runtime.spec.ts 170 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072207320742075207620772078207920802081208220832084208520862087208820892090209120922093209420952096209720982099210021012102210321042105210621072108210921102111211221132114211521162117211821192120212121222123212421252126212721282129213021312132213321342135213621372138213921402141214221432144214521462147214821492150215121522153215421552156215721582159216021612162216321642165216621672168216921702171217221732174217521762177217821792180218121822183218421852186218721882189219021912192219321942195219621972198219922002201220222032204220522062207220822092210221122122213221422152216221722182219222022212222222322242225222622272228222922302231223222332234223522362237223822392240224122422243224422452246224722482249225022512252225322542255225622572258225922602261226222632264226522662267226822692270227122722273227422752276227722782279228022812282228322842285228622872288228922902291229222932294229522962297229822992300230123022303230423052306230723082309231023112312231323142315231623172318231923202321232223232324232523262327232823292330233123322333233423352336233723382339234023412342234323442345234623472348234923502351235223532354235523562357235823592360236123622363236423652366236723682369237023712372237323742375237623772378237923802381238223832384238523862387238823892390239123922393239423952396239723982399240024012402240324042405240624072408240924102411241224132414241524162417241824192420242124222423242424252426242724282429243024312432243324342435243624372438243924402441244224432444244524462447244824492450245124522453245424552456245724582459246024612462246324642465246624672468246924702471247224732474247524762477247824792480248124822483248424852486248724882489249024912492249324942495249624972498249925002501250225032504250525062507250825092510251125122513251425152516251725182519252025212522252325242525252625272528252925302531253225332534253525362537253825392540254125422543254425452546254725482549255025512552255325542555255625572558255925602561256225632564256525662567256825692570257125722573257425752576257725782579258025812582258325842585258625872588258925902591259225932594259525962597259825992600260126022603260426052606260726082609261026112612261326142615261626172618261926202621262226232624262526262627262826292630263126322633263426352636263726382639264026412642264326442645264626472648264926502651265226532654265526562657265826592660266126622663266426652666266726682669267026712672267326742675267626772678267926802681268226832684268526862687268826892690269126922693269426952696269726982699270027012702270327042705270627072708270927102711271227132714271527162717271827192720272127222723272427252726272727282729273027312732273327342735273627372738273927402741274227432744274527462747274827492750275127522753275427552756275727582759276027612762276327642765276627672768276927702771277227732774277527762777277827792780278127822783278427852786278727882789279027912792279327942795279627972798279928002801280228032804280528062807280828092810281128122813281428152816281728182819282028212822282328242825282628272828282928302831283228332834283528362837283828392840284128422843284428452846284728482849285028512852285328542855285628572858285928602861286228632864286528662867286828692870287128722873287428752876287728782879288028812882288328842885288628872888288928902891289228932894289528962897289828992900290129022903290429052906290729082909291029112912291329142915291629172918291929202921292229232924292529262927292829292930293129322933293429352936293729382939294029412942294329442945294629472948294929502951295229532954295529562957295829592960296129622963296429652966296729682969297029712972297329742975297629772978297929802981298229832984298529862987298829892990299129922993299429952996299729982999300030013002300330043005300630073008300930103011301230133014301530163017301830193020302130223023302430253026302730283029303030313032303330343035303630373038303930403041304230433044304530463047304830493050305130523053305430553056305730583059306030613062306330643065306630673068306930703071307230733074307530763077307830793080308130823083308430853086308730883089309030913092309330943095309630973098309931003101310231033104310531063107310831093110311131123113311431153116311731183119312031213122312331243125312631273128312931303131313231333134313531363137313831393140314131423143314431453146314731483149315031513152315331543155315631573158315931603161316231633164316531663167316831693170317131723173317431753176317731783179318031813182318331843185318631873188318931903191319231933194319531963197319831993200320132023203320432053206320732083209321032113212321332143215321632173218321932203221322232233224322532263227322832293230323132323233323432353236323732383239324032413242324332443245324632473248324932503251325232533254325532563257325832593260326132623263326432653266326732683269327032713272327332743275327632773278327932803281328232833284328532863287328832893290329132923293329432953296329732983299330033013302330333043305330633073308330933103311331233133314331533163317331833193320332133223323332433253326332733283329333033313332333333343335333633373338333933403341334233433344334533463347334833493350335133523353335433553356335733583359336033613362336333643365336633673368336933703371337233733374337533763377337833793380338133823383338433853386338733883389339033913392339333943395339633973398339934003401340234033404340534063407340834093410341134123413341434153416341734183419342034213422342334243425342634273428342934303431343234333434343534363437343834393440344134423443344434453446344734483449345034513452345334543455345634573458345934603461346234633464346534663467346834693470347134723473347434753476347734783479348034813482348334843485348634873488348934903491349234933494349534963497349834993500350135023503350435053506350735083509351035113512351335143515351635173518351935203521352235233524352535263527352835293530353135323533353435353536353735383539354035413542354335443545354635473548354935503551355235533554355535563557355835593560356135623563356435653566356735683569357035713572357335743575357635773578357935803581358235833584358535863587358835893590359135923593359435953596359735983599360036013602
  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 'cordis'
  7. import { PythonCodeRuntime } 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. // 256 MiB framing ceiling 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 `ceiling - envelope`, not `(ceiling - envelope) / 6` (that
  103. // divided in escape expansion the charge already counts).
  104. const admissible = 256 * 1024 * 1024 - 64
  105. const ctx = new Context()
  106. await expect(ctx.plugin(PythonCodeRuntime, { maxLogBytes: admissible + 1 }))
  107. .rejects.toThrow(/maxLogBytes must not exceed 268435392 .*fd-3 frame ceiling/)
  108. await expect(ctx.plugin(PythonCodeRuntime, { maxValueBytes: admissible + 1 }))
  109. .rejects.toThrow(/maxValueBytes must not exceed 268435392 .*fd-3 frame ceiling/)
  110. // The boundary value itself loads: the bound is the largest cap a frame can
  111. // still carry, not one below it.
  112. const boundary = await ctx.plugin(PythonCodeRuntime, { maxValueBytes: admissible })
  113. await boundary.dispose()
  114. })
  115. it('rejects a pythonBin that spawn() would throw on, at load', async () => {
  116. // Both values pass the string schema and both make `spawn` throw
  117. // SYNCHRONOUSLY from inside run() — ERR_INVALID_ARG_VALUE for the empty
  118. // path, ERR_INVALID_ARG_TYPE for the NUL — so run() would REJECT instead of
  119. // resolving the worker-exit the seam promises for a child that cannot
  120. // start. Both are self-contained configuration errors, so they fail here.
  121. const ctx = new Context()
  122. await expect(ctx.plugin(PythonCodeRuntime, { pythonBin: '' }))
  123. .rejects.toThrow(/pythonBin must be a non-empty path without NUL bytes/)
  124. await expect(ctx.plugin(PythonCodeRuntime, { pythonBin: 'py\u0000thon3' }))
  125. .rejects.toThrow(/pythonBin must be a non-empty path without NUL bytes/)
  126. })
  127. it('rejects a timer budget setTimeout would silently clamp to 1 ms', async () => {
  128. // Node stores a setTimeout delay as a signed 32-bit value and substitutes
  129. // 1 ms for anything larger, inverting the knob's meaning: a huge maxWallMs
  130. // would time every run out at once, and a huge graceMs would SIGKILL one
  131. // millisecond after SIGTERM. Both must fail at load instead.
  132. const ctx = new Context()
  133. await expect(ctx.plugin(PythonCodeRuntime, { maxWallMs: 2_147_483_648 }))
  134. .rejects.toThrow(/maxWallMs must not exceed 2147483647/)
  135. // graceMs is bounded by the close deadline's added margin, not by the raw
  136. // timer maximum, because that sum is what gets armed.
  137. await expect(ctx.plugin(PythonCodeRuntime, { graceMs: 2_147_481_648 }))
  138. .rejects.toThrow(/graceMs must not exceed 2147481647/)
  139. // The exact maxima still load.
  140. await expect(ctx.plugin(PythonCodeRuntime, { maxWallMs: 2_147_483_647, graceMs: 2_147_481_647 }))
  141. .resolves.toBeDefined()
  142. })
  143. it('rejects loading this Unix-only backend on Windows', async () => {
  144. // The bootstrap needs the POSIX `resource` module, a positional fd 3, and
  145. // negative-PID process-group signals — none on Windows. The constructor
  146. // must throw at load rather than register ctx.codeRuntime and defer the
  147. // failure to the first run.
  148. const original = process.platform
  149. Object.defineProperty(process, 'platform', { value: 'win32', configurable: true })
  150. try {
  151. const ctx = new Context()
  152. await expect(ctx.plugin(PythonCodeRuntime, {})).rejects.toThrow(/requires a Unix platform/)
  153. } finally {
  154. Object.defineProperty(process, 'platform', { value: original, configurable: true })
  155. }
  156. })
  157. it('rejects a binding global that is not a Python identifier or is reserved', async () => {
  158. const { runtime } = await setup()
  159. await expect(runtime.run({
  160. program: 'return 1',
  161. bindings: [{ global: '1bad', functions: {} }],
  162. })).rejects.toThrow(/is not a usable Python identifier/)
  163. await expect(runtime.run({
  164. program: 'return 1',
  165. bindings: [{ global: 'class', functions: {} }],
  166. })).rejects.toThrow(/is not a usable Python identifier/)
  167. })
  168. it('rejects duplicate binding namespaces', async () => {
  169. const { runtime } = await setup()
  170. await expect(runtime.run({
  171. program: 'return 1',
  172. bindings: [
  173. { global: 'tools', functions: {} },
  174. { global: 'tools', functions: {} },
  175. ],
  176. })).rejects.toThrow(/duplicate binding global/)
  177. })
  178. it('rejects run() after disposal, and unregisters ctx.codeRuntime', async () => {
  179. const { ctx, fiber, runtime } = await setup()
  180. await fiber.dispose()
  181. await expect(runtime.run({ program: 'return 1', bindings: [] }))
  182. .rejects.toThrow(/after disposal/)
  183. expect(ctx.get('codeRuntime')).toBeUndefined()
  184. })
  185. it('short-circuits when the request signal is already aborted', async () => {
  186. const { runtime } = await setup()
  187. const signal = AbortSignal.abort('already-cancelled')
  188. const result = await runtime.run({ program: 'return 1', bindings: [], signal })
  189. expect(result.error?.kind).toBe('abort')
  190. expect(result.error?.message).toContain('already-cancelled')
  191. expect(result.logs).toEqual([])
  192. })
  193. it('short-circuits on an already-aborted signal whose reason cannot be converted', async () => {
  194. // The pre-flight arm converted the reason with a bare `String()`, so a
  195. // hostile reason threw out of `run()` — the seam promises to reject only for
  196. // misuse, and a caller's cancellation token is not misuse.
  197. const { runtime } = await setup()
  198. const signal = AbortSignal.abort({
  199. [Symbol.toPrimitive]() { throw new Error('reason blew up') },
  200. })
  201. const result = await runtime.run({ program: 'return 1', bindings: [], signal })
  202. expect(result.error?.kind).toBe('abort')
  203. expect(result.error?.message).toBe('<unrenderable rejection value>')
  204. expect(result.logs).toEqual([])
  205. })
  206. it('runs the interpreter from materialized scripts outside the package, and removes them per run', async () => {
  207. // The interpreter is an EXTERNAL process, so it can only open paths the OS
  208. // resolves. Inside the single-file Python-SDK executable the packaged `py/`
  209. // directory lives in pkg's virtual filesystem, which Node reads through its
  210. // patched `fs` but `python3` cannot see, so spawning from that path fails
  211. // with ENOENT. The scripts are therefore copied to a real directory first.
  212. //
  213. // The path is read from the child's own `__main__` module, so it proves
  214. // where the interpreter actually loaded the entry script — asserting on a
  215. // host-side constant would only restate the source. The program namespace
  216. // seeds `__name__` but no `__file__`, hence the module lookup.
  217. // `protocol.py` must land in the SAME directory, since `bootstrap.py` puts
  218. // its own directory on `sys.path` to import it; the run completing at all
  219. // already exercises that import.
  220. const { runtime } = await setup()
  221. const entryOf = async (): Promise<string> => {
  222. const result = await runtime.run({ program: 'import sys\nreturn sys.modules["__main__"].__file__', bindings: [] })
  223. expect(result.error).toBeUndefined()
  224. return result.value as string
  225. }
  226. const entry = await entryOf()
  227. expect(entry.endsWith('/bootstrap.py')).toBe(true)
  228. const dir = dirname(entry)
  229. expect(dir.startsWith(realpathSync(tmpdir()))).toBe(true)
  230. expect(dir).not.toContain('/packages/')
  231. // Staging is per RUN and removed at settlement, so by the time `run()`
  232. // resolved the directory is already gone — nothing survives to be rewritten
  233. // by a later run. `protocol.py` had to be beside the entry script for the run
  234. // to complete at all, since `bootstrap.py` imports it off `sys.path`.
  235. expect(existsSync(dir)).toBe(false)
  236. // A second run stages its own copy rather than reusing the first.
  237. expect(dirname(await entryOf())).not.toBe(dir)
  238. })
  239. it('contains a program that rewrites its own bootstrap to the run that did it', async () => {
  240. // The child runs as the same UID as the host, so `0o700` does not stop model
  241. // code from rewriting the scripts it was started from —
  242. // `sys.modules['__main__'].__file__` names them. While all runs shared one
  243. // staged copy, a program that overwrote `bootstrap.py` broke the NEXT run
  244. // (measured: it settled as `worker-exit`), and substituted code would have
  245. // run before the resource limits were applied.
  246. const { runtime } = await setup({ maxWallMs: 10_000 })
  247. const sabotage = await runtime.run({
  248. program: [
  249. 'import sys',
  250. 'path = sys.modules["__main__"].__file__',
  251. 'open(path, "w").write("raise SystemExit(1)\\n")',
  252. 'return path',
  253. ].join('\n'),
  254. bindings: [],
  255. })
  256. expect(sabotage.error).toBeUndefined()
  257. // The damage stayed inside the run that caused it.
  258. const after = await runtime.run({ program: 'return 1 + 1', bindings: [] })
  259. expect(after.error).toBeUndefined()
  260. expect(after.value).toBe(2)
  261. }, 20_000)
  262. it('leaves no subprocess or scripts behind when disposal races the first run', async () => {
  263. // Staging runs SYNCHRONOUSLY so no async boundary opens between `run()` and
  264. // the point where `execute` registers the run in `live` and installs the
  265. // abort listener. With an `await` there, a disposal landing in that window
  266. // saw an empty `live`, returned, removed the script directory, and let the
  267. // continuation spawn a subprocess after the fiber was gone.
  268. //
  269. // `dispose()` is called in the same synchronous turn as `run()`, with no
  270. // `await` between them, so it lands exactly in that window.
  271. //
  272. // The leak assertion compares before and after rather than requiring an
  273. // empty tmpdir: other tests in this file build runtimes they never dispose,
  274. // so only the directories this test adds are its own evidence.
  275. const staged = (): string[] =>
  276. readdirSync(realpathSync(tmpdir())).filter(name => name.startsWith('dsh-code-runtime-python-'))
  277. const before = new Set(staged())
  278. const { fiber, runtime } = await setup({ maxWallMs: 8_000 })
  279. const pending = runtime.run({ program: 'import time\nwhile True: time.sleep(0.1)', bindings: [] })
  280. const disposed = fiber.dispose()
  281. const result = await pending
  282. await disposed
  283. // Whatever the run reports, it must be terminal and must not be a success.
  284. expect(result.value).toBeUndefined()
  285. expect(['abort', 'worker-exit', 'timeout']).toContain(result.error?.kind)
  286. // Disposal is to quiescence, so this run's directory is gone once it
  287. // resolves, and nothing recreated it afterwards.
  288. expect(staged().filter(name => !before.has(name))).toEqual([])
  289. }, 15_000)
  290. it('settles as abort when the signal fires in the same turn as the first run', async () => {
  291. // Same window, the other listener. `addEventListener('abort')` does not
  292. // replay an event that already fired, so an abort landing before the
  293. // listener was installed used to be missed entirely and the program ran to
  294. // success or the wall ceiling instead of resolving as `abort`. Synchronous
  295. // staging keeps the pre-flight check and the listener in one turn, leaving
  296. // no gap for the signal to slip through.
  297. const { runtime } = await setup({ maxWallMs: 4_000, graceMs: 200 })
  298. const controller = new AbortController()
  299. const pending = runtime.run({
  300. program: 'import time\nwhile True: time.sleep(0.1)',
  301. bindings: [],
  302. signal: controller.signal,
  303. })
  304. controller.abort('same-turn-abort')
  305. const result = await pending
  306. expect(result.error?.kind).toBe('abort')
  307. expect(result.error?.message).toContain('same-turn-abort')
  308. }, 15_000)
  309. it('reports a staging failure as worker-exit instead of rejecting run()', async () => {
  310. // Staging touches the filesystem, so it can fail for reasons that are not
  311. // the caller's doing: a full or read-only temp filesystem, or a deployment
  312. // that failed to ship the packaged scripts. Those are SUBSTRATE failures,
  313. // the same class as a child that cannot start, and the seam reserves
  314. // rejection for misuse — so `run()` must resolve, not throw.
  315. //
  316. // `TMPDIR` is the honest lever: `mkdtempSync` builds its path from
  317. // `os.tmpdir()`, so pointing it at a path that is not a directory makes the
  318. // real call fail without stubbing the module under test.
  319. const previous = process.env.TMPDIR
  320. const notADirectory = join(await mkdtemp(join(tmpdir(), 'dsh-staging-')), 'file')
  321. await writeFile(notADirectory, '')
  322. process.env.TMPDIR = notADirectory
  323. try {
  324. const { runtime } = await setup()
  325. const result = await runtime.run({ program: 'return 1', bindings: [] })
  326. expect(result.error?.kind).toBe('worker-exit')
  327. expect(result.error?.message).toContain('failed to stage the python bootstrap')
  328. expect(result.logs).toEqual([])
  329. } finally {
  330. if (previous === undefined) delete process.env.TMPDIR
  331. else process.env.TMPDIR = previous
  332. }
  333. })
  334. it('leaves no staging directory behind when a script copy fails', async () => {
  335. // `mkdtempSync` succeeding and a later `copyFileSync` failing is its own
  336. // case: the directory exists but is only partially populated. Recording it
  337. // before the copies would leak it, because `run` retries staging on the next
  338. // call and overwrites the single recorded path — teardown could then remove
  339. // only the newest attempt. Staging must clean up its own partial directory.
