueError("R" * 100000)',
  4792. bindings: [],
  4793. })
  4794. expect(raised.error?.kind).toBe('exception')
  4795. const raisedMessage = raised.error?.message ?? ''
  4796. expect(raisedMessage.endsWith('… [truncated]')).toBe(true)
  4797. expect(Buffer.byteLength(raisedMessage, 'utf8')).toBeLessThanOrEqual(maxValueBytes)
  4798. const forged = await runtime.run({
  4799. program: [
  4800. 'import os, json',
  4801. 'msg = "F" * 100000',
  4802. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":msg}}).encode() + b"\\n")',
  4803. 'import time',
  4804. 'time.sleep(5)',
  4805. ].join('\n'),
  4806. bindings: [],
  4807. })
  4808. expect(forged.error?.kind).toBe('exception')
  4809. const forgedMessage = forged.error?.message ?? ''
  4810. expect(forgedMessage.endsWith('… [truncated]')).toBe(true)
  4811. expect(Buffer.byteLength(forgedMessage, 'utf8')).toBe(maxValueBytes)
  4812. }, 15_000)
  4813. it('emits the marker alone when the cap is smaller than the marker itself', async () => {
  4814. // With maxValueBytes below the marker's own 15 bytes there is no room for
  4815. // message text; the marker still goes out, so the truncation stays reported
  4816. // instead of the diagnostic silently becoming empty. Both producers agree.
  4817. const { runtime } = await setup({ maxValueBytes: 4 })
  4818. const raised = await runtime.run({ program: 'raise ValueError("R" * 500)', bindings: [] })
  4819. expect(raised.error?.kind).toBe('exception')
  4820. expect(raised.error?.message).toBe('… [truncated]')
  4821. const forged = await runtime.run({
  4822. program: [
  4823. 'import os, json',
  4824. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":"F" * 500}}).encode() + b"\\n")',
  4825. 'import time',
  4826. 'time.sleep(5)',
  4827. ].join('\n'),
  4828. bindings: [],
  4829. })
  4830. expect(forged.error?.kind).toBe('exception')
  4831. expect(forged.error?.message).toBe('… [truncated]')
  4832. }, 15_000)
  4833. it('caps a forged done error.message without splitting a surrogate pair', async () => {
  4834. // At exactly maxValueBytes code units the prefix can end on a high
  4835. // surrogate whose low half sits just outside it. `Buffer.from` encodes that
  4836. // orphan as U+FFFD — the same corruption a mid-sequence byte cut causes —
  4837. // and those three replacement bytes sit past the marker-reserved budget, so
  4838. // the byte trim-back drops them.
  4839. const maxValueBytes = 32
  4840. const { runtime } = await setup({ maxValueBytes })
  4841. const result = await runtime.run({
  4842. program: [
  4843. 'import os, json',
  4844. // 32 ASCII chars then astral characters: code unit 32 is the first
  4845. // character's high surrogate (Python spells it as one code point, so
  4846. // json.dumps emits the raw 4 bytes the host reads back as a pair).
  4847. 'msg = "A" * 32 + "\\U0001f600" * 4',
  4848. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":msg}}).encode() + b"\\n")',
  4849. 'import time',
  4850. 'time.sleep(5)',
  4851. ].join('\n'),
  4852. bindings: [],
  4853. })
  4854. expect(result.error?.kind).toBe('exception')
  4855. // 17 A's fill the marker-reserved budget; no orphaned half, no U+FFFD.
  4856. expect(result.error?.message).toBe(`${'A'.repeat(17)}… [truncated]`)
  4857. expect(Buffer.byteLength(result.error?.message ?? '', 'utf8')).toBe(maxValueBytes)
  4858. }, 8000)
  4859. it('returns a diagnostic under a third of the cap untouched, skipping the encode', async () => {
  4860. // Under maxValueBytes/3 code units a message cannot overflow the cap
  4861. // whatever it holds (3 bytes is the per-code-unit maximum), so the fast
  4862. // path returns it without encoding anything. Non-ASCII proves the bound is
  4863. // the code-unit count, not a byte assumption: 6 characters at 3 bytes each
  4864. // is 18 bytes, inside the 64-byte cap.