  340. //
  341. // Only `copyFileSync` is stubbed, and only for the second script, so
  342. // `mkdtempSync` really runs and the directory under assertion is real.
  343. const staged = (): string[] =>
  344. readdirSync(realpathSync(tmpdir())).filter(name => name.startsWith('dsh-code-runtime-python-'))
  345. const before = new Set(staged())
  346. failNextCopyOf.value = 'protocol.py'
  347. try {
  348. const { runtime } = await setup()
  349. const result = await runtime.run({ program: 'return 1', bindings: [] })
  350. expect(result.error?.kind).toBe('worker-exit')
  351. expect(result.error?.message).toContain('failed to stage the python bootstrap')
  352. // The partial directory is gone, so nothing accumulates across retries.
  353. expect(staged().filter(name => !before.has(name))).toEqual([])
  354. } finally {
  355. failNextCopyOf.value = undefined
  356. }
  357. }, 15_000)
  358. })
  359. describe('PythonCodeRuntime — inherited resource limits', () => {
  360. it('runs under an inherited hard limit tighter than addressSpaceMb', async () => {
  361. // An unprivileged process may lower a hard rlimit but never raise it. Under
  362. // a harness started with `ulimit -v` below `addressSpaceBytes`, requesting
  363. // the configured cap made `setrlimit` raise `ValueError` and every run
  364. // returned a bootstrap exception — even though the inherited limit is
  365. // STRONGER than the one asked for. The bootstrap clamps to the inherited
  366. // hard limit instead, so the run proceeds under the stricter bound.
  367. //
  368. // `pythonBin` is the honest lever: a wrapper that lowers RLIMIT_AS and then
  369. // execs the real interpreter reproduces the inherited-limit condition
  370. // without touching this test process's own limits.
  371. const dir = await mkdtemp(join(tmpdir(), 'dsh-rlimit-'))
  372. const wrapper = join(dir, 'python3-capped')
  373. // 256 MiB, half the 512 MiB addressSpaceMb default, so the requested cap is
  374. // unambiguously above the inherited ceiling.
  375. await writeFile(wrapper, '#!/bin/sh\nulimit -v 262144\nexec python3 "$@"\n', { mode: 0o755 })
  376. const { runtime } = await setup({ pythonBin: wrapper })
  377. const result = await runtime.run({
  378. program: 'import resource\nreturn resource.getrlimit(resource.RLIMIT_AS)[1]',
  379. bindings: [],
  380. })
  381. expect(result.error).toBeUndefined()
  382. // The applied hard limit is the inherited one, not the configured 512 MiB.
  383. expect(result.value).toBe(256 * 1024 * 1024)
  384. }, 15_000)
  385. it('applies the configured limits when nothing tighter is inherited', async () => {
  386. // The clamp must not weaken the normal path: with an infinite inherited hard
  387. // limit there is nothing to clamp against, and RLIM_INFINITY compares as -1,
  388. // so treating it as a numeric bound would collapse every limit to -1.
  389. const { runtime } = await setup({ cpuSeconds: 42, addressSpaceMb: 400 })
  390. const result = await runtime.run({
  391. // `getrlimit` returns a tuple, which the lossless-JSON completion check
  392. // rejects; the pair is listed explicitly rather than converted.
  393. program: 'import resource\ncpu = resource.getrlimit(resource.RLIMIT_CPU)\nreturn [cpu[0], cpu[1], resource.getrlimit(resource.RLIMIT_AS)[1]]',
  394. bindings: [],
  395. })
  396. expect(result.error).toBeUndefined()
  397. // Soft at cpuSeconds, hard at +1 (the SIGKILL backstop), address space at
  398. // the configured megabytes — exactly what the unclamped path applied.
  399. expect(result.value).toEqual([42, 43, 400 * 1024 * 1024])
  400. }, 15_000)
  401. it('preserves an inherited soft limit stricter than the configured cap', async () => {
  402. // Clamping reads BOTH inherited bounds, not just the hard one. A deployment
  403. // that inherited a soft rlimit below the configured cap must keep that
  404. // stricter soft: returning the configured value would RAISE the effective
  405. // soft limit, loosening containment. The wrapper lowers only the SOFT CPU
  406. // limit (`ulimit -S -t`) and leaves the hard limit unlimited, so the
  407. // requested soft (`cpuSeconds`) sits above the inherited soft — the case that
  408. // exposed the bug. RLIMIT_CPU is used because macOS ignores `ulimit -v`
  409. // (RLIMIT_AS), which is exactly why the backend skips address space there.
  410. const dir = await mkdtemp(join(tmpdir(), 'dsh-rlimit-soft-'))
  411. const wrapper = join(dir, 'python3-soft-capped')
  412. // Soft CPU 5 s, well below the configured 30 s, hard left unlimited.
  413. await writeFile(wrapper, '#!/bin/sh\nulimit -S -t 5\nexec python3 "$@"\n', { mode: 0o755 })
  414. const { runtime } = await setup({ pythonBin: wrapper, cpuSeconds: 30 })
  415. const result = await runtime.run({
  416. program: 'import resource\nreturn resource.getrlimit(resource.RLIMIT_CPU)[0]',
  417. bindings: [],
  418. })
  419. expect(result.error).toBeUndefined()
  420. // The applied SOFT limit is the inherited 5 s, not the configured 30 s.
  421. expect(result.value).toBe(5)
  422. }, 15_000)
  423. it('rechecks CPU at settlement against the effective inherited soft limit', async () => {
  424. // The settlement-time CPU recheck must compare against the EFFECTIVE soft
  425. // limit (`_clamped` may have lowered it to a stricter inherited value), not
  426. // the configured `cpuSeconds`. A program that traps SIGXCPU, burns past the
  427. // inherited soft, and returns inside the soft-to-hard gap would otherwise be
  428. // compared to the configured value and falsely reported successful, bypassing
  429. // the inherited limit. The wrapper sets a 1 s soft CPU limit; the program
  430. // traps SIGXCPU and busy-loops past it, then returns — the recheck must
  431. // re-deliver SIGXCPU so the host classifies the run as a timeout.
  432. const dir = await mkdtemp(join(tmpdir(), 'dsh-cpu-recheck-'))
  433. const wrapper = join(dir, 'python3-cpu-capped')
  434. await writeFile(wrapper, '#!/bin/sh\nulimit -S -t 1\nexec python3 "$@"\n', { mode: 0o755 })
  435. const { runtime } = await setup({ pythonBin: wrapper, cpuSeconds: 30, maxWallMs: 12_000 })
  436. const result = await runtime.run({
  437. program: [
  438. 'import signal, time',
  439. // Trap SIGXCPU so the soft limit does not terminate the program; burn
  440. // CPU well past the inherited 1 s soft, then return normally.
  441. 'signal.signal(signal.SIGXCPU, lambda *a: None)',
  442. 'end = time.process_time() + 2.5',
  443. 'while time.process_time() < end:',
  444. ' pass',
  445. 'return "returned"',
  446. ].join('\n'),
  447. bindings: [],
  448. })
  449. // The recheck compares spent CPU against the effective 1 s soft, not 30 s, so
  450. // the run is a timeout rather than a false success.
  451. expect(result.error?.kind).toBe('timeout')
  452. }, 20_000)
  453. })
  454. describe('PythonCodeRuntime — programs and bindings', () => {
  455. it('runs a top-level script, captures print output, and returns `result`', async () => {
  456. const { runtime } = await setup()
  457. const result = await runtime.run({
  458. program: [
  459. 'x = 40',
  460. 'y = 2',
  461. 'print("hello", x + y)',
  462. 'return {"answer": x + y}',
  463. ].join('\n'),
  464. bindings: [],
  465. })
  466. expect(result.error).toBeUndefined()
  467. expect(result.value).toEqual({ answer: 42 })
  468. // `print` in Python emits: text, ' ', text, '\n'. Concat the captured
  469. // fragments and assert the model-visible message survives.
  470. expect(result.logs.join('')).toContain('hello 42')
  471. // 15s: this is usually the suite's first real subprocess — a cold python3
  472. // start (interpreter + asyncio import) on a loaded CI runner can exceed
  473. // the 5s default alone; later tests reuse the warm page cache.
  474. }, 15_000)
  475. it('bridges binding calls both ways and rejects the program-side call on a host rejection', async () => {
  476. const { runtime } = await setup()
  477. const calls: unknown[] = []
  478. const result = await runtime.run({
  479. program: [
  480. 'first = await tools.echo({"n": 1})',
  481. 'caught = ""',
  482. 'try:',
  483. ' await tools.fail({})',
  484. 'except RuntimeError as e:',
  485. ' caught = str(e)',
  486. 'return {"first": first, "caught": caught}',
  487. ].join('\n'),
  488. bindings: tools({
  489. echo: async (args) => { calls.push(args); return { echoed: args as CodeJsonValue } },
  490. fail: async () => { throw new Error('nope') },
  491. }),
  492. })
  493. expect(result.error).toBeUndefined()
  494. expect(result.value).toEqual({ first: { echoed: { n: 1 } }, caught: 'nope' })
  495. expect(calls).toEqual([{ n: 1 }])
  496. })
  497. it('still answers the call when the rejection value cannot be converted to a string', async () => {
  498. // `messageOf` calls `String(error)`, which runs the value's own conversion,
  499. // and this call site is a DETACHED async reply callback. A rejection whose
  500. // `Symbol.toPrimitive` throws therefore escaped as an unhandled rejection:
  501. // the reply frame was never written, the program stayed blocked on `await`,
  502. // and the run degraded to a `maxWallMs` timeout (observed) — a host with no
  503. // `unhandledRejection` listener would exit instead. The rejection must reach
  504. // the program as an ordinary error carrying a fixed placeholder.
  505. const { runtime } = await setup({ maxWallMs: 8_000 })
  506. const result = await runtime.run({
  507. program: [
  508. 'try:',
  509. ' await tools.hostile({})',
  510. 'except RuntimeError as e:',
  511. ' return "rejected: " + str(e)',
  512. 'return "no rejection"',
  513. ].join('\n'),
  514. bindings: tools({
  515. hostile: async () => {
  516. throw { [Symbol.toPrimitive]() { throw new Error('toPrimitive blew up') } }
  517. },
  518. }),
  519. })
  520. expect(result.error).toBeUndefined()
  521. expect(result.value).toBe('rejected: <unrenderable rejection value>')
  522. }, 15_000)
  523. it('still answers the call when an Error carries a cyclic value in place of its message', async () => {
  524. // `Error.message` is typed `string` but is a plain writable property, so a
  525. // rejection can carry any value there. Returning it verbatim handed a
  526. // non-string to `sendReply`, breaching `encodeJsonPlain`'s JSON-plain
  527. // precondition: a cyclic object grew the encoder stack until the host threw
  528. // RangeError from the detached reply callback, so no reply frame was written
  529. // and the run degraded to a `maxWallMs` timeout (observed). The conversion
  530. // must contain it — `String()` on a cycle throws inside the guard and lands
  531. // on the placeholder, so the program sees an ordinary error.
  532. const { runtime } = await setup({ maxWallMs: 8_000 })
  533. const result = await runtime.run({
  534. program: [
  535. 'try:',
  536. ' await tools.hostile({})',
  537. 'except RuntimeError as e:',
  538. ' return "rejected: " + str(e)',
  539. 'return "no rejection"',
  540. ].join('\n'),
  541. bindings: tools({
  542. hostile: async () => {
  543. const cyclic: { self?: unknown; [Symbol.toPrimitive]: () => string } = {
  544. // A cycle alone is inert for `String()`; the throwing conversion is
  545. // what proves the guard runs rather than the encoder.
  546. [Symbol.toPrimitive]: () => { throw new Error('cyclic message') },
  547. }
  548. cyclic.self = cyclic
  549. const error = new Error('placeholder')
  550. // Writable per spec, so no cast is needed to install a non-string.
  551. ;(error as unknown as { message: unknown }).message = cyclic
  552. throw error
  553. },
  554. }),
  555. })
  556. expect(result.error).toBeUndefined()
  557. expect(result.value).toBe('rejected: <unrenderable rejection value>')
  558. }, 15_000)
  559. it('renders an Error whose message is a value with no JSON form', async () => {
  560. // The non-cyclic arm. A number would not discriminate: `scalarJson` renders
  561. // it as digits and the child `str()`s the field back, so it survives the
  562. // wire either way. `undefined` is the value that separates the two orders —
  563. // `scalarJson` emits a bare `undefined` token, so the reply line is not JSON
  564. // at all, the child's parse drops the frame, and the program stays blocked
  565. // on `await` until the wall ceiling (observed). Converting first sends the
  566. // string "undefined", which the program receives as an ordinary rejection.
  567. const { runtime } = await setup({ maxWallMs: 8_000 })
  568. const result = await runtime.run({
  569. program: [
  570. 'try:',
  571. ' await tools.absent({})',
  572. 'except RuntimeError as e:',
  573. ' return "rejected: " + str(e)',
  574. 'return "no rejection"',
  575. ].join('\n'),
  576. bindings: tools({
  577. absent: async () => {
  578. const error = new Error('placeholder')
  579. ;(error as unknown as { message: unknown }).message = undefined
  580. throw error
  581. },
  582. }),
  583. })
  584. expect(result.error).toBeUndefined()
  585. expect(result.value).toBe('rejected: undefined')
  586. }, 15_000)
  587. it('runs a program with no await', async () => {
  588. const { runtime } = await setup()
  589. const result = await runtime.run({
  590. program: 'return 2 + 2',
  591. bindings: [],
  592. })
  593. expect(result.error).toBeUndefined()
  594. expect(result.value).toBe(4)
  595. })
  596. it('returns JSON null whether the program returns None or falls off the end', async () => {
  597. // Python has no `undefined`: an async body that returns None and one that
  598. // never returns both yield None, so both complete as an exact JSON null.
  599. // (The worker/TS backend can tell `return undefined` from `return null`;
  600. // Python cannot, and reporting null for both is the honest rendering.)
  601. const { runtime } = await setup()
  602. const explicit = await runtime.run({ program: 'return None', bindings: [] })
  603. expect(explicit.error).toBeUndefined()
  604. expect(explicit.value).toBeNull()
  605. const noReturn = await runtime.run({ program: 'x = 1', bindings: [] })
  606. expect(noReturn.error).toBeUndefined()
  607. expect(noReturn.value).toBeNull()
  608. })
  609. it('settles with no value on a forged valueless done frame', async () => {
  610. // The child always sends a value now (return None → JSON null), so a done
  611. // frame with no value key can only be forged; the host settles it as a
  612. // value-less completion rather than crashing on the absent field.
  613. const { runtime } = await setup()
  614. const result = await runtime.run({
  615. program: [
  616. 'import os',
  617. 'os.write(3, b\'{"type":"done"}\\n\')',
  618. 'import time',
  619. 'time.sleep(5)',
  620. ].join('\n'),
  621. bindings: [],
  622. })
  623. expect(result.error).toBeUndefined()
  624. expect(result.value).toBeUndefined()
  625. })
  626. it('coalesces print arguments into one log line, not per-write fragments', async () => {
  627. // print("a","b") calls write() per arg/sep/newline; the stream must emit
  628. // one logical line "a b" so Code Mode's join(newline) does not insert
  629. // spurious blank lines. Two prints → exactly two entries, no empties.
  630. const { runtime } = await setup()
  631. const result = await runtime.run({
  632. program: ['print("a", "b")', 'print("c")', 'return None'].join('\n'),
  633. bindings: [],
  634. })
  635. expect(result.error).toBeUndefined()
  636. expect(result.logs).toEqual(['a b', 'c'])
  637. })
  638. it('flushes a print with no trailing newline', async () => {
  639. const { runtime } = await setup()
  640. const result = await runtime.run({
  641. program: ['print("partial", end="")', 'return None'].join('\n'),
  642. bindings: [],
  643. })
  644. expect(result.error).toBeUndefined()
  645. expect(result.logs).toEqual(['partial'])
  646. })
  647. it('fails a completion dict with a non-string key as invalid-output (no key coercion)', async () => {
  648. // json.dumps would coerce {1: "a", "1": "b"} to a single "1" key, silently
  649. // dropping data. The shape validator rejects it before encoding.
  650. const { runtime } = await setup()
  651. const result = await runtime.run({
  652. program: 'return {1: "first", "1": "second"}',
  653. bindings: [],
  654. })
  655. expect(result.value).toBeUndefined()
  656. expect(result.error?.kind).toBe('invalid-output')
  657. expect(result.error?.message).toContain('non-string dict key')
  658. })
  659. it('rejects a binding argument with a non-string dict key before dispatch', async () => {
  660. const { runtime } = await setup()
  661. let called = false
  662. const result = await runtime.run({
  663. program: [
  664. 'caught = ""',
  665. 'try:',
  666. ' await tools.sink({1: "x"})',
  667. 'except RuntimeError as e:',
  668. ' caught = str(e)',
  669. 'return caught',
  670. ].join('\n'),
  671. bindings: tools({ sink: async () => { called = true; return null } }),
  672. })
  673. expect(result.error).toBeUndefined()
  674. expect(result.value).toContain('lossless JSON')
  675. expect(called).toBe(false)
  676. })
  677. it('fails a non-JSON completion value as invalid-output (no repr substitution)', async () => {
  678. // A set is not lossless JSON. The old draft substituted repr(); the seam
  679. // now requires refusing the run instead.
  680. const { runtime } = await setup()
  681. const result = await runtime.run({
  682. program: 'return {1, 2, 3}',
  683. bindings: [],
  684. })
  685. expect(result.value).toBeUndefined()
  686. expect(result.error?.kind).toBe('invalid-output')
  687. expect(result.error?.message).toContain('lossless JSON')
  688. expect(result.error?.message).toContain('set')
  689. })
  690. it('fails a negative-zero completion value as invalid-output (sign bit is lossy over JSON)', async () => {
  691. // JSON serialization turns -0.0 into 0 (or JS -0), silently changing the
  692. // sign bit; the canonical lossless-JSON boundary rejects it, so the
  693. // Python side must too — as a completion and as a binding argument.
  694. const { runtime } = await setup()
  695. const completion = await runtime.run({
  696. program: 'return -0.0',
  697. bindings: [],
  698. })
  699. expect(completion.error?.kind).toBe('invalid-output')
  700. expect(completion.error?.message).toContain('negative zero')
  701. const argument = await runtime.run({
  702. program: [
  703. 'try:',
  704. ' await tools.echo(-0.0)',
  705. ' return "accepted"',
  706. 'except RuntimeError as e:',
  707. ' return str(e)',
  708. ].join('\n'),
  709. bindings: tools({ echo: async args => args as never }),
  710. })
  711. expect(argument.error).toBeUndefined()
  712. expect(argument.value).toContain('negative zero')
  713. })
  714. it('fails a NaN completion value as invalid-output (allow_nan=False)', async () => {
  715. // json.dumps would happily emit NaN by default, but NaN is not JSON; the
  716. // bootstrap passes allow_nan=False so it fails as invalid-output.
  717. const { runtime } = await setup()
  718. const result = await runtime.run({
  719. program: 'return float("nan")',
  720. bindings: [],
  721. })
  722. expect(result.error?.kind).toBe('invalid-output')
  723. })
  724. it('fails an over-budget completion value as output-limit (child-side check)', async () => {
  725. const { runtime } = await setup({ maxValueBytes: 64 })
  726. const result = await runtime.run({
  727. program: 'return "V" * 5000',
  728. bindings: [],
  729. })
  730. expect(result.value).toBeUndefined()
  731. expect(result.error?.kind).toBe('output-limit')
  732. expect(result.error?.message).toContain('exceeded 64 bytes')
  733. })
  734. it('rejects a wide completion as output-limit before materializing its traversal state', async () => {
  735. // `[0] * 2000000` sits far above maxValueBytes but well below the frame
  736. // ceiling. The folded checker must reject it via the pre-enqueue bound —
  737. // BEFORE pushing two million elements onto the walk — so a small
  738. // addressSpaceMb does not turn the check itself into an RLIMIT_AS death.
  739. const { runtime } = await setup({ maxValueBytes: 64, addressSpaceMb: 256, maxWallMs: 15_000 })
  740. const result = await runtime.run({
  741. program: 'return [0] * 2000000',
  742. bindings: [],
  743. })
  744. expect(result.value).toBeUndefined()
  745. expect(result.error?.kind).toBe('output-limit')
  746. expect(result.error?.message).toContain('exceeded 64 bytes')
  747. }, 20_000)
  748. it('rejects a wide dict as output-limit without materializing its items list', async () => {
  749. // Same pre-enqueue bound on the dict branch: `len(current)` replaces
  750. // `list(current.items())`, which allocated one tuple per member before the
  751. // bound could reject the value. Two million entries under a 64-byte cap
  752. // fits the 256 MiB address space as a dict but not as a dict PLUS a
  753. // two-million-tuple list.
  754. const { runtime } = await setup({ maxValueBytes: 64, addressSpaceMb: 256, maxWallMs: 15_000 })
  755. const result = await runtime.run({
  756. program: 'return {str(i): 0 for i in range(2000000)}',
  757. bindings: [],
  758. })
  759. expect(result.value).toBeUndefined()
  760. expect(result.error?.kind).toBe('output-limit')
  761. expect(result.error?.message).toContain('exceeded 64 bytes')
  762. }, 20_000)
  763. it('meters a float completion in the host\'s number spelling', async () => {
  764. // CPython's repr disagrees with the host's String(number): `1.0` is three
  765. // bytes here and one there, `1e-07` pads the exponent the host writes as
  766. // `1e-7`. Both sides meter the SAME budget, so the child must count the
  767. // bytes the host will receive — otherwise a boundary-sized value is
  768. // falsely reported as output-limit.
  769. const { runtime } = await setup({ maxValueBytes: 1 })
  770. const integral = await runtime.run({ program: 'return 1.0', bindings: [] })
  771. expect(integral.error).toBeUndefined()
  772. expect(integral.value).toBe(1)
  773. const exponent = await setup({ maxValueBytes: 4 })
  774. const small = await exponent.runtime.run({ program: 'return 1e-7', bindings: [] })
  775. expect(small.error).toBeUndefined()
  776. expect(small.value).toBe(1e-7)
  777. // The spelling is a meter input, not a licence to overshoot: `1.5` is three
  778. // bytes on both sides and still fails a two-byte budget.
  779. const tight = await setup({ maxValueBytes: 2 })
  780. const over = await tight.runtime.run({ program: 'return 1.5', bindings: [] })
  781. expect(over.error?.kind).toBe('output-limit')
  782. })
  783. it('carries floats across the wire in the host\'s number spelling', async () => {
  784. // The child ENCODES with the same speller it meters with, so the frame the
  785. // host parses must reproduce every double exactly — including the branches
  786. // where CPython and ECMAScript disagree (integral floats, sub-1e-6
  787. // exponents, >= 1e21, and beyond-safe-range integral doubles whose exact
  788. // digits differ from the shortest round-trip form).
  789. const { runtime } = await setup()
  790. const result = await runtime.run({
  791. 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]',
  792. bindings: [],
  793. })
  794. expect(result.error).toBeUndefined()
  795. 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])
  796. })
  797. it('rejects a forged non-lossless done value host-side as invalid-output', async () => {
  798. // A forged done frame bypasses the child's _check_done_value. JSON.parse
  799. // turns 1e400 into Infinity; validateChildFrame no longer scans done.value,
  800. // so the host's own checkDoneValue must catch the non-lossless number.
  801. const { runtime } = await setup()
  802. const result = await runtime.run({
  803. program: [
  804. 'import os',
  805. String.raw`os.write(3, b'{"type":"done","value":1e400}' + b'\n')`,
  806. 'import time',
  807. 'time.sleep(5)',
  808. ].join('\n'),
  809. bindings: [],
  810. })
  811. expect(result.value).toBeUndefined()
  812. expect(result.error?.kind).toBe('invalid-output')
  813. expect(result.error?.message).toContain('non-lossless number')
  814. })
  815. it('reports a syntax error as an exception without settling with a value', async () => {
  816. const { runtime } = await setup()
  817. const result = await runtime.run({
  818. program: '$$invalid python$$',
  819. bindings: [],
  820. })
  821. expect(result.error?.kind).toBe('exception')
  822. expect(result.error?.message).toContain('SyntaxError')
  823. expect(result.value).toBeUndefined()
  824. })
  825. it('reports a runtime raise as an exception with the traceback', async () => {
  826. const { runtime } = await setup()
  827. const result = await runtime.run({
  828. program: 'raise ValueError("intentional")',
  829. bindings: [],
  830. })
  831. expect(result.error?.kind).toBe('exception')
  832. expect(result.error?.message).toContain('ValueError')
  833. expect(result.error?.message).toContain('intentional')
  834. })
  835. it('bounds a deep exception cause chain instead of burning the wall budget formatting it', async () => {
  836. // A chain thousands of links deep would make the rendering walk and
  837. // format() linear in its length, consuming maxWallMs. Rendering is capped
  838. // at 100 links with a marker; the run reports the exception well within
  839. // budget rather than timing out.
  840. const { runtime } = await setup({ maxValueBytes: 1024 * 1024, maxWallMs: 20_000 })
  841. const start = Date.now()
  842. const result = await runtime.run({
  843. program: [
  844. 'err = None',
  845. 'for i in range(3000):',
  846. ' try:',
  847. ' raise ValueError(i) from err',
  848. ' except ValueError as e:',
  849. ' err = e',
  850. 'raise err',
  851. ].join('\n'),
  852. bindings: [],
  853. })
  854. expect(result.error?.kind).toBe('exception')
  855. expect(result.error?.message).toContain('exception chain truncated at 100 links')
  856. expect(Date.now() - start).toBeLessThan(15_000)
  857. }, 25_000)
  858. it('bounds an over-cap chain without assigning to the live exception', async () => {
  859. // The cap used to be applied by severing the over-cap link ON the live
  860. // exception. An exception class overriding __setattr__ to raise turned that
  861. // assignment into model code running inside the bootstrap's failure
  862. // handler; the throw skipped the `done` send that sits after the handler,
  863. // so the host blocked on fd 3 and reported a maxWallMs timeout instead of
  864. // the model's own exception. Cutting the chain on the TracebackException
  865. // COPY touches no model hook, so the marker still appears and the run
  866. // reports `exception`.
  867. const { runtime } = await setup({ maxValueBytes: 1024 * 1024, maxWallMs: 15_000 })
  868. const start = Date.now()
  869. const result = await runtime.run({
  870. program: [
  871. 'class Sealed(Exception):',
  872. ' def __setattr__(self, name, value):',
  873. ' raise RuntimeError("live mutation refused")',
  874. 'err = None',
  875. 'for i in range(150):',
  876. ' try:',
  877. ' raise Sealed(i) from err',
  878. ' except Sealed as e:',
  879. ' err = e',
  880. 'raise err',
  881. ].join('\n'),
  882. bindings: [],
  883. })
  884. expect(result.error?.kind).toBe('exception')
  885. expect(result.error?.message).toContain('Sealed')
  886. expect(result.error?.message).toContain('exception chain truncated at 100 links')
  887. // The sever attempt is what used to leak: its message must not appear, and
  888. // the run must settle well inside the wall budget rather than timing out.
  889. expect(result.error?.message).not.toContain('live mutation refused')
  890. expect(Date.now() - start).toBeLessThan(10_000)
  891. }, 20_000)
  892. it('still sends done when rendering the diagnostic itself raises', async () => {
  893. // format() reaches the exception's own __str__, so a model class whose
  894. // __str__ raises can throw from inside the failure handler. CPython's
  895. // _safe_string absorbs a raising __str__ during formatting, but the
  896. // fallback must hold for any throw on that path (a raising __repr__ of an
  897. // argument, a MemoryError under RLIMIT_AS), so the assertion is the
  898. // invariant that matters: a `done` frame carrying `exception`, never a
  899. // timeout, and never the failing renderer's own message.
  900. const { runtime } = await setup({ maxWallMs: 10_000 })
  901. const result = await runtime.run({
  902. program: [
  903. 'class Unprintable(Exception):',
  904. ' def __str__(self):',
  905. ' raise RuntimeError("str refused")',
  906. ' def __repr__(self):',
  907. ' raise RuntimeError("repr refused")',
  908. 'raise Unprintable()',
  909. ].join('\n'),
  910. bindings: [],
  911. })
  912. expect(result.error?.kind).toBe('exception')
  913. expect(result.error?.message).toContain('Unprintable')
  914. expect(result.error?.message).not.toContain('str refused')
  915. expect(result.error?.message).not.toContain('repr refused')
  916. }, 15_000)
  917. it('sends done with an inert diagnostic when the whole rendering path raises', async () => {
  918. // Drive the fallback itself. `TracebackException.format` reads the
  919. // exception class's `__module__` to decide whether to qualify the name, and
  920. // a metaclass property can raise there — a throw INSIDE the formatter,
  921. // reached with no rebinding of anything the bootstrap owns. Without the
  922. // wrapper it escapes the handler, the `done` send never runs, and the host
  923. // times out at maxWallMs.
  924. const { runtime } = await setup({ maxWallMs: 10_000 })
  925. const result = await runtime.run({
  926. program: [
  927. 'class Meta(type):',
  928. ' @property',
  929. ' def __module__(cls):',
  930. ' raise RuntimeError("renderer refused")',
  931. 'class Hostile(ValueError, metaclass=Meta):',
  932. ' pass',
  933. 'raise Hostile("original failure")',
  934. ].join('\n'),
  935. bindings: [],
  936. })
  937. expect(result.error?.kind).toBe('exception')
  938. // The inert fallback names the class and a fixed literal; it must not carry
  939. // the renderer's message, and must not have become a timeout. `__name__` is
  940. // still a plain str here, so the class name survives.
  941. expect(result.error?.message).toBe('Hostile: <diagnostic rendering failed>')
  942. }, 15_000)
  943. it('falls back to a placeholder class name when __name__ itself raises', async () => {
  944. // The fallback reads type(exc).__name__, which a metaclass property can
  945. // hijack. It must neither run that override's failure into the handler nor
  946. // format a non-str __name__ into the message. The hostile `__module__` is
  947. // what drives execution into the fallback in the first place.
  948. const { runtime } = await setup({ maxWallMs: 10_000 })
  949. const result = await runtime.run({
  950. program: [
  951. 'class Meta(type):',
  952. ' @property',
  953. ' def __module__(cls):',
  954. ' raise RuntimeError("renderer refused")',
  955. ' @property',
  956. ' def __name__(cls):',
  957. ' raise RuntimeError("name refused")',
  958. 'class Nameless(Exception, metaclass=Meta):',
  959. ' pass',
  960. 'raise Nameless()',
  961. ].join('\n'),
  962. bindings: [],
  963. })
  964. expect(result.error?.kind).toBe('exception')
  965. expect(result.error?.message).toBe('<unknown>: <diagnostic rendering failed>')
  966. }, 15_000)
  967. it('reports the real exception when the program rebinds every name the failure path uses', async () => {
  968. // The bootstrap IS __main__, so `import __main__; __main__._X = ...` reaches
  969. // any module global a call-time lookup would read. The failure path is the
  970. // worst place for that: the reporter, the byte cap, the traceback formatter,
  971. // the settlement flush and the `done` send all run AFTER the `except` block,
  972. // so a replacement that raises skips the send, leaves the host blocked on
  973. // fd 3, and the run reports a maxWallMs timeout instead of the model's own
  974. // exception. Rebind all of them at once; the run must still carry the real
  975. // ValueError.
  976. const { runtime } = await setup({ maxWallMs: 10_000 })
  977. const result = await runtime.run({
  978. program: [
  979. 'import __main__',
  980. 'def boom(*a, **k):',
  981. ' raise RuntimeError("hijacked")',
  982. '__main__._SAFE_MODEL_TRACEBACK = boom',
  983. '__main__._cap_message = boom',
  984. '__main__._model_traceback = boom',
  985. '__main__._UNRENDERABLE_DIAGNOSTIC = boom',
  986. '__main__._LogStream.flush_line = boom',
  987. '__main__.ProtocolChannel.send_sync = boom',
  988. 'raise ValueError("real failure")',
  989. ].join('\n'),
  990. bindings: [],
  991. })
  992. expect(result.error?.kind).toBe('exception')
  993. expect(result.error?.message).toContain('ValueError: real failure')
  994. expect(result.error?.message).not.toContain('hijacked')
  995. }, 15_000)
  996. it('bounds an over-cap exception-group nesting on the copy', async () => {
  997. // Exception groups link through `exceptions`, not the cause/context
  998. // dunders, so the cap has to count that edge too — otherwise a deeply
  999. // nested group walks past the bound the marker claims to enforce.
  1000. const { runtime } = await setup({ maxValueBytes: 1024 * 1024, maxWallMs: 15_000 })
  1001. const result = await runtime.run({
  1002. program: [
  1003. 'group = ValueError("leaf")',
  1004. 'for i in range(150):',
  1005. ' group = ExceptionGroup(f"g{i}", [group])',
  1006. 'raise group',
  1007. ].join('\n'),
  1008. bindings: [],
  1009. })
  1010. expect(result.error?.kind).toBe('exception')
  1011. expect(result.error?.message).toContain('exception chain truncated at 100 links')
  1012. }, 20_000)
  1013. it('filters every bootstrap frame from the traceback of an uncaught binding rejection', async () => {
  1014. // A rejection re-raised by the bootstrap's dispatch adds bootstrap frames
  1015. // AFTER the model's own; only <model> frames may reach model-visible,
  1016. // durable output — a bootstrap.py path would leak host absolutes and make
  1017. // transcripts machine-dependent.
  1018. const { runtime } = await setup()
  1019. const result = await runtime.run({
  1020. program: 'await tools.boom({})',
  1021. bindings: tools({ boom: async () => { throw new Error('exploded') } }),
  1022. })
  1023. expect(result.error?.kind).toBe('exception')
  1024. expect(result.error?.message).toContain('exploded')
  1025. expect(result.error?.message).toContain('<model>')
  1026. expect(result.error?.message).not.toContain('bootstrap.py')
  1027. })
  1028. it('renders a non-Error thrown value from a host binding as its String form', async () => {
  1029. const { runtime } = await setup()
  1030. const result = await runtime.run({
  1031. program: [
  1032. 'caught = ""',
  1033. 'try:',
  1034. ' await tools.failRaw({})',
  1035. 'except RuntimeError as e:',
  1036. ' caught = str(e)',
  1037. 'return caught',
  1038. ].join('\n'),
  1039. bindings: tools({
  1040. failRaw: async () => { throw 'raw-nope' },
  1041. }),
  1042. })
  1043. expect(result.error).toBeUndefined()
  1044. expect(result.value).toContain('raw-nope')
  1045. })
  1046. it('reassembles a frame split across writes behind a completed one', async () => {
  1047. // One os.write carrying "<frame>\n<partial...>" leaves a non-empty
  1048. // residual after the newline loop; the tail must survive until its own
  1049. // newline arrives and then parse as a normal frame.
  1050. const { runtime } = await setup()
  1051. const result = await runtime.run({
  1052. program: [
  1053. 'import os, json',
  1054. 'head = json.dumps({"type":"log","text":"first"}).encode()',
  1055. 'tail = json.dumps({"type":"log","text":"second"}).encode()',
  1056. 'import time',
  1057. 'os.write(3, head + b"\\n" + tail[:5])',
  1058. 'time.sleep(0.2)',
  1059. 'os.write(3, tail[5:] + b"\\n")',
  1060. 'return "ok"',
  1061. ].join('\n'),
  1062. bindings: [],
  1063. })
  1064. expect(result.error).toBeUndefined()
  1065. expect(result.value).toBe('ok')
  1066. expect(result.logs).toContain('first')
  1067. expect(result.logs).toContain('second')
  1068. })
  1069. it('raises the declared errorClass with the member name on rejection', async () => {
  1070. // Code Mode declares { name: ToolCallError, memberNameProperty: toolName };
  1071. // a host rejection must surface as that class, carrying the failed tool.