  4865. const { runtime } = await setup({ maxValueBytes: 64 })
  4866. const result = await runtime.run({
  4867. program: [
  4868. 'import os, json',
  4869. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":"中文中文中文"}}).encode() + b"\\n")',
  4870. 'import time',
  4871. 'time.sleep(5)',
  4872. ].join('\n'),
  4873. bindings: [],
  4874. })
  4875. expect(result.error?.kind).toBe('exception')
  4876. expect(result.error?.message).toBe('中文中文中文')
  4877. }, 8000)
  4878. it('re-caps a forged done error.message host-side on a UTF-8 boundary', async () => {
  4879. // A forged done frame can carry an arbitrarily long error message; the
  4880. // host caps it to maxValueBytes and appends the shared marker. The
  4881. // message is emoji-dense and the cap is chosen so the marker-reserved
  4882. // 51-byte cut lands INSIDE a 4-byte sequence (one ASCII byte then 4-byte
  4883. // runs, so only a cut at 1 + 4k is aligned) — the cap must trim back to a
  4884. // code-point boundary rather than decode a replacement character, which
  4885. // would also exceed the cap.
  4886. const maxValueBytes = 66
  4887. const { runtime } = await setup({ maxValueBytes })
  4888. const result = await runtime.run({
  4889. program: [
  4890. 'import os, json',
  4891. 'msg = "E" + "\\U0001f600" * 2000',
  4892. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":msg}}).encode() + b"\\n")',
  4893. 'import time',
  4894. 'time.sleep(5)',
  4895. ].join('\n'),
  4896. bindings: [],
  4897. })
  4898. expect(result.error?.kind).toBe('exception')
  4899. const message = result.error?.message ?? ''
  4900. expect(message.endsWith('… [truncated]')).toBe(true)
  4901. const marker = '… [truncated]'
  4902. const body = message.slice(0, message.length - marker.length)
  4903. // The WHOLE message, marker included, honors the cap.
  4904. expect(Buffer.byteLength(message, 'utf8')).toBeLessThanOrEqual(maxValueBytes)
  4905. // 'E' plus 12 emoji is 49 bytes: the trim-back walked the 51-byte budget
  4906. // down past two continuation bytes rather than splitting the 13th.
  4907. expect(body).toBe(`E${'\u{1f600}'.repeat(12)}`)
  4908. // The cut landed on a code-point boundary — no replacement character.
  4909. expect(body).not.toContain('\ufffd')
  4910. }, 8000)
  4911. it('bounds a single oversized newline-terminated line on fd 3', async () => {
  4912. // The same cap applies to one giant framed line. Write EXACTLY the
  4913. // cap with no newline — at the limit, not past it, so nothing trips —
  4914. // then a small newline tail, which is the chunk that crosses.
  4915. const ceiling = 64 * 1024 * 1024
  4916. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  4917. const result = await runtime.run({
  4918. program: [
  4919. 'import os',
  4920. 'chunk = b"A" * (8 * 1024 * 1024)',
  4921. `for _ in range(${ceiling / (8 * 1024 * 1024)}):`,
  4922. ' os.write(3, chunk)',
  4923. 'os.write(3, b"AAAA\\n")',
  4924. 'return "never"',
  4925. ].join('\n'),
  4926. bindings: [],
  4927. })
  4928. expect(result.error?.kind).toBe('worker-exit')
  4929. expect(result.error?.message).toContain(`protocol frame exceeded ${ceiling} bytes`)
  4930. }, 90_000)
  4931. it('rejects an over-cap newline-free fd-3 buffer without first joining it into one line', async () => {
  4932. // The cap has to be enforced on the byte COUNTER before Buffer.concat,
  4933. // not on the joined line afterwards: the join is a second copy of
  4934. // everything held, so a program could force roughly twice the advertised
  4935. // 64 MiB of host memory before anything rejected it.
  4936. //
  4937. // This program writes past the cap with no newline: the counter crosses on
  4938. // the 9th 8 MiB write (72 MiB) while the buffer is still a single unframed
  4939. // line, so the pre-join check rejects it without concat-ing a second copy.