  1072. const { runtime } = await setup()
  1073. const result = await runtime.run({
  1074. program: [
  1075. 'caught = ""',
  1076. 'try:',
  1077. ' await tools.fail({})',
  1078. 'except ToolCallError as e:',
  1079. ' caught = f"{type(e).__name__}:{e.toolName}:{e}"',
  1080. 'return caught',
  1081. ].join('\n'),
  1082. bindings: [{
  1083. global: 'tools',
  1084. functions: { fail: async () => { throw new Error('typed-nope') } },
  1085. errorClass: { name: 'ToolCallError', memberNameProperty: 'toolName' },
  1086. }],
  1087. })
  1088. expect(result.error).toBeUndefined()
  1089. expect(result.value).toBe('ToolCallError:fail:typed-nope')
  1090. })
  1091. it('rejects an errorClass name colliding with its namespace global at the seam', async () => {
  1092. const { runtime } = await setup()
  1093. await expect(runtime.run({
  1094. program: 'return 1',
  1095. bindings: [{ global: 'tools', functions: {}, errorClass: { name: 'tools', memberNameProperty: 'toolName' } }],
  1096. })).rejects.toThrow(/collides with another injected global/)
  1097. })
  1098. it('rejects a namespace global colliding with a runtime-owned name at the seam', async () => {
  1099. // `__dsh_main__` passes the identifier check, but exec()ing the generated
  1100. // wrapper would silently overwrite the binding after injection. `console`
  1101. // is the WORKER backend's slot — refused here too so a namespace list
  1102. // valid on one backend is valid on all.
  1103. const { runtime } = await setup()
  1104. // `__debug__` is refused for a different reason than a collision: CPython
  1105. // compiles a bare `__debug__` reference to the constant True and refuses to
  1106. // assign the name at compile time, so an injected global under it is
  1107. // unreachable from the program — accepted by the seam, unusable here.
  1108. for (const global of ['__dsh_main__', 'console', '__debug__']) {
  1109. await expect(runtime.run({
  1110. program: 'x = 1',
  1111. bindings: [{ global, functions: {} }],
  1112. })).rejects.toThrow(/collides with a runtime-owned global/)
  1113. }
  1114. })
  1115. it('accepts a non-identifier memberNameProperty and rejects only an empty one', async () => {
  1116. // The seam permits any non-empty own property except the reserved
  1117. // members; Python setattr/getattr carry exotic names like `tool-name`,
  1118. // and the worker backend accepts them, so this backend must too.
  1119. const { runtime } = await setup()
  1120. const result = await runtime.run({
  1121. program: [
  1122. 'try:',
  1123. ' await tools.boom({})',
  1124. 'except ToolCallError as e:',
  1125. ' return getattr(e, "tool-name")',
  1126. ].join('\n'),
  1127. bindings: [{
  1128. global: 'tools',
  1129. functions: { boom: async () => { throw new Error('nope') } },
  1130. errorClass: { name: 'ToolCallError', memberNameProperty: 'tool-name' },
  1131. }],
  1132. })
  1133. expect(result.error).toBeUndefined()
  1134. expect(result.value).toBe('boom')
  1135. await expect(runtime.run({
  1136. program: 'return 1',
  1137. bindings: [{ global: 'tools', functions: {}, errorClass: { name: 'ToolCallError', memberNameProperty: '' } }],
  1138. })).rejects.toThrow(/memberNameProperty must be a non-empty attribute name/)
  1139. })
  1140. it('resolves a basename pythonBin to an absolute path (runs a real program)', async () => {
  1141. // A bare `python3` basename must resolve against PATH and actually launch
  1142. // under the empty-env spawn — exercises the accessSync success branch.
  1143. const { runtime } = await setup({ pythonBin: 'python3' })
  1144. const result = await runtime.run({ program: 'return 7', bindings: [] })
  1145. expect(result.error).toBeUndefined()
  1146. expect(result.value).toBe(7)
  1147. })
  1148. it('spawns via an absolute python path resolved from a basename against PATH', async () => {
  1149. // resolvePythonBin turns the default basename into an absolute path before
  1150. // the empty-env spawn; a basename with no PATH match falls through to the
  1151. // normal ENOENT worker-exit rather than throwing.
  1152. const { runtime } = await setup({ pythonBin: 'definitely-no-such-python-xyz' })
  1153. const result = await runtime.run({ program: 'return 1', bindings: [] })
  1154. expect(result.error?.kind).toBe('worker-exit')
  1155. })
  1156. it('rejects a memberNameProperty naming a constrained BaseException attribute', async () => {
  1157. // `__dict__`/`__class__` are constrained descriptors alongside
  1158. // `__traceback__` — setattr of a string raises TypeError while
  1159. // constructing the rejection — so every dunder is refused at the seam.
  1160. const { runtime } = await setup()
  1161. // name/message/stack are the seam's own exclusions (CodeBindingErrorClass
  1162. // forbids replacing them; the worker backend rejects them identically).
  1163. for (const member of ['__traceback__', '__dict__', '__class__', 'args', 'name', 'message', 'stack']) {
  1164. await expect(runtime.run({
  1165. program: 'return 1',
  1166. bindings: [{ global: 'tools', functions: {}, errorClass: { name: 'ToolCallError', memberNameProperty: member } }],
  1167. })).rejects.toThrow(/reserved error member/)
  1168. }
  1169. })
  1170. it('rejects a lossy binding resolution (NaN) instead of coercing it to null', async () => {
  1171. // JSON.stringify would turn NaN into null and drop undefined fields; the
  1172. // seam requires a descriptive rejection so data cannot silently corrupt.
  1173. const { runtime } = await setup()
  1174. const result = await runtime.run({
  1175. program: [
  1176. 'caught = ""',
  1177. 'try:',
  1178. ' await tools.bad({})',
  1179. 'except RuntimeError as e:',
  1180. ' caught = str(e)',
  1181. 'return caught',
  1182. ].join('\n'),
  1183. bindings: tools({ bad: async () => Number.NaN }),
  1184. })
  1185. expect(result.error).toBeUndefined()
  1186. expect(result.value).toContain('lossless JSON')
  1187. })
  1188. it('contains a forged pathological done value without crashing the host', async () => {
  1189. // A ~20k-deep nested array forged onto fd 3 would overflow a recursive
  1190. // JSON.stringify; the host's iterative encoder measures it stack-safely
  1191. // and fails it deterministically on the byte budget (40 kB > 32 KiB).
  1192. const { runtime } = await setup()
  1193. const result = await runtime.run({
  1194. program: [
  1195. 'import os, json',
  1196. 'depth = 20000',
  1197. 'payload = "[" * depth + "]" * depth',
  1198. 'os.write(3, b\'{"type":"done","value":\' + payload.encode() + b\'}\\n\')',
  1199. 'import time',
  1200. 'time.sleep(5)',
  1201. ].join('\n'),
  1202. bindings: [],
  1203. })
  1204. expect(result.value).toBeUndefined()
  1205. expect(result.error?.kind).toBe('output-limit')
  1206. })
  1207. it('preserves a deeply nested completion value below the byte budget', async () => {
  1208. // CodeJsonValue has no depth limit: a 10000-deep nested list is only
  1209. // ~20 kB — under maxValueBytes — and must cross intact. That depth
  1210. // overflows BOTH recursive serializers the pipeline used to rely on
  1211. // (CPython's json.dumps recursion limit ~1000s, V8's JSON.stringify), so
  1212. // it proves the child-side _encode_json_plain and the host-side
  1213. // encodeJsonPlain together. The host JSON.parse of the frame is iterative
  1214. // in V8 for arrays, so only the two encoders were at risk.
  1215. const { runtime } = await setup()
  1216. const result = await runtime.run({
  1217. program: [
  1218. 'v = None',
  1219. 'for _ in range(10000):',
  1220. ' v = [v]',
  1221. 'return v',
  1222. ].join('\n'),
  1223. bindings: [],
  1224. })
  1225. expect(result.error).toBeUndefined()
  1226. // Walk down iteratively (a recursive toEqual would itself overflow).
  1227. let depth = 0
  1228. let cursor: unknown = result.value
  1229. while (Array.isArray(cursor)) {
  1230. expect(cursor).toHaveLength(1)
  1231. cursor = cursor[0]
  1232. depth++
  1233. }
  1234. expect(depth).toBe(10000)
  1235. expect(cursor).toBeNull()
  1236. })
  1237. it('bridges a deeply nested binding resolution back into the program stack-safely', async () => {
  1238. // A binding resolution has no seam-level depth or byte cap; neither the
  1239. // host's reply serialization nor the CHILD's reply decode may die on
  1240. // recursion (json.loads raises RecursionError ~10k levels deep; the
  1241. // bootstrap decodes frames iteratively). 12000 levels sits past that
  1242. // limit while staying tiny in bytes.
  1243. const { runtime } = await setup()
  1244. const deep = ((): unknown => {
  1245. let v: unknown = null
  1246. for (let i = 0; i < 12000; i++) v = [v]
  1247. return v
  1248. })()
  1249. const result = await runtime.run({
  1250. program: [
  1251. 'v = await tools.deep({})',
  1252. 'depth = 0',
  1253. 'while isinstance(v, list):',
  1254. ' v = v[0]',
  1255. ' depth += 1',
  1256. 'return depth',
  1257. ].join('\n'),
  1258. bindings: tools({ deep: async () => deep as never }),
  1259. })
  1260. expect(result.error).toBeUndefined()
  1261. expect(result.value).toBe(12000)
  1262. })
  1263. it('rejects a reserved errorClass name at the seam', async () => {
  1264. const { runtime } = await setup()
  1265. await expect(runtime.run({
  1266. program: 'return 1',
  1267. bindings: [{
  1268. global: 'tools',
  1269. functions: {},
  1270. errorClass: { name: 'class', memberNameProperty: 'toolName' },
  1271. }],
  1272. })).rejects.toThrow(/errorClass.name "class" is not a usable Python identifier/)
  1273. })
  1274. it('routes a declared inherited-attribute name through the bridge via subscript', async () => {
  1275. // __class__ resolves on `object` before any fallback hook; the proxy's
  1276. // __getattribute__ intercepts declared names first, and subscript access
  1277. // is the SDK-advertised route for underscore names.
  1278. const { runtime } = await setup()
  1279. const seen: string[] = []
  1280. const result = await runtime.run({
  1281. program: [
  1282. 'a = await tools["__class__"]({"via": "subscript"})',
  1283. 'b = await tools.__class__({"via": "dot"})',
  1284. 'return [a, b]',
  1285. ].join('\n'),
  1286. bindings: tools({
  1287. '__class__': async () => { seen.push('called'); return 'bridged' },
  1288. }),
  1289. })
  1290. expect(result.error).toBeUndefined()
  1291. expect(result.value).toEqual(['bridged', 'bridged'])
  1292. expect(seen).toEqual(['called', 'called'])
  1293. })
  1294. it('rejects NaN binding arguments immediately instead of hanging', async () => {
  1295. // Default json.dumps would emit a non-standard NaN token that the host
  1296. // JSON.parse drops silently, hanging the call until the wall clock;
  1297. // allow_nan=False raises in-program right away.
  1298. const { runtime } = await setup({ maxWallMs: 8000 })
  1299. const start = Date.now()
  1300. const result = await runtime.run({
  1301. program: [
  1302. 'caught = ""',
  1303. 'try:',
  1304. ' await tools.echo({"x": float("nan")})',
  1305. 'except RuntimeError as e:',
  1306. ' caught = str(e)',
  1307. 'return caught',
  1308. ].join('\n'),
  1309. bindings: tools({ echo: async args => args as CodeJsonValue }),
  1310. })
  1311. expect(result.error).toBeUndefined()
  1312. expect(result.value).toContain('lossless JSON')
  1313. expect(Date.now() - start).toBeLessThan(5000)
  1314. })
  1315. it('carries large binding arguments well past maxValueBytes', async () => {
  1316. // Binding traffic has no seam byte cap: a call frame far larger than the
  1317. // completion budget must reach the host intact (the fd-3 ceiling is a
  1318. // fixed memory-safety bound, not an output budget).
  1319. const maxValueBytes = 4096
  1320. const { runtime } = await setup({ maxValueBytes })
  1321. let receivedLength = 0
  1322. const result = await runtime.run({
  1323. program: [
  1324. `big = "B" * ${maxValueBytes * 50}`,
  1325. 'r = await tools.measure({"payload": big})',
  1326. 'return r',
  1327. ].join('\n'),
  1328. bindings: tools({
  1329. measure: async (args) => {
  1330. receivedLength = ((args as { payload: string }).payload).length
  1331. return receivedLength
  1332. },
  1333. }),
  1334. })
  1335. expect(result.error).toBeUndefined()
  1336. expect(receivedLength).toBe(maxValueBytes * 50)
  1337. expect(result.value).toBe(maxValueBytes * 50)
  1338. })
  1339. it('rejects an unknown binding name inside the program with a matching error', async () => {
  1340. const { runtime } = await setup()
  1341. const result = await runtime.run({
  1342. program: [
  1343. 'caught = ""',
  1344. 'try:',
  1345. ' await tools.nope({})',
  1346. 'except (AttributeError, RuntimeError) as e:',
  1347. ' caught = str(e)',
  1348. 'return caught',
  1349. ].join('\n'),
  1350. bindings: tools({ known: async () => 'ok' }),
  1351. })
  1352. expect(result.error).toBeUndefined()
  1353. expect(result.value).toContain('nope')
  1354. })
  1355. it('bounds an unknown-binding diagnostic built from a forged call frame', async () => {
  1356. // `call.global` and `call.name` carry no byte cap of their own, only the
  1357. // 256 MiB fd-3 frame ceiling, and the reply interpolated them raw: one copy
  1358. // into the template result, one into the `JSON.stringify` escape, one into
  1359. // the `encodeJsonPlain` frame, one into the pipe write. Slicing each field
  1360. // to `maxValueBytes` code units first makes an 8 MiB forged name a
  1361. // 128-byte reply. The observable effect is the reply the child then has to
  1362. // READ: its fd-3 reader is unbuffered, so `readline` consumes an oversized
  1363. // reply one `read(2)` per byte and the run's own legitimate call never gets
  1364. // answered — measured under a 60 s ceiling, the 8 MiB case timed out and a
  1365. // 64 MiB case cost the host 509.9 MiB of heap against 120.3 MiB with the
  1366. // slices in place. The child's address space stays generous enough to BUILD
  1367. // the forgery, which is not what is under test.
  1368. const { runtime } = await setup({ maxValueBytes: 128, addressSpaceMb: 1024, maxWallMs: 20_000 })
  1369. const result = await runtime.run({
  1370. program: [
  1371. 'import os',
  1372. 'frame = b\'{"type":"call","id":9001,"global":"tools","name":"\' + b"n" * (8 * 1024 * 1024) + b\'","args":{}}\\n\'',
  1373. // One os.write returns short past the pipe buffer, and a partial frame
  1374. // would glue itself to the next one and be dropped as malformed, so the
  1375. // forgery goes out through a drain loop.
  1376. 'view = memoryview(frame)',
  1377. 'while view:',
  1378. ' view = view[os.write(3, view):]',
  1379. // A legitimate call after the forgery: its reply can only arrive once
  1380. // the child has read past whatever the forged frame was answered with.
  1381. 'await tools.known({})',
  1382. 'return "settled"',
  1383. ].join('\n'),
  1384. bindings: tools({ known: async () => 'ok' }),
  1385. })
  1386. expect(result.error).toBeUndefined()
  1387. expect(result.value).toBe('settled')
  1388. }, 40_000)
  1389. it('bridges a binding call reached via subscript access (tools["name"])', async () => {
  1390. // The SDK tells the model `await tools["my-tool"](args)` works for exotic
  1391. // names; the proxy's __getitem__ must route it through the bridge.
  1392. const { runtime } = await setup()
  1393. const result = await runtime.run({
  1394. program: [
  1395. 'r = await tools["my-tool"]({"n": 7})',
  1396. 'return r',
  1397. ].join('\n'),
  1398. bindings: tools({ 'my-tool': async args => ({ got: args as CodeJsonValue }) }),
  1399. })
  1400. expect(result.error).toBeUndefined()
  1401. expect(result.value).toEqual({ got: { n: 7 } })
  1402. })
  1403. it('raises KeyError for an undeclared subscript name', async () => {
  1404. const { runtime } = await setup()
  1405. const result = await runtime.run({
  1406. program: [
  1407. 'caught = ""',
  1408. 'try:',
  1409. ' await tools["absent"]({})',
  1410. 'except KeyError as e:',
  1411. ' caught = str(e)',
  1412. 'return caught',
  1413. ].join('\n'),
  1414. bindings: tools({ known: async () => 'ok' }),
  1415. })
  1416. expect(result.error).toBeUndefined()
  1417. expect(result.value).toContain('absent')
  1418. })
  1419. })
  1420. describe('PythonCodeRuntime — budgets, termination, disposal', () => {
  1421. it('kills a wall-clock runaway program via SIGTERM/SIGKILL and reports timeout', async () => {
  1422. const { runtime } = await setup({ maxWallMs: 500, graceMs: 200 })
  1423. const start = Date.now()
  1424. const result = await runtime.run({
  1425. program: 'import time\nwhile True: time.sleep(1)',
  1426. bindings: [],
  1427. })
  1428. const elapsed = Date.now() - start
  1429. // The wall timer may fire first or the exit-after-signal may resolve; both are ok.
  1430. expect(['timeout', 'worker-exit']).toContain(result.error?.kind)
  1431. // We got somewhere in the neighborhood of maxWallMs, not the underlying `sleep(1)`.
  1432. expect(elapsed).toBeLessThan(2000)
  1433. }, 5000)
  1434. it('aborts a run when the outer signal fires mid-flight', async () => {
  1435. const { runtime } = await setup({ maxWallMs: 10_000 })
  1436. const controller = new AbortController()
  1437. const settled: Promise<CodeRunResult> = runtime.run({
  1438. program: 'import time\nwhile True: time.sleep(0.1)',
  1439. bindings: [],
  1440. signal: controller.signal,
  1441. })
  1442. setTimeout(() => { controller.abort('outer-abort') }, 200)
  1443. const result = await settled
  1444. expect(['abort', 'worker-exit']).toContain(result.error?.kind)
  1445. }, 5000)
  1446. it('settles the run when a mid-flight abort reason cannot be converted', async () => {
  1447. // The listener converted the reason before calling `finish()`, so a hostile
  1448. // reason threw from inside an `AbortSignal` listener. Node reports that as an
  1449. // uncaught exception — it can terminate the host — and `finish()` never ran,
  1450. // so the run stayed live until the wall ceiling and misreported as `timeout`
  1451. // (observed) instead of the caller's cancellation. `maxWallMs` is short so
  1452. // that misreport is a fast assertion failure rather than a suite timeout.
  1453. const uncaught: unknown[] = []
  1454. const record = (error: unknown): void => { uncaught.push(error) }
  1455. process.on('uncaughtException', record)
  1456. try {
  1457. const { runtime } = await setup({ maxWallMs: 4_000, graceMs: 200 })
  1458. const controller = new AbortController()
  1459. const settled: Promise<CodeRunResult> = runtime.run({
  1460. program: 'import time\nwhile True: time.sleep(0.1)',
  1461. bindings: [],
  1462. signal: controller.signal,
  1463. })
  1464. setTimeout(() => {
  1465. controller.abort({ [Symbol.toPrimitive]() { throw new Error('reason blew up') } })
  1466. }, 200)
  1467. const result = await settled
  1468. expect(result.error?.kind).toBe('abort')
  1469. expect(result.error?.message).toBe('<unrenderable rejection value>')
  1470. expect(uncaught).toEqual([])
  1471. } finally {
  1472. process.off('uncaughtException', record)
  1473. }
  1474. }, 15_000)
  1475. it('disposes to quiescence: an in-flight run resolves as abort and the child exits', async () => {
  1476. const { fiber, runtime } = await setup({ maxWallMs: 10_000 })
  1477. const pending = runtime.run({
  1478. program: 'import time\nwhile True: time.sleep(0.1)',
  1479. bindings: [],
  1480. })
  1481. // Give the process time to spawn and start running.
  1482. await new Promise(resolve => setTimeout(resolve, 200))
  1483. await fiber.dispose()
  1484. const result = await pending
  1485. expect(['abort', 'worker-exit']).toContain(result.error?.kind)
  1486. }, 5000)
  1487. it('reports a spawn failure via a bogus python binary as worker-exit', async () => {
  1488. const { runtime } = await setup({ pythonBin: '/nonexistent/python-binary', maxWallMs: 3000 })
  1489. const result = await runtime.run({
  1490. program: 'return 1',
  1491. bindings: [],
  1492. })
  1493. expect(result.error?.kind).toBe('worker-exit')
  1494. }, 8000)
  1495. it('applies the strictest of the configured and inherited resource limits', async () => {
  1496. // This case used to drive the bootstrap's `applying resource limits failed`
  1497. // handler with `cpuSeconds: 2 ** 63`, asserting that a cap the child cannot
  1498. // apply fails the run rather than running it uncapped. That premise no longer
  1499. // holds, for two independent reasons, so the test now pins what is actually
  1500. // guaranteed instead of a path no admissible input reaches.
  1501. //
  1502. // First, `2 ** 63` is not a safe integer, so it is now rejected at LOAD as a
  1503. // configuration error — it can never reach the child at all. Second, even the
  1504. // largest admissible values are applied successfully, because `_clamped`
  1505. // bounds every requested pair by the inherited hard limit: an unprivileged
  1506. // process may lower a hard limit but never raise one, so the child keeps the
  1507. // stricter of the two rather than asking for something `setrlimit` refuses.
  1508. // The failure handler remains as a substrate guard (a platform whose kernel
  1509. // refuses the call for its own reasons), but it is no longer reachable from
  1510. // configuration, and a test that pretends otherwise documents a contract the
  1511. // code does not have.
  1512. //
  1513. // What is observable: a very large cap still yields a working run, and the
  1514. // containment it promises is met by the inherited ceiling.
  1515. const { runtime } = await setup({ cpuSeconds: Number.MAX_SAFE_INTEGER - 1, maxWallMs: 10_000 })
  1516. const result = await runtime.run({ program: 'return 1', bindings: [] })
  1517. expect(result.error).toBeUndefined()
  1518. expect(result.value).toBe(1)
  1519. }, 20_000)
  1520. it('settles as worker-exit when the child exits before sending done (no hang)', async () => {
  1521. // Regression: settlement must key off `close` (process reaped AND stdio
  1522. // drained), not `exit`. With `exit`, finish() re-armed a second exit
  1523. // listener that never fired — run() hung forever whenever the exit event
  1524. // beat the final fd-3 data (deterministic on macOS, a lost race elsewhere).
  1525. const { runtime } = await setup({ maxWallMs: 10_000 })
  1526. const result = await runtime.run({
  1527. program: 'import os\nos._exit(7)',
  1528. bindings: [],
  1529. })
  1530. expect(result.error?.kind).toBe('worker-exit')
  1531. expect(result.error?.message).toContain('code=7')
  1532. }, 5000)
  1533. it('classifies RLIMIT_CPU soft-limit expiry (SIGXCPU) as a timeout', async () => {
  1534. // A CPU hot loop burns the soft limit; the kernel delivers SIGXCPU, whose
  1535. // close signal the host maps to `timeout`. macOS re-delivers SIGXCPU
  1536. // differently, so we assert only kind/message here — CI's darwin leg
  1537. // validates real delivery. cpuSeconds must be an integer for setrlimit.
  1538. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 20_000 })
  1539. const result = await runtime.run({
  1540. program: 'while True: pass',
  1541. bindings: [],
  1542. })
  1543. expect(result.error?.kind).toBe('timeout')
  1544. expect(result.error?.message).toContain('CPU time exhausted')
  1545. }, 8000)
  1546. it('keeps an early self-inflicted SIGKILL a worker-exit, not a CPU timeout', async () => {
  1547. // The unsolicited-SIGKILL-as-timeout classification applies only when the
  1548. // CPU budget could have expired (wall time >= cpuSeconds). A SIGKILL
  1549. // seconds before that (cgroup OOM, an operator, os.kill) is substrate
  1550. // death and stays worker-exit per the orthogonal taxonomy.
  1551. const { runtime } = await setup({ cpuSeconds: 60, maxWallMs: 10_000 })
  1552. const result = await runtime.run({
  1553. program: [
  1554. 'import os, signal',
  1555. 'os.kill(os.getpid(), signal.SIGKILL)',
  1556. ].join('\n'),
  1557. bindings: [],
  1558. })
  1559. expect(result.error?.kind).toBe('worker-exit')
  1560. expect(result.error?.message).toContain('SIGKILL')
  1561. })
  1562. it('charges a forked descendant against the run CPU budget', async () => {
  1563. // RLIMIT_CPU is per-process and every child inherits a FRESH budget, so a
  1564. // program that shells out multiplies `cpuSeconds` by the number of
  1565. // descendants it starts. Measured before the aggregate meter existed: with
  1566. // cpuSeconds 1, two sequential busy children burned 2.0 CPU-seconds
  1567. // (RUSAGE_CHILDREN) and the run still returned a SUCCESS completion. The
  1568. // settle-time check meters RUSAGE_SELF + RUSAGE_CHILDREN and converts the
  1569. // overrun into the same SIGXCPU the untrapped soft limit sends.
  1570. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 30_000 })
  1571. const result = await runtime.run({
  1572. program: [
  1573. 'import subprocess, sys',
  1574. 'for _ in range(2):',
  1575. ' subprocess.run([sys.executable, "-c", "import time\\nt=time.time()\\nwhile time.time()-t<1.2: pass"])',
  1576. 'return "escaped the cpu budget"',
  1577. ].join('\n'),
  1578. bindings: [],
  1579. })
  1580. // Darwin's SIGXCPU re-delivery differs, so accept either terminal
  1581. // classification; what must NOT happen is the completion crossing.
  1582. expect(['timeout', 'worker-exit']).toContain(result.error?.kind)
  1583. expect(result.value).toBeUndefined()
  1584. }, 40_000)
  1585. it('does not charge wall time or a cheap descendant against the CPU budget', async () => {
  1586. // The meter is CPU, not wall clock, and it must not fire on a child that
  1587. // burns almost nothing: a sleeping program and a trivial subprocess both
  1588. // have to complete normally, or the check would reject every program that
  1589. // shells out.
  1590. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 30_000 })
  1591. const slept = await runtime.run({
  1592. program: 'import time\ntime.sleep(1.5)\nreturn "slept"',
  1593. bindings: [],
  1594. })
  1595. expect(slept.error).toBeUndefined()
  1596. expect(slept.value).toBe('slept')
  1597. const cheap = await runtime.run({
  1598. program: [
  1599. 'import subprocess, sys',
  1600. 'subprocess.run([sys.executable, "-c", "pass"])',
  1601. 'return "cheap child"',
  1602. ].join('\n'),
  1603. bindings: [],
  1604. })
  1605. expect(cheap.error).toBeUndefined()
  1606. expect(cheap.value).toBe('cheap child')
  1607. }, 40_000)
  1608. it('spends no part of addressSpaceMb on bootstrap machinery', async () => {
  1609. // RLIMIT_AS counts RESERVED address space, so anything the bootstrap maps
  1610. // for its own accounting is subtracted from the program's `addressSpaceMb`.
  1611. // A sampling thread for the descendant-CPU meter cost 72 MiB here (an 8 MiB
  1612. // stack plus a 64 MiB glibc per-thread malloc arena reservation) and turned
  1613. // the 2-million-entry dict rejection below into a MemoryError under a
  1614. // 256 MiB cap on a slower runner. Assert the child's own mappings directly
  1615. // rather than inferring the budget from a near-cap allocation, so the bound
  1616. // is read from /proc instead of from how much headroom one machine happens
  1617. // to have; 48 MiB is well above the ~30 MiB a bare interpreter maps and
  1618. // well below the 102 MiB the thread produced. `addressSpaceMb` itself is
  1619. // skipped on darwin (the dyld shared cache makes any practical cap
  1620. // unsettable) and /proc/self/maps does not exist there, so the mapping
  1621. // assertion is Linux-only; the completion path is checked everywhere.
  1622. const { runtime } = await setup({ maxValueBytes: 4096, addressSpaceMb: 256 })
  1623. const mapped = await runtime.run({
  1624. program: [
  1625. 'import sys',
  1626. 'if sys.platform != "linux":',
  1627. ' return 0',
  1628. 'total = 0',
  1629. 'with open("/proc/self/maps") as handle:',
  1630. ' for line in handle:',
  1631. ' low, high = (int(part, 16) for part in line.split(" ", 1)[0].split("-"))',
  1632. ' total += high - low',
  1633. 'return total // (1024 * 1024)',
  1634. ].join('\n'),
  1635. bindings: [],
  1636. })
  1637. expect(mapped.error).toBeUndefined()
  1638. expect(mapped.value).toBeLessThan(48)
  1639. }, 20_000)
  1640. it('spends no part of addressSpaceMb on the reply pump, across a binding await', async () => {
  1641. // The test above measures BEFORE the program yields, so it could not see the
  1642. // reply pump's cost: `loop.run_in_executor(None, read_frame)` created the
  1643. // default executor's first thread on the first `await tools.*`, and that
  1644. // thread's 8 MiB stack plus a 64 MiB glibc per-thread malloc arena are
  1645. // charged to RLIMIT_AS while the limit is already in force — measured, the
  1646. // child went from 30.34 MiB to 102.39 MiB across one binding call. Under a
  1647. // small `addressSpaceMb` the thread cannot start and a legitimate call hangs
  1648. // to `maxWallMs`; under a larger one an allocation that should have fit dies
  1649. // as MemoryError. `loop.add_reader` watches the fd with no thread at all.
  1650. //
  1651. // Measuring both sides inside one run is what discriminates: a single
  1652. // after-the-fact number cannot separate the pump's cost from the
  1653. // interpreter's own footprint. Linux-only for the same reason as above.
  1654. const { runtime } = await setup({ addressSpaceMb: 256, maxWallMs: 20_000 })
  1655. const result = await runtime.run({
  1656. program: [
  1657. 'import sys',
  1658. 'def mapped():',
  1659. ' if sys.platform != "linux":',
  1660. ' return 0',
  1661. ' total = 0',
  1662. ' with open("/proc/self/maps") as handle:',
  1663. ' for line in handle:',
  1664. ' low, high = (int(part, 16) for part in line.split(" ", 1)[0].split("-"))',
  1665. ' total += high - low',
  1666. ' return total // (1024 * 1024)',
  1667. 'before = mapped()',
  1668. 'echoed = await tools.echo({"ping": True})',
  1669. 'return {"before": before, "after": mapped(), "echoed": echoed}',
  1670. ].join('\n'),
  1671. bindings: tools({ echo: async args => args as CodeJsonValue }),
  1672. })
  1673. expect(result.error).toBeUndefined()
  1674. const value = result.value as { before: number; after: number; echoed: unknown }
  1675. // The binding call really happened, so the pump really ran.
  1676. expect(value.echoed).toEqual({ ping: true })
  1677. // Awaiting a binding maps nothing extra. The 8 MiB allowance absorbs ordinary
  1678. // heap growth while staying far below the 72 MiB a pump thread cost.
  1679. expect(value.after - value.before).toBeLessThan(8)
  1680. }, 30_000)
  1681. it('still terminates a program that ignores SIGXCPU (hard-limit backstop)', async () => {
  1682. // A hot loop under SIG_IGN burns through the soft limit; the kernel's
  1683. // hard limit (cpuSeconds + 1) SIGKILLs it. Only a kernel-authoritative
  1684. // SIGXCPU close classifies as the CPU timeout — a bare SIGKILL is
  1685. // indistinguishable from a cgroup OOM kill, so it reports worker-exit
  1686. // (Darwin re-delivers SIGXCPU instead, where the wall clock settles it
  1687. // as timeout). Either way the run TERMINATES within the budget — the
  1688. // backstop holds even when the classification is the opaque one.
  1689. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 6_000 })
  1690. const result = await runtime.run({
  1691. program: [
  1692. 'import signal',
  1693. 'signal.signal(signal.SIGXCPU, signal.SIG_IGN)',
  1694. 'while True: pass',
  1695. ].join('\n'),
  1696. bindings: [],
  1697. })
  1698. expect(['timeout', 'worker-exit']).toContain(result.error?.kind)
  1699. }, 12_000)
  1700. it('enforces the CPU budget even when the program monkeypatches the enforcement primitives', async () => {
  1701. // The check uses import-time-captured references, so replacing
  1702. // resource.getrusage / signal.signal / os.kill on the modules cannot
  1703. // defang it: a trapping program that also swaps the callables and burns
  1704. // past the budget still dies by the authoritative SIGXCPU.
  1705. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 15_000 })
  1706. const result = await runtime.run({
  1707. program: [
  1708. 'import signal, os, resource, time',
  1709. 'signal.signal(signal.SIGXCPU, lambda *a: None)',
  1710. 'resource.getrusage = lambda *a: (_ for _ in ()).throw(RuntimeError("nope"))',
  1711. 'os.kill = lambda *a: None',
  1712. 'signal.signal = lambda *a: None',
  1713. 'deadline = time.process_time() + 1.05',
  1714. 'while time.process_time() < deadline: pass',
  1715. 'return "escaped"',
  1716. ].join('\n'),
  1717. bindings: [],
  1718. })
  1719. expect(result.error?.kind).toBe('timeout')
  1720. expect(result.value).toBeUndefined()
  1721. }, 15_000)
  1722. it('re-delivers SIGXCPU when a trapping program returns inside the soft-to-hard gap', async () => {
  1723. // A program can trap SIGXCPU and settle during the one-second gap; the
  1724. // bootstrap re-checks the kernel CPU meter (getrusage) after settlement
  1725. // and dies by SIGXCPU with the default disposition restored, so the host
  1726. // still classifies the exhausted budget as a timeout instead of success.
  1727. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 15_000 })
  1728. const result = await runtime.run({
  1729. program: [
  1730. 'import signal, time',
  1731. 'fired = []',
  1732. 'signal.signal(signal.SIGXCPU, lambda *a: fired.append(1))',
  1733. 'deadline = time.process_time() + 1.05',
  1734. 'while time.process_time() < deadline: pass',
  1735. 'return "escaped"',
  1736. ].join('\n'),
  1737. bindings: [],
  1738. })
  1739. expect(result.error?.kind).toBe('timeout')
  1740. expect(result.error?.message).toContain('CPU time exhausted')
  1741. expect(result.value).toBeUndefined()
  1742. }, 15_000)
  1743. it('enforces the CPU budget when the program rebinds the enforcer on __main__', async () => {
  1744. // The bootstrap IS `__main__`, so `import __main__` reaches its globals.
  1745. // The enforcement callable holds its primitives in closure cells (not
  1746. // module attributes) and `_run` reads the callable into a frame local
  1747. // before the program starts, so neither replacing the global nor swapping
  1748. // the module's captured names changes what runs after settlement.
  1749. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 15_000 })
  1750. const result = await runtime.run({
  1751. program: [
  1752. 'import signal, time, __main__',
  1753. 'signal.signal(signal.SIGXCPU, lambda *a: None)',
  1754. '__main__._DIE_IF_CPU_EXHAUSTED = lambda *_: None',
  1755. 'deadline = time.process_time() + 1.05',
  1756. 'while time.process_time() < deadline: pass',
  1757. 'return "escaped"',
  1758. ].join('\n'),
  1759. bindings: [],
  1760. })
  1761. expect(result.error?.kind).toBe('timeout')
  1762. expect(result.value).toBeUndefined()
  1763. }, 15_000)
  1764. it('bounds a program that defeats the post-check by writing its closure cell', async () => {
  1765. // The closure-cell capture raises the cost of defeating the post-check; it
  1766. // does NOT make it unreachable, and nothing in-process could: a cell is
  1767. // writable through `fn.__closure__[i].cell_contents`, and `sys._getframe`
  1768. // reads _run's frame locals. This program does exactly that — walks to
  1769. // _run's frame, takes the enforcement callable, and replaces its captured
  1770. // `getrusage` with one reporting zero CPU used — then burns past cpuSeconds
  1771. // with SIGXCPU trapped. The run must still fail, because the bound that
  1772. // model code cannot forge is outside the interpreter: the RLIMIT_CPU HARD
  1773. // limit at cpuSeconds + 1, whose SIGKILL admits no handler. No success is
  1774. // reportable either way.
  1775. const { runtime } = await setup({ cpuSeconds: 1, maxWallMs: 20_000 })
  1776. const start = Date.now()
  1777. const result = await runtime.run({
  1778. program: [
  1779. 'import signal, sys, time',
  1780. 'signal.signal(signal.SIGXCPU, lambda *a: None)',
  1781. // Walk out of __dsh_main__ to _run's frame and take its local.
  1782. 'die = None',
  1783. 'depth = 1',
  1784. 'while depth < 12:',
  1785. ' frame = sys._getframe(depth)',
  1786. ' if "die_if_cpu_exhausted" in frame.f_locals:',
  1787. ' die = frame.f_locals["die_if_cpu_exhausted"]',
  1788. ' break',
  1789. ' depth += 1',
  1790. 'assert die is not None, "enforcer not reachable from the frame chain"',
  1791. 'class Zero:',
  1792. ' ru_utime = 0.0',
  1793. ' ru_stime = 0.0',
  1794. 'names = die.__code__.co_freevars',
  1795. 'die.__closure__[names.index("getrusage")].cell_contents = lambda *a: Zero()',
  1796. // Burn well past the soft limit into the hard limit's SIGKILL.