  4940. // Checking the joined line instead would have produced a 72 MiB FIRST LINE
  4941. // that the per-line bound then dropped only after the doubling had happened.
  4942. const ceiling = 64 * 1024 * 1024
  4943. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  4944. const result = await runtime.run({
  4945. program: [
  4946. 'import os',
  4947. 'chunk = b"A" * (8 * 1024 * 1024)',
  4948. `for _ in range(${ceiling / (8 * 1024 * 1024) + 1}):`,
  4949. ' os.write(3, chunk)',
  4950. 'return "never"',
  4951. ].join('\n'),
  4952. bindings: [],
  4953. })
  4954. expect(result.value).toBeUndefined()
  4955. expect(result.error?.kind).toBe('worker-exit')
  4956. expect(result.error?.message).toContain(`protocol frame exceeded ${ceiling} bytes`)
  4957. }, 120_000)
  4958. it('keeps two within-cap frames whose combined buffer crosses the cap', async () => {
  4959. // The unframed byte counter charges the WHOLE buffer, which legitimately
  4960. // holds several frames each within FRAME_PARSE_CAP_BYTES. A first frame of
  4961. // exactly the cap followed by a second frame crosses the counter without
  4962. // either frame exceeding the cap; the first-frame check (not the counter)
  4963. // must let them through, or a legitimate near-cap frame plus a trailing
  4964. // frame would be misreported as a worker-exit.
  4965. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  4966. const result = await runtime.run({
  4967. program: [
  4968. 'import os',
  4969. 'chunk = b"A" * (8 * 1024 * 1024)',
  4970. // Exactly the cap, no newline — at the limit, so nothing trips.
  4971. 'for _ in range(8):',
  4972. ' os.write(3, chunk)',
  4973. // A newline, then a small legitimate log frame.
  4974. 'os.write(3, b"\\n{\\"type\\":\\"log\\",\\"text\\":\\"after-cap-frames\\"}\\n")',
  4975. 'return "done"',
  4976. ].join('\n'),
  4977. bindings: [],
  4978. })
  4979. expect(result.error).toBeUndefined()
  4980. expect(result.value).toBe('done')
  4981. expect(result.logs).toContain('after-cap-frames')
  4982. }, 120_000)
  4983. it('rejects an oversized first frame that lands on the sealing threshold with a newline', async () => {
  4984. // The fragment-count seal runs only on newline-free chunks (the ELSE half
  4985. // of the newline branch), so a chunk that carries the first newline always
  4986. // reaches the join and its first-frame check; sealing it into a block
  4987. // would empty pendingChunks, leave sawNewline false, and skip that check.
  4988. // Whether the pipe delivers exactly 1024 chunks is timing-dependent, but
  4989. // the oversized first frame (63.9 MiB of A's + 12289 more before the
  4990. // newline) exceeds FRAME_PARSE_CAP_BYTES no matter how it arrives — the
  4991. // case pins the worker-exit settlement, not a pre/post copy-count
  4992. // distinction (both orders reject an over-cap frame).
  4993. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  4994. const result = await runtime.run({
  4995. program: [
  4996. 'import os',
  4997. // 4 KiB writes are <= PIPE_BUF, so each os.write is atomic and the
  4998. // host sees one chunk per write; 16384 of them accumulate 64 MiB of
  4999. // newline-free bytes (16 fragment-count seals of 1024 chunks).
  5000. 'chunk = b"A" * 4096',
  5001. 'for _ in range(16384):',
  5002. ' os.write(3, chunk)',
  5003. // 12289 more A's push the first frame past 64 MiB; drain-loop so the
  5004. // write cannot truncate, then a newline and a small legitimate frame.
  5005. "data = b'A' * 12289 + b'\\n' + b'{\"type\":\"log\",\"text\":\"after-seal\"}\\n'",
  5006. 'view = memoryview(data)',
  5007. 'while view:',
  5008. ' view = view[os.write(3, view):]',
  5009. 'return "done"',
  5010. ].join('\n'),
  5011. bindings: [],
  5012. })
  5013. expect(result.error?.kind).toBe('worker-exit')
  5014. expect(result.error?.message).toContain('protocol frame exceeded')
  5015. }, 120_000)
  5016. })