  1797. 'while True: pass',
  1798. ].join('\n'),
  1799. bindings: [],
  1800. })
  1801. // What holds on EVERY platform: the tampering bought no success. The run
  1802. // failed, carried no value, and the reported kind is one of the two
  1803. // kernel-level outcomes — never a completion.
  1804. expect(result.value).toBeUndefined()
  1805. expect(result.error?.kind === 'worker-exit' || result.error?.kind === 'timeout').toBe(true)
  1806. if (process.platform === 'linux') {
  1807. // Linux enforces the RLIMIT_CPU HARD limit at cpuSeconds + 1 promptly, so
  1808. // the CPU bound — not the 20 s wall ceiling — is what stops the program.
  1809. // Its SIGKILL is not SIGXCPU, so the orthogonal-failure taxonomy reports
  1810. // `worker-exit`: a bare SIGKILL is not evidence of CPU burn.
  1811. expect(result.error?.kind).toBe('worker-exit')
  1812. expect(Date.now() - start).toBeLessThan(15_000)
  1813. } else {
  1814. // Darwin does not deliver the hard limit's SIGKILL on the same schedule;
  1815. // observed on the macOS lane, a program that patches the post-check runs
  1816. // to the WALL ceiling instead. The CPU budget is therefore not the
  1817. // binding constraint against a tampering program there — the wall clock
  1818. // is. Asserted rather than skipped so the difference stays visible.
  1819. expect(result.error?.kind).toBe('timeout')
  1820. }
  1821. }, 30_000)
  1822. it('keeps a finished-but-not-closed run live so dispose awaits the child\'s death', async () => {
  1823. // finish() no longer drops the run from `live`; settle() (at close) does.
  1824. // A SIGTERM-trapping program with a small graceMs sits in the grace window
  1825. // after finish() fires — dispose() must not resolve until the SIGKILL
  1826. // backstop actually reaps the child. The program prints its pid (captured
  1827. // as a log even on abort); once dispose() resolves, that pid must be dead
  1828. // (process.kill(pid, 0) throws ESRCH).
  1829. const { fiber, runtime } = await setup({ maxWallMs: 10_000, graceMs: 400 })
  1830. // Deterministic readiness: the program reports its pid through a binding
  1831. // AFTER installing the trap, so dispose cannot race the spawn (a fixed
  1832. // sleep lost that race on slow CI runners — SIGTERM landed pre-trap).
  1833. let reportedPid!: (pid: number) => void
  1834. const trapReady = new Promise<number>((resolve) => { reportedPid = resolve })
  1835. const pending = runtime.run({
  1836. program: [
  1837. 'import signal, time, os',
  1838. 'signal.signal(signal.SIGTERM, lambda *a: None)',
  1839. 'await tools.ready({"pid": os.getpid()})',
  1840. 'while True: time.sleep(0.05)',
  1841. ].join('\n'),
  1842. bindings: tools({
  1843. ready: async (args) => {
  1844. reportedPid((args as { pid: number }).pid)
  1845. return 'ok'
  1846. },
  1847. }),
  1848. })
  1849. const pid = await trapReady
  1850. const start = Date.now()
  1851. await fiber.dispose()
  1852. const elapsed = Date.now() - start
  1853. const result = await pending
  1854. expect(['abort', 'worker-exit', 'timeout']).toContain(result.error?.kind)
  1855. // dispose() returned only after the grace window elapsed (the SIGTERM trap
  1856. // forces the SIGKILL backstop path), proving the run stayed live past finish().
  1857. expect(elapsed).toBeGreaterThanOrEqual(300)
  1858. expect(Number.isInteger(pid) && pid > 0).toBe(true)
  1859. // The child is fully reaped by the time dispose() resolved.
  1860. expect(() => process.kill(pid, 0)).toThrow(/ESRCH/)
  1861. }, 8000)
  1862. it('settles on the decided result even when a setsid-escaped orphan holds stdio open past close', async () => {
  1863. // `close` only fires once every inherited stdio stream drains. A descendant
  1864. // started with start_new_session=True escapes the child's process group, so
  1865. // the SIGTERM/SIGKILL aimed at that group never reaches it; if it inherited
  1866. // our stdout/stderr/fd 3 and outlives the run, `close` would never fire and
  1867. // run() would hang forever. The close-deadline backstop (graceMs + margin)
  1868. // must force settlement on the value the `done` frame already decided.
  1869. const { runtime } = await setup({ graceMs: 100 })
  1870. const start = Date.now()
  1871. const result = await runtime.run({
  1872. program: [
  1873. 'import subprocess, sys',
  1874. // Orphan in a fresh session, inheriting our stdout/stderr/fd 3, alive
  1875. // well past the close-deadline so `close` cannot fire on its own.
  1876. 'subprocess.Popen([sys.executable, "-c", "import time; time.sleep(10)"],',
  1877. ' start_new_session=True)',
  1878. 'return "escaped"',
  1879. ].join('\n'),
  1880. bindings: [],
  1881. })
  1882. const elapsed = Date.now() - start
  1883. // The done frame decided the value; the deadline settled it despite the
  1884. // orphan pinning the pipes open.
  1885. expect(result.error).toBeUndefined()
  1886. expect(result.value).toBe('escaped')
  1887. // Settlement waited for the backstop (graceMs + CLOSE_REAP_MARGIN_MS ≈ 2.1s),
  1888. // not the wall-clock ceiling — proving the deadline, not the ceiling, fired.
  1889. expect(elapsed).toBeGreaterThanOrEqual(1_500)
  1890. expect(elapsed).toBeLessThan(5_000)
  1891. }, 8000)
  1892. it('reaps a same-group child that ignores SIGTERM and releases the pipes before close', async () => {
  1893. // The same-group counterpart to the setsid-orphan case above. A descendant
  1894. // left in the child's OWN process group (no setsid, so `kill(-pid)` reaches
  1895. // it) can ignore SIGTERM yet still release the inherited stdout/stderr/fd 3
  1896. // it does not hold — here by giving the Popen child DEVNULL streams and
  1897. // letting close_fds drop fd 3. The leader then writes `done` and exits, its
  1898. // `close` fires because the pipes drained, and settle() runs while that
  1899. // descendant is still alive. settle() then keeps a REF'd poll alive until the
  1900. // grace-window SIGKILL has emptied the whole process group, so the host cannot
  1901. // exit and reparent the survivor to init: no subprocess outlives the fiber.
  1902. //
  1903. // The descendant must have SIG_IGN installed BEFORE the host sends SIGTERM,
  1904. // or it dies from the default SIGTERM whether the fix is present or not — so
  1905. // it writes a readiness marker after trapping and the leader waits for that
  1906. // marker before returning. While alive it bumps a heartbeat file every 50 ms;
  1907. // the test asserts the heartbeat STOPS, which is what "no longer executing"
  1908. // means whether the killed descendant is reaped or lingers as a zombie (a
  1909. // SIGKILL'd process runs no more code either way). It sleeps 30 s as a safety
  1910. // net so a broken fix cannot leak it forever.
  1911. const handoff = await mkdtemp(join(tmpdir(), 'dsh-samegroup-'))
  1912. const readyMarker = join(handoff, 'ready')
  1913. const heartbeat = join(handoff, 'heartbeat')
  1914. const { runtime } = await setup({ maxWallMs: 10_000, graceMs: 300 })
  1915. const result = await runtime.run({
  1916. program: [
  1917. 'import subprocess, sys, os, time',
  1918. `marker = ${JSON.stringify(readyMarker)}`,
  1919. `heartbeat = ${JSON.stringify(heartbeat)}`,
  1920. // Same group (no start_new_session); ignores SIGTERM; holds none of the
  1921. // leader's pipes (DEVNULL std streams, close_fds drops fd 3). It writes
  1922. // the marker (argv[1]) only AFTER the trap is installed — so the leader
  1923. // cannot return, and the host cannot send SIGTERM, before it is ignored —
  1924. // then rewrites the heartbeat (argv[2]) every 50 ms for up to 30 s.
  1925. 'code = ("import signal, sys, time\\n"',
  1926. ' "signal.signal(signal.SIGTERM, signal.SIG_IGN)\\n"',
  1927. ' "open(sys.argv[1], \'w\').close()\\n"',
  1928. ' "end = time.time() + 30\\n"',
  1929. ' "while time.time() < end:\\n"',
  1930. ' " open(sys.argv[2], \'w\').close()\\n"',
  1931. ' " time.sleep(0.05)\\n")',
  1932. 'child = subprocess.Popen([sys.executable, "-c", code, marker, heartbeat],',
  1933. ' stdin=subprocess.DEVNULL,',
  1934. ' stdout=subprocess.DEVNULL,',
  1935. ' stderr=subprocess.DEVNULL)',
  1936. 'deadline = time.time() + 5',
  1937. 'while not os.path.exists(marker) and time.time() < deadline:',
  1938. ' time.sleep(0.02)',
  1939. 'return "spawned"',
  1940. ].join('\n'),
  1941. bindings: [],
  1942. })
  1943. expect(result.error).toBeUndefined()
  1944. expect(result.value).toBe('spawned')
  1945. // The trap really installed before the leader returned, so this is the
  1946. // SIGTERM-ignoring descendant, not one that would have died to the default.
  1947. expect(existsSync(readyMarker)).toBe(true)
  1948. // The grace-window SIGKILL (graceMs 300 + reap margin) empties the group. Once
  1949. // it has, the descendant stops bumping the heartbeat. Poll the heartbeat's
  1950. // mtime: two consecutive reads far enough apart with no change means it is no
  1951. // longer executing — true whether it was reaped or lingers as a zombie, so
  1952. // the assertion holds in a container whose init does not wait() orphans. The
  1953. // window (well under the 30 s self-timeout) proves the SIGKILL did the work.
  1954. const mtime = (): number => { try { return statSync(heartbeat).mtimeMs } catch { return 0 } }
  1955. const stopDeadline = Date.now() + 8_000
  1956. let last = mtime()
  1957. let still = false
  1958. while (Date.now() < stopDeadline) {
  1959. await new Promise(resolve => setTimeout(resolve, 400))
  1960. const now = mtime()
  1961. if (now === last && now !== 0) { still = true; break }
  1962. last = now
  1963. }
  1964. expect(still).toBe(true)
  1965. }, 20_000)
  1966. it('dispose awaits reaping of a same-group survivor from a completed run', async () => {
  1967. // The quiescence contract also holds for a run that ALREADY resolved: the run
  1968. // stays tracked in `live` until its process group is reaped, so a `dispose()`
  1969. // that races a just-returned run() still awaits the survivor rather than
  1970. // snapshotting an empty `live` and returning while it lives. Here the run
  1971. // completes (leaving a SIGTERM-ignoring same-group descendant), then dispose()
  1972. // is called; the heartbeat must be stale BY THE TIME dispose() resolves —
  1973. // proving teardown waited for the reap, not merely that the reap eventually
  1974. // happened.
  1975. const handoff = await mkdtemp(join(tmpdir(), 'dsh-dispose-quiesce-'))
  1976. const readyMarker = join(handoff, 'ready')
  1977. const heartbeat = join(handoff, 'heartbeat')
  1978. const { runtime, fiber } = await setup({ maxWallMs: 10_000, graceMs: 300 })
  1979. const result = await runtime.run({
  1980. program: [
  1981. 'import subprocess, sys, os, time',
  1982. `marker = ${JSON.stringify(readyMarker)}`,
  1983. `heartbeat = ${JSON.stringify(heartbeat)}`,
  1984. 'code = ("import signal, sys, time\\n"',
  1985. ' "signal.signal(signal.SIGTERM, signal.SIG_IGN)\\n"',
  1986. ' "open(sys.argv[1], \'w\').close()\\n"',
  1987. ' "end = time.time() + 30\\n"',
  1988. ' "while time.time() < end:\\n"',
  1989. ' " open(sys.argv[2], \'w\').close()\\n"',
  1990. ' " time.sleep(0.05)\\n")',
  1991. 'child = subprocess.Popen([sys.executable, "-c", code, marker, heartbeat],',
  1992. ' stdin=subprocess.DEVNULL,',
  1993. ' stdout=subprocess.DEVNULL,',
  1994. ' stderr=subprocess.DEVNULL)',
  1995. 'deadline = time.time() + 5',
  1996. 'while not os.path.exists(marker) and time.time() < deadline:',
  1997. ' time.sleep(0.02)',
  1998. 'return "spawned"',
  1999. ].join('\n'),
  2000. bindings: [],
  2001. })
  2002. expect(result.error).toBeUndefined()
  2003. expect(existsSync(readyMarker)).toBe(true)
  2004. // dispose() must not return until the group is reaped. After it resolves, the
  2005. // heartbeat must already be stale: read its mtime, wait past the heartbeat
  2006. // interval, and confirm it did not advance — the descendant is no longer
  2007. // executing (reaped or zombie), so teardown was genuinely quiescent.
  2008. await fiber.dispose()
  2009. const mtime = (): number => { try { return statSync(heartbeat).mtimeMs } catch { return 0 } }
  2010. const afterDispose = mtime()
  2011. await new Promise(resolve => setTimeout(resolve, 500))
  2012. expect(mtime()).toBe(afterDispose)
  2013. }, 20_000)
  2014. it('sends SIGKILL at the poll deadline when the event loop was blocked past both timers', async () => {
  2015. // If the host event loop is blocked (a big synchronous computation) from
  2016. // before the group-reap poll was scheduled until after the deadline, both the
  2017. // poll timer and the grace-window SIGKILL timer are overdue when the loop
  2018. // resumes. Node runs the earlier-scheduled poll first, so the SIGKILL timer
  2019. // may not have fired yet. The deadline arm must then send SIGKILL ITSELF
  2020. // rather than cancel the unfired escalation — otherwise a SIGTERM-ignoring
  2021. // same-group survivor is released for good. A synchronous busy-loop after
  2022. // run() resolves reproduces the block deterministically.
  2023. const handoff = await mkdtemp(join(tmpdir(), 'dsh-deadline-'))
  2024. const readyMarker = join(handoff, 'ready')
  2025. const heartbeat = join(handoff, 'heartbeat')
  2026. const graceMs = 300
  2027. const { runtime } = await setup({ maxWallMs: 10_000, graceMs })
  2028. const result = await runtime.run({
  2029. program: [
  2030. 'import subprocess, sys, os, time',
  2031. `marker = ${JSON.stringify(readyMarker)}`,
  2032. `heartbeat = ${JSON.stringify(heartbeat)}`,
  2033. 'code = ("import signal, sys, time\\n"',
  2034. ' "signal.signal(signal.SIGTERM, signal.SIG_IGN)\\n"',
  2035. ' "open(sys.argv[1], \'w\').close()\\n"',
  2036. ' "end = time.time() + 30\\n"',
  2037. ' "while time.time() < end:\\n"',
  2038. ' " open(sys.argv[2], \'w\').close()\\n"',
  2039. ' " time.sleep(0.05)\\n")',
  2040. 'child = subprocess.Popen([sys.executable, "-c", code, marker, heartbeat],',
  2041. ' stdin=subprocess.DEVNULL,',
  2042. ' stdout=subprocess.DEVNULL,',
  2043. ' stderr=subprocess.DEVNULL)',
  2044. 'deadline = time.time() + 5',
  2045. 'while not os.path.exists(marker) and time.time() < deadline:',
  2046. ' time.sleep(0.02)',
  2047. 'return "spawned"',
  2048. ].join('\n'),
  2049. bindings: [],
  2050. })
  2051. expect(result.error).toBeUndefined()
  2052. expect(existsSync(readyMarker)).toBe(true)
  2053. // Block the event loop synchronously past graceMs + CLOSE_REAP_MARGIN_MS
  2054. // (2000) with margin, so both timers are overdue when the loop resumes.
  2055. const blockUntil = Date.now() + graceMs + 2_000 + 800
  2056. while (Date.now() < blockUntil) { /* busy-wait, no yield */ }
  2057. // Yield: the overdue poll runs (group still non-empty, deadline passed) and
  2058. // must send SIGKILL itself. The survivor then stops bumping the heartbeat.
  2059. const mtime = (): number => { try { return statSync(heartbeat).mtimeMs } catch { return 0 } }
  2060. const stopDeadline = Date.now() + 5_000
  2061. let last = mtime()
  2062. let stopped = false
  2063. while (Date.now() < stopDeadline) {
  2064. await new Promise(resolve => setTimeout(resolve, 400))
  2065. const now = mtime()
  2066. if (now === last && now !== 0) { stopped = true; break }
  2067. last = now
  2068. }
  2069. expect(stopped).toBe(true)
  2070. }, 20_000)
  2071. })
  2072. describe('PythonCodeRuntime — hostile peer', () => {
  2073. it('drops garbage bytes and unknown-shape frames posted directly to fd 3', async () => {
  2074. // The model program can reach fd 3 and write anything. We inject a
  2075. // non-JSON line, a valid JSON but unknown-shape frame, and a broken done
  2076. // frame; the host must not crash, and the real `done` still settles the run.
  2077. const { runtime } = await setup()
  2078. const result = await runtime.run({
  2079. program: [
  2080. 'import os',
  2081. 'os.write(3, b"not-json\\n")',
  2082. 'os.write(3, b\'{"type":"unknown"}\\n\')',
  2083. 'os.write(3, b\'{"type":"done","error":{"message":42}}\\n\')',
  2084. 'return "survived"',
  2085. ].join('\n'),
  2086. bindings: [],
  2087. })
  2088. expect(result.error).toBeUndefined()
  2089. expect(result.value).toBe('survived')
  2090. })
  2091. it('answers a forged call frame for an unknown binding and never crashes', async () => {
  2092. // The unknown-binding reply path, driven through the id the host expects:
  2093. // the program lets its own first call claim id 0 and forges id 1, which the
  2094. // host answers with the `unknown binding` rejection the honest call would
  2095. // have received. A forged id out of sequence is dropped instead — that is
  2096. // the id-bound test below, not this one.
  2097. const { runtime } = await setup({ maxWallMs: 8_000 })
  2098. let seenLegitCall = false
  2099. const result = await runtime.run({
  2100. program: [
  2101. 'import os, json',
  2102. 'x = await tools.echo({"ping": True})',
  2103. 'os.write(3, json.dumps({"type":"call","id":1,"global":"tools","name":"forged","args":{}}).encode() + b"\\n")',
  2104. // The forged frame is answered, but nothing in the child awaits id 1, so
  2105. // the reply is ignored and the run completes on its own value.
  2106. 'return x',
  2107. ].join('\n'),
  2108. bindings: tools({
  2109. echo: async (args) => { seenLegitCall = true; return args as CodeJsonValue },
  2110. }),
  2111. })
  2112. expect(result.error).toBeUndefined()
  2113. expect(result.value).toEqual({ ping: true })
  2114. expect(seenLegitCall).toBe(true)
  2115. }, 15_000)
  2116. it('drops forged call frames whose ids are not the next in sequence, retaining no per-id state', async () => {
  2117. // The host used to remember every answered id in a Set, so a program could
  2118. // write an unbounded run of unique forged ids — each frame far below the
  2119. // 256 MiB ceiling, so nothing rejected them — and grow host memory for the
  2120. // whole run. Ids are consecutive from 0, so one counter replaces the set.
  2121. //
  2122. // The discriminator is that the forgeries must not be answered. Each names a
  2123. // binding that does exist, so a host answering them would run `echo` once
  2124. // per forgery; the count proves only the legitimate call was dispatched.
  2125. // Ids also run DESCENDING, so a high-water-mark test would drop the honest
  2126. // call that follows rather than the forgeries.
  2127. const { runtime } = await setup()
  2128. let echoCalls = 0
  2129. const result = await runtime.run({
  2130. program: [
  2131. 'import os, json',
  2132. 'for i in range(2000, 0, -1):',
  2133. ' os.write(3, json.dumps({"type":"call","id":i,"global":"tools","name":"echo","args":{"forged":i}}).encode() + b"\\n")',
  2134. 'x = await tools.echo({"ping": True})',
  2135. 'return x',
  2136. ].join('\n'),
  2137. bindings: tools({
  2138. echo: async (args) => { echoCalls += 1; return args as CodeJsonValue },
  2139. }),
  2140. })
  2141. expect(result.error).toBeUndefined()
  2142. expect(result.value).toEqual({ ping: true })
  2143. expect(echoCalls).toBe(1)
  2144. }, 15_000)
  2145. it('keeps answering calls a program makes after one with unserializable arguments', async () => {
  2146. // The child claims an id only once its write succeeds, so a call rejected
  2147. // child-side for non-lossless arguments leaves no gap. Were a gap possible,
  2148. // the host's exact-successor test would drop every later call and the run
  2149. // would hang to the wall ceiling instead of completing.
  2150. const { runtime } = await setup({ maxWallMs: 8_000 })
  2151. const seen: unknown[] = []
  2152. const result = await runtime.run({
  2153. program: [
  2154. 'caught = ""',
  2155. 'try:',
  2156. ' await tools.echo({"bad": float("inf")})',
  2157. 'except RuntimeError as e:',
  2158. ' caught = str(e)',
  2159. 'after = await tools.echo({"ok": True})',
  2160. 'return {"caught": caught, "after": after}',
  2161. ].join('\n'),
  2162. bindings: tools({
  2163. echo: async (args) => { seen.push(args); return args as CodeJsonValue },
  2164. }),
  2165. })
  2166. expect(result.error).toBeUndefined()
  2167. const value = result.value as { caught: string; after: unknown }
  2168. expect(value.caught).toContain('lossless JSON')
  2169. expect(value.after).toEqual({ ok: true })
  2170. // The rejected call never reached the host; the one after it did.
  2171. expect(seen).toEqual([{ ok: true }])
  2172. }, 15_000)
  2173. it('drops a forged frame carrying an integer outside JavaScript safe range', async () => {
  2174. // JSON.parse would silently round 9007199254740993 to ...992 BEFORE any
  2175. // validation, corrupting a dispatched argument or completion. The host
  2176. // scans the raw line and drops such frames as hostile traffic; the honest
  2177. // child cannot produce one (its validator rejects unsafe ints).
  2178. const { runtime } = await setup()
  2179. let dispatched: unknown
  2180. const result = await runtime.run({
  2181. program: [
  2182. 'import os',
  2183. // Forged call frame with an unsafe int argument, then a forged done
  2184. // frame with an unsafe int value — both must be dropped whole.
  2185. 'os.write(3, b\'{"type":"call","id":7,"global":"tools","name":"echo","args":9007199254740993}\\n\')',
  2186. 'os.write(3, b\'{"type":"done","value":9007199254740993}\\n\')',
  2187. 'x = await tools.echo({"ok": True})',
  2188. 'return x',
  2189. ].join('\n'),
  2190. bindings: tools({
  2191. echo: async (args) => { dispatched = args; return args as CodeJsonValue },
  2192. }),
  2193. })
  2194. expect(result.error).toBeUndefined()
  2195. // The forged done did not settle the run; the legit call and completion did.
  2196. expect(result.value).toEqual({ ok: true })
  2197. expect(dispatched).toEqual({ ok: true })
  2198. })
  2199. it('truncates host-side logs once the budget is exhausted and emits the marker', async () => {
  2200. // Set a tiny host-side budget; the Python side has a much larger one, so
  2201. // its LogBuffer will not truncate — the host ledger fires first.
  2202. const { runtime } = await setup({ maxLogBytes: 32 })
  2203. const result = await runtime.run({
  2204. program: [
  2205. 'for _ in range(50):',
  2206. ' print("aaaaaaaaaa")',
  2207. 'return "done"',
  2208. ].join('\n'),
  2209. bindings: [],
  2210. })
  2211. expect(result.error).toBeUndefined()
  2212. const markers = result.logs.filter(line => line.includes('log capture truncated at 32 bytes'))
  2213. expect(markers.length).toBeGreaterThanOrEqual(1)
  2214. })
  2215. it('reports an exception whose message holds an unpaired surrogate instead of stranding to the wall clock', async () => {
  2216. // A strict UTF-8 encode of "\ud800" throws while BUILDING the failure
  2217. // frame; the run would then hang to maxWallMs and misreport as timeout.
  2218. const { runtime } = await setup({ maxWallMs: 8_000 })
  2219. const result = await runtime.run({
  2220. program: String.raw`raise Exception("bad \ud800 surrogate")`,
  2221. bindings: [],
  2222. })
  2223. expect(result.error?.kind).toBe('exception')
  2224. expect(result.error?.message).toContain('bad')
  2225. expect(result.error?.message).toContain('surrogate')
  2226. })
  2227. it('carries a lone-surrogate completion string across the wire as its JSON escape', async () => {
  2228. // UTF-8 has no encoding for a lone surrogate, but JSON does: the ASCII
  2229. // `\ud800` escape, which JSON.parse reads back as the same UTF-16 code
  2230. // unit. `CodeJsonValue`, `snapshotJsonValue`, and the worker backend all
  2231. // accept such a string, so this backend must not narrow the shared seam.
  2232. const { runtime } = await setup()
  2233. const result = await runtime.run({
  2234. program: String.raw`return {"lone": "a\ud800b", "spelled": "😀"}`,
  2235. bindings: [],
  2236. })
  2237. expect(result.error).toBeUndefined()
  2238. // The lone half survives as the code unit itself; a spelled-out high-low
  2239. // PAIR folds into the astral character the host would hold for it.
  2240. expect(result.value).toEqual({ lone: 'a\ud800b', spelled: '\u{1f600}' })
  2241. })
  2242. it('meters a lone surrogate at its six escaped bytes, matching the host', async () => {
  2243. // The child and the host share maxValueBytes, so the child must charge the
  2244. // escape's six ASCII bytes (plus two quotes): eight fits, nine does not.
  2245. const { runtime } = await setup({ maxValueBytes: 8 })
  2246. const ok = await runtime.run({ program: String.raw`return "\ud800"`, bindings: [] })
  2247. expect(ok.error).toBeUndefined()
  2248. expect(ok.value).toBe('\ud800')
  2249. const over = await setup({ maxValueBytes: 7 })
  2250. const result = await over.runtime.run({ program: String.raw`return "\ud800"`, bindings: [] })
  2251. expect(result.error?.kind).toBe('output-limit')
  2252. })
  2253. it('passes a lone-surrogate binding argument through instead of failing the call', async () => {
  2254. // The argument validator shared the same over-narrow rejection; a host
  2255. // binding must receive the code unit the program passed.
  2256. const seen: unknown[] = []
  2257. const { runtime } = await setup()
  2258. const result = await runtime.run({
  2259. program: String.raw`return await tools.echo({"text": "x\udfff"})`,
  2260. bindings: tools({ echo: async (args: unknown) => { seen.push(args); return args as CodeJsonValue } }),
  2261. })
  2262. expect(result.error).toBeUndefined()
  2263. expect(seen).toEqual([{ text: 'x\udfff' }])
  2264. expect(result.value).toEqual({ text: 'x\udfff' })
  2265. })
  2266. it('meters a non-ASCII completion in UTF-8 JSON bytes, matching the host', async () => {
  2267. // json.dumps' default \uXXXX escaping would count "é" as 8 bytes while
  2268. // the host meter counts its UTF-8 JSON form (4); the shared budget must
  2269. // agree, so a 4-byte-fitting value passes a maxValueBytes of 4.
  2270. const { runtime } = await setup({ maxValueBytes: 4 })
  2271. const ok = await runtime.run({ program: 'return "é"', bindings: [] })
  2272. expect(ok.error).toBeUndefined()
  2273. expect(ok.value).toBe('é')
  2274. const over = await runtime.run({ program: 'return "éx"', bindings: [] })
  2275. expect(over.error?.kind).toBe('output-limit')
  2276. })
  2277. it('filters bootstrap frames from exception-group members (TaskGroup)', async () => {
  2278. // Python 3.11+ stores member stacks under TracebackException.exceptions;
  2279. // the <model>-frame filter must recurse into them too.
  2280. const { runtime } = await setup()
  2281. const result = await runtime.run({
  2282. program: [
  2283. 'import asyncio, sys',
  2284. 'if sys.version_info < (3, 11):',
  2285. ' raise ValueError("skip-old <model>")',
  2286. 'async def boom():',
  2287. ' raise ValueError("group-member")',
  2288. 'async with asyncio.TaskGroup() as tg:',
  2289. ' tg.create_task(boom())',
  2290. ].join('\n'),
  2291. bindings: [],
  2292. })
  2293. expect(result.error?.kind).toBe('exception')
  2294. expect(result.error?.message).toContain('<model>')
  2295. expect(result.error?.message).not.toContain('bootstrap.py')
  2296. })
  2297. it('keeps frames intact when a model thread floods logs while a large frame drains', async () => {
  2298. // os.write releases the GIL and a frame beyond PIPE_BUF is not atomic:
  2299. // without the writer lock + full-write loop, the printing thread could
  2300. // interleave bytes mid-frame and the host would drop the malformed JSON,
  2301. // hanging the run to the wall clock (or losing the completion).
  2302. const { runtime } = await setup({ maxValueBytes: 1024 * 1024, maxLogBytes: 4 * 1024 * 1024, maxWallMs: 15_000 })
  2303. const result = await runtime.run({
  2304. program: [
  2305. 'import threading',
  2306. 'stop = False',
  2307. 'def spam():',
  2308. ' while not stop:',
  2309. ' print("spam-line-" + "y" * 100)',
  2310. 't = threading.Thread(target=spam)',
  2311. 't.start()',
  2312. // A ~300 KiB completion — several PIPE_BUF units — while spam runs.
  2313. 'big = "x" * (300 * 1024)',
  2314. 'stop = True',
  2315. 't.join()',
  2316. 'return big',
  2317. ].join('\n'),
  2318. bindings: [],
  2319. })
  2320. expect(result.error).toBeUndefined()
  2321. expect(result.value).toBe('x'.repeat(300 * 1024))
  2322. }, 20_000)
  2323. it('settles cleanly while a daemon thread keeps writing unterminated log text', async () => {
  2324. // WARNING regression: the settlement `flush_out()/flush_err()` on the main
  2325. // coroutine read and clear `_LogStream._pending` and the shared LogBuffer
  2326. // ledger with NO lock, while a model daemon thread's `print`/`write` mutate
  2327. // the same state. Capturing the bound method (`out_stream.flush_line`) only
  2328. // fixes WHICH callable runs, not what it reads mid-flight: the flush could
  2329. // interleave with a concurrent write and join a `_pending` list being
  2330. // mutated under it, corrupting the ledger and costing the `done` frame — the
  2331. // run would then strand to the wall clock instead of completing. The shared
  2332. // re-entrant lock serializes them.
  2333. //
  2334. // A pure data race has no single bad input to reject deterministically, so
  2335. // this maximizes overlap: daemon threads emit UNTERMINATED writes (which
  2336. // pile into `_pending` rather than flushing per line) right up to the moment
  2337. // the body returns and settlement flushes. Repeated so the interleave lands.
  2338. for (let attempt = 0; attempt < 5; attempt++) {
  2339. const { runtime, fiber } = await setup({ maxLogBytes: 4 * 1024 * 1024, maxWallMs: 15_000 })
  2340. const result = await runtime.run({
  2341. program: [
  2342. 'import sys, threading',
  2343. 'stop = False',
  2344. 'def spam():',
  2345. ' while not stop:',
  2346. // No newline: the text accumulates in the stream's `_pending`, which is
  2347. // exactly the state the settlement flush also touches.
  2348. ' sys.stdout.write("tail-fragment-" + "z" * 64)',
  2349. 'workers = [threading.Thread(target=spam, daemon=True) for _ in range(4)]',
  2350. 'for t in workers: t.start()',
  2351. // Let the daemons build up pending writes, then return so settlement
  2352. // flushes while they are still mid-write.
  2353. 'import time; time.sleep(0.05)',
  2354. 'return "settled"',
  2355. ].join('\n'),
  2356. bindings: [],
  2357. })
  2358. expect(result.error).toBeUndefined()
  2359. expect(result.value).toBe('settled')
  2360. await fiber.dispose()
  2361. }
  2362. }, 30_000)
  2363. it('completes a binding called from a worker thread on its own event loop', async () => {
  2364. // A binding reply Future is created on the loop that ran `dispatch`. When the
  2365. // model calls a binding from a worker THREAD via `asyncio.run(tools.x(...))`,
  2366. // that Future belongs to the thread's loop, not the main loop where
  2367. // `_pump_replies` reads the reply. `asyncio.Future` is not thread-safe:
  2368. // completing it from another thread does not wake its own loop, so a direct
  2369. // `set_result` would strand the awaiting thread and the run would degrade to a
  2370. // wall-clock timeout. The pump must schedule completion on the Future's own
  2371. // loop via `call_soon_threadsafe`. The tight maxWallMs makes the pre-fix
  2372. // failure a fast timeout rather than a hang.
  2373. //
  2374. // The main coroutine yields with `await asyncio.sleep` while the worker runs,
  2375. // rather than a synchronous `t.join()`: joining would block the main thread,
  2376. // so the main loop could not run `_pump_replies` and the call would deadlock
  2377. // regardless of the fix — that blocks the pump, not the cross-loop delivery
  2378. // this test pins.
  2379. const { runtime } = await setup({ maxWallMs: 8_000 })
  2380. const seen: unknown[] = []
  2381. const result = await runtime.run({
  2382. program: [
  2383. 'import asyncio, threading',
  2384. 'result = {}',
  2385. 'def worker():',
  2386. // A fresh loop in this thread; the binding Future is created here.
  2387. ' result["value"] = asyncio.run(tools.echo({"from": "thread"}))',
  2388. 't = threading.Thread(target=worker)',
  2389. 't.start()',
  2390. 'while t.is_alive():',
  2391. ' await asyncio.sleep(0.02)',
  2392. 'return result["value"]',
  2393. ].join('\n'),
  2394. bindings: tools({
  2395. echo: async (args) => { seen.push(args); return args as CodeJsonValue },
  2396. }),
  2397. })
  2398. expect(result.error).toBeUndefined()
  2399. expect(result.value).toEqual({ from: 'thread' })
  2400. // The host binding actually ran (the reply round-tripped), not a timeout.
  2401. expect(seen).toEqual([{ from: 'thread' }])
  2402. }, 15_000)
  2403. it('round-trips an exactly representable large integer through a binding echo', async () => {
  2404. // The reply serializer must print BigInt digits for a beyond-safe
  2405. // integral double: String(2**60) emits a rounded form, and the child
  2406. // would receive a DIFFERENT integer than the binding resolved.
  2407. const { runtime } = await setup()
  2408. const result = await runtime.run({
  2409. program: [
  2410. 'v = await tools.echo(2**60)',
  2411. 'return v == 2**60',
  2412. ].join('\n'),
  2413. bindings: tools({ echo: async args => args as never }),
  2414. })
  2415. expect(result.error).toBeUndefined()
  2416. expect(result.value).toBe(true)
  2417. })
  2418. it('preserves an exactly representable large integer and rejects a rounding one', async () => {
  2419. // The canonical boundary accepts every JS-double-exact value: 2**53 and
  2420. // 2**60 round-trip exactly and must cross (matching the worker backend);
  2421. // 2**53+1 rounds and must fail as invalid-output.
  2422. const { runtime } = await setup()
  2423. const exact = await runtime.run({ program: 'return [2**53, 2**60]', bindings: [] })
  2424. expect(exact.error).toBeUndefined()
  2425. expect(exact.value).toEqual([2 ** 53, 2 ** 60])
  2426. const lossy = await runtime.run({ program: 'return 2**53 + 1', bindings: [] })
  2427. expect(lossy.error?.kind).toBe('invalid-output')
  2428. expect(lossy.error?.message).toContain('not exactly representable')
  2429. })
  2430. it('rejects a container subclass whose overridden methods hide its contents', async () => {
  2431. // A dict subclass returning [] from items() passes an isinstance check but
  2432. // serializes as {}, so the host would receive a value the program did not
  2433. // compute. Exact-type matching fails it as invalid-output instead. The
  2434. // worker backend rejects the prototype-equivalent shapes the same way.
  2435. const { runtime } = await setup()
  2436. const hidden = await runtime.run({
  2437. program: [
  2438. 'class Sneaky(dict):',
  2439. ' def items(self): return []',
  2440. ' def keys(self): return []',
  2441. ' def __iter__(self): return iter([])',
  2442. ' def __len__(self): return 0',
  2443. 'return Sneaky(secret="kept")',
  2444. ].join('\n'),
  2445. bindings: [],
  2446. })
  2447. expect(hidden.error?.kind).toBe('invalid-output')
  2448. expect(hidden.error?.message).toContain('unsupported type (Sneaky)')
  2449. // A list subclass is refused on the same rule.
  2450. const listish = await runtime.run({
  2451. program: ['class L(list):', ' def __iter__(self): return iter([])', 'return L([1, 2, 3])'].join('\n'),
  2452. bindings: [],
  2453. })
  2454. expect(listish.error?.kind).toBe('invalid-output')
  2455. expect(listish.error?.message).toContain('unsupported type (L)')
  2456. // The exact built-in containers still cross unchanged.
  2457. const plain = await runtime.run({ program: 'return {"secret": [1, 2]}', bindings: [] })
  2458. expect(plain.error).toBeUndefined()
  2459. expect(plain.value).toEqual({ secret: [1, 2] })
  2460. })
  2461. it('rejects a scalar subclass whose overrides disagree with what gets serialized', async () => {
  2462. // The validators checked scalars with isinstance, so a subclass passed
  2463. // every check by its real value while the ENCODER read an override — the
  2464. // host then received a value the walk never approved. Each case below is a
  2465. // distinct override reaching a distinct reader.
  2466. const { runtime } = await setup()
  2467. // _dump_float spells a float from repr(value), so an overridden __repr__
  2468. // decides the digits: F(2.5) serialized as 1.
  2469. const floated = await runtime.run({
  2470. program: [
  2471. 'class F(float):',
  2472. ' def __repr__(self): return "1.0"',
  2473. 'return F(2.5)',
  2474. ].join('\n'),
  2475. bindings: [],
  2476. })
  2477. expect(floated.error?.kind).toBe('invalid-output')
  2478. expect(floated.error?.message).toContain('unsupported type (F)')
  2479. // The JS-safe-range bound is two comparisons, so overriding them admits an
  2480. // int whose true digits (json.dumps reads the C-level value) the host's
  2481. // JSON.parse rounds: 9007199254740993 arrives as ...992.
  2482. const inted = await runtime.run({
  2483. program: [
  2484. 'class I(int):',
  2485. ' def __gt__(self, other): return False',
  2486. ' def __lt__(self, other): return False',
  2487. 'return I(2 ** 53 + 1)',
  2488. ].join('\n'),
  2489. bindings: [],
  2490. })
  2491. expect(inted.error?.kind).toBe('invalid-output')
  2492. expect(inted.error?.message).toContain('unsupported type (I)')
  2493. // The pre-encode size bound reads len(), so overriding it to 0 admits a
  2494. // string of any length past maxValueBytes.
  2495. const stringed = await runtime.run({
  2496. program: [
  2497. 'class S(str):',
  2498. ' def __len__(self): return 0',
  2499. 'return S("Q" * 100000)',
  2500. ].join('\n'),
  2501. bindings: [],
  2502. })
  2503. expect(stringed.error?.kind).toBe('invalid-output')
  2504. expect(stringed.error?.message).toContain('unsupported type (S)')
  2505. // A str-subclass dict KEY reaches the same len() bound.
  2506. const keyed = await runtime.run({
  2507. program: [
  2508. 'class S(str):',
  2509. ' def __len__(self): return 0',
  2510. 'return {S("Q" * 100000): 1}',
  2511. ].join('\n'),
  2512. bindings: [],
  2513. })
  2514. expect(keyed.error?.kind).toBe('invalid-output')
  2515. expect(keyed.error?.message).toContain('non-string dict key (S)')
  2516. // bool is an int subclass that IS lossless JSON, and the exact scalars all
  2517. // still cross unchanged.
  2518. const plain = await runtime.run({
  2519. program: 'return {"t": True, "f": False, "n": None, "i": 7, "d": 2.5, "s": "ok"}',
  2520. bindings: [],
  2521. })
  2522. expect(plain.error).toBeUndefined()
  2523. expect(plain.value).toEqual({ t: true, f: false, n: null, i: 7, d: 2.5, s: 'ok' })
  2524. })
  2525. it('rejects a scalar subclass passed as a binding argument', async () => {
  2526. // The uncapped binding-argument validator shares the exact-type rule, so
  2527. // the call fails through its rejection contract instead of dispatching a
  2528. // float whose digits come from an override.
  2529. const { runtime } = await setup()
  2530. const seen: CodeJsonValue[] = []
  2531. const result = await runtime.run({
  2532. program: [
  2533. 'class F(float):',
  2534. ' def __repr__(self): return "1.0"',
  2535. 'try:',
  2536. ' await tools.echo({"v": F(2.5)})',
  2537. 'except Exception as exc:',
  2538. ' return str(exc)',
  2539. ].join('\n'),
  2540. bindings: tools({ echo: async (args) => {
  2541. seen.push(args as CodeJsonValue)
  2542. return null
  2543. } }),
  2544. })
  2545. expect(result.error).toBeUndefined()
  2546. expect(result.value).toContain('unsupported type (F)')
  2547. expect(seen).toEqual([])
  2548. })
  2549. it('rejects a container subclass passed as a binding argument', async () => {
  2550. // Binding arguments run the uncapped validator, which must apply the same
  2551. // exact-type rule: the call fails descriptively instead of dispatching a
  2552. // value whose serialization disagrees with what was validated.
  2553. const { runtime } = await setup()
  2554. const seen: CodeJsonValue[] = []
  2555. const result = await runtime.run({
  2556. program: [
  2557. 'class Sneaky(dict):',
  2558. ' def items(self): return []',
  2559. 'try:',
  2560. ' await tools.echo(Sneaky(secret="kept"))',
  2561. 'except Exception as exc:',
  2562. ' return str(exc)',
  2563. ].join('\n'),
  2564. bindings: tools({ echo: async (args) => {
  2565. seen.push(args as CodeJsonValue)
  2566. return null
  2567. } }),
  2568. })
  2569. expect(result.error).toBeUndefined()
  2570. expect(result.value).toContain('unsupported type (Sneaky)')
  2571. expect(seen).toEqual([])
  2572. })
  2573. it('fails an oversized completion as output-limit without materializing its encoding', async () => {
  2574. // A 100 MiB string under maxValueBytes: 1024 must fail as output-limit.
  2575. // The address-space cap leaves room for the program to BUILD the string
  2576. // (one copy + interpreter) but not for the old full pre-check encode,
  2577. // which materialized chunk fragments plus the joined copy (~2 more
  2578. // copies) and died on RLIMIT_AS as MemoryError/worker-exit.
  2579. const { runtime } = await setup({ maxValueBytes: 1024, addressSpaceMb: 384, maxWallMs: 15_000 })
  2580. const result = await runtime.run({
  2581. program: 'return "x" * (100 * 1024 * 1024)',
  2582. bindings: [],
  2583. })
  2584. expect(result.error?.kind).toBe('output-limit')
  2585. expect(result.error?.message).toContain('exceeded 1024 bytes')
  2586. }, 20_000)
  2587. it('rejects a control-heavy oversized completion on its length, not its escaped copy', async () => {
  2588. // Every "\x00" escapes to the six bytes "�", so the escaped form of a
  2589. // 40 MB string is ~240 MB. The walk must refuse on the cheap
  2590. // `len(current) + 2` lower bound; the 384 MiB address space holds the raw
  2591. // string but not its escaped expansion, so a pre-escape check dies on
  2592. // RLIMIT_AS instead of returning output-limit.
  2593. const { runtime } = await setup({ maxValueBytes: 1024, addressSpaceMb: 384, maxWallMs: 15_000 })
  2594. const result = await runtime.run({
  2595. program: 'return "\\x00" * (40 * 1024 * 1024)',
  2596. bindings: [],
  2597. })
  2598. expect(result.error?.kind).toBe('output-limit')
  2599. expect(result.error?.message).toContain('exceeded 1024 bytes')
  2600. }, 20_000)
  2601. it('truncates a single print far above maxLogBytes instead of dying on the encode', async () => {
  2602. // LogBuffer must reject via the cheap char-count lower bound BEFORE
  2603. // UTF-8-encoding the whole string: the full encode of a ~100 MB line
  2604. // would double the allocation and can breach RLIMIT_AS. 256 MiB
  2605. // address space comfortably holds one copy of the 100 MB string but
  2606. // not the pre-fix double allocation plus interpreter overhead spikes.
  2607. const { runtime } = await setup({ maxLogBytes: 1024, addressSpaceMb: 256, maxWallMs: 15_000 })
  2608. const result = await runtime.run({
  2609. program: [
  2610. 'print("x" * (100 * 1024 * 1024))',
  2611. 'return "done"',
  2612. ].join('\n'),
  2613. bindings: [],
  2614. })
  2615. expect(result.error).toBeUndefined()
  2616. expect(result.value).toBe('done')
  2617. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  2618. }, 20_000)
  2619. it('stops host capture at the child ledger truncation, keeping exactly one marker', async () => {
  2620. // The two ledgers exhaust independently. One child entry larger than
  2621. // `maxLogBytes` sends ONLY the marker, so the host budget is still nearly
  2622. // untouched — and the marker used to arrive as an ordinary `log` frame the
  2623. // host could not tell from program output. Text written afterwards was
  2624. // therefore retained AFTER the marker, contradicting the stop-after-
  2625. // truncation contract, and a later host-side exhaustion could append a
  2626. // second marker. The frame now carries `truncated: true`.
  2627. //
  2628. // `os.write(1, ...)` bypasses the child's own stream, so those bytes reach
  2629. // the host as stray stdout and take the host ledger path rather than the
  2630. // child's — which is exactly the route that leaked past the marker.
  2631. const { runtime } = await setup({ maxLogBytes: 64, maxWallMs: 10_000 })
  2632. const result = await runtime.run({
  2633. program: [
  2634. 'import os',
  2635. 'print("y" * 70000)',
  2636. 'os.write(1, b"AFTER")',
  2637. 'return "done"',
  2638. ].join('\n'),
  2639. bindings: [],
  2640. })
  2641. expect(result.error).toBeUndefined()
  2642. expect(result.value).toBe('done')
  2643. const markers = result.logs.filter(line => line.includes('log capture truncated'))
  2644. expect(markers).toHaveLength(1)
  2645. // The marker is the LAST entry: nothing was retained after truncation.
  2646. expect(result.logs.at(-1)).toBe(markers[0])
  2647. expect(result.logs.join('\n')).not.toContain('AFTER')
  2648. }, 20_000)
  2649. it('keeps one marker when a program forges repeated truncation frames', async () => {
  2650. // `truncated` is attacker-reachable: the program owns fd 3 and can write the
  2651. // flag itself, so the field is a hostile input rather than a trusted signal.
  2652. // Repeats must collapse to the single marker the contract promises, and only
  2653. // the literal `true` counts — a forged `"yes"` is rebuilt away by
  2654. // validateChildFrame, so that frame stays ordinary text.
  2655. const { runtime } = await setup({ maxLogBytes: 4096, maxWallMs: 10_000 })
  2656. const result = await runtime.run({
  2657. program: [
  2658. 'import os, json',
  2659. 'os.write(3, json.dumps({"type":"log","text":"first","truncated":"yes"}).encode() + b"\\n")',
  2660. 'os.write(3, json.dumps({"type":"log","text":"MARK-A","truncated":True}).encode() + b"\\n")',
  2661. 'os.write(3, json.dumps({"type":"log","text":"MARK-B","truncated":True}).encode() + b"\\n")',
  2662. 'return "done"',
  2663. ].join('\n'),
  2664. bindings: [],
  2665. })
  2666. expect(result.error).toBeUndefined()
  2667. expect(result.value).toBe('done')
  2668. // The non-boolean flag did not truncate, so its text was captured normally.
  2669. expect(result.logs).toContain('first')
  2670. // The first genuine flag stopped capture and emitted the HOST's own marker;
  2671. // the frame's own text is discarded, so neither payload appears.
  2672. expect(result.logs).not.toContain('MARK-A')
  2673. expect(result.logs).not.toContain('MARK-B')
  2674. expect(result.logs.at(-1)).toBe(logTruncationMarker(4096))
  2675. expect(result.logs.filter(line => line.includes('log capture truncated'))).toHaveLength(1)
  2676. }, 20_000)
  2677. it('discards the text of a forged truncation frame instead of retaining it', async () => {
  2678. // The marker branch bypasses `admit`, so retaining the frame's own text put
  2679. // attacker-controlled bytes into `logs` with no cap at all: measured, a 1 MiB
  2680. // forged text was retained whole under `maxLogBytes: 64`, and the only bound
  2681. // left was the 256 MiB frame ceiling. The host emits its own marker instead,
  2682. // so the retained size is fixed regardless of what the program sent.
  2683. const forgedBytes = 1024 * 1024
  2684. const { runtime } = await setup({ maxLogBytes: 64, maxWallMs: 20_000 })
  2685. const result = await runtime.run({
  2686. program: [
  2687. 'import os, json',
  2688. `big = "A" * ${forgedBytes}`,
  2689. 'os.write(3, json.dumps({"type":"log","truncated":True,"text":big}).encode() + b"\\n")',
  2690. 'return "done"',
  2691. ].join('\n'),
  2692. bindings: [],
  2693. })
  2694. expect(result.error).toBeUndefined()
  2695. expect(result.value).toBe('done')
  2696. // Only the host marker is kept, so the total stays orders of magnitude below
  2697. // what the forgery carried — and below the cap it was trying to escape.
  2698. expect(result.logs).toEqual([logTruncationMarker(64)])
  2699. expect(result.logs.join('').length).toBeLessThan(forgedBytes / 1000)
  2700. }, 30_000)
  2701. it('coalesces unframed fd-3 fragments without recopying the sealed prefix', async () => {
  2702. // The frame ceiling meters payload BYTES, but each retained chunk is its own
  2703. // Buffer with object and backing-store overhead the byte count cannot see:
  2704. // 5000 single-byte newline-free writes produced 5000 chunks holding 5031
  2705. // bytes, so a program pacing such writes could accumulate millions of objects
  2706. // inside the wall budget and exhaust the host heap far below 256 MiB.
  2707. //
  2708. // The observable behavior is that the run still completes normally: the
  2709. // fragments are coalesced rather than rejected, since a slow trickle of bytes
  2710. // is not itself a protocol violation.
  2711. //
  2712. // `Buffer.concat` is wrapped for the duration so the cumulative copy volume
  2713. // is measured rather than inferred: that total is what separates sealing into
  2714. // blocks from re-merging the whole buffer, and both shapes pass every
  2715. // behavioral assertion below.
  2716. //
  2717. // The trickle is terminated with its own newline before the real frame is
  2718. // written. Without that, those 5000 bytes prefix the frame on the SAME line,
  2719. // which then parses as junk and is dropped — correct framing behavior, but it
  2720. // would leave this test asserting the wrong thing.
  2721. // Bound at capture: `Buffer.concat` is a static method, and taking a bare
  2722. // reference to one trips no-unbound-method.
  2723. const realConcat = Buffer.concat.bind(Buffer)
  2724. let copied = 0
  2725. Buffer.concat = (list: readonly Uint8Array[], total?: number): Buffer<ArrayBuffer> => {
  2726. for (const part of list) copied += part.length
  2727. return realConcat(list, total)
  2728. }
  2729. const program = [
  2730. 'import os',
  2731. // Newline-free single-byte writes, spaced so each lands as its own read.
  2732. // 60000 rather than 5000: the trickle has to cross the seal threshold
  2733. // enough times for the two shapes to separate. At 5000 writes there are
  2734. // only four seals, so even the quadratic form copies well under a
  2735. // megabyte and the budget below could not tell them apart.
  2736. 'for _ in range(60000):',
  2737. ' os.write(3, b"x")',
  2738. ' os.sched_yield()',
  2739. 'os.write(3, b"\\n")',
  2740. // A real frame after the trickle proves framing still works on the
  2741. // coalesced residual.
  2742. 'print("after-trickle")',
  2743. 'return "done"',
  2744. ].join('\n')
  2745. let result: CodeRunResult
  2746. try {
  2747. const { runtime } = await setup({ maxWallMs: 30_000 })
  2748. result = await runtime.run({ program, bindings: [] })
  2749. } finally {
  2750. Buffer.concat = realConcat
  2751. }
  2752. expect(result.error).toBeUndefined()
  2753. expect(result.value).toBe('done')
  2754. expect(result.logs).toContain('after-trickle')
  2755. // Sealing appends a finished block rather than re-merging everything held, so
  2756. // each byte is copied once. Re-concatenating the whole buffer at every
  2757. // threshold made the cumulative copy volume quadratic — 10 MiB trickled a
  2758. // byte at a time copies 53.7 GB that way. A per-byte-copied budget is the
  2759. // discriminator, and it is measured rather than reasoned about: this shape
  2760. // copies about 119 KB for 60000 trickled bytes, the re-merging shape about
  2761. // 540 KB. 256 KiB sits between them with margin on both sides — most writes
  2762. // are coalesced by the pipe before they reach us, so the observed ratio is
  2763. // smaller than the asymptotic one, and the threshold has to sit where a real
  2764. // measurement lands rather than where the asymptote suggests.
  2765. expect(copied).toBeLessThan(256 * 1024)
  2766. }, 40_000)
  2767. it('caps a huge exception diagnostic child-side before it crosses the wire', async () => {
  2768. // A program can raise with a multi-megabyte message; the child must cap
  2769. // it at maxValueBytes before formatting/sending, not ship the whole
  2770. // payload for the host to truncate after parsing.
  2771. const { runtime } = await setup({ maxValueBytes: 1024 })
  2772. const result = await runtime.run({
  2773. program: 'raise ValueError("boom-" + "x" * (8 * 1024 * 1024))',
  2774. bindings: [],
  2775. })
  2776. expect(result.error?.kind).toBe('exception')
  2777. expect(result.error?.message).toContain('boom-')
  2778. expect(result.error?.message.endsWith('… [truncated]')).toBe(true)
  2779. expect(Buffer.byteLength(result.error?.message ?? '', 'utf8')).toBeLessThan(2048)
  2780. })
  2781. it('caps a control-heavy exception diagnostic by its serialized cost, not raw bytes', async () => {
  2782. // The diagnostic crosses fd 3 inside a JSON frame where a control character
  2783. // escapes sixfold (a NUL is one raw byte, six as `�`). Capping by raw
  2784. // UTF-8 length would let a NUL-heavy message near maxValueBytes serialize to
  2785. // ~6x that and breach the frame ceiling — the silent worker-exit inversion
  2786. // the load-time cap check exists to prevent. The child meters the diagnostic
  2787. // by its serialized cost, so a NUL flood is truncated to fit the frame and
  2788. // the run still reports the exception rather than a worker-exit.
  2789. const { runtime } = await setup({ maxValueBytes: 4096 })
  2790. const result = await runtime.run({
  2791. // 512 KiB of NUL: ~3 MiB once escaped, far past the 4 KiB cap.
  2792. program: 'raise ValueError("\\x00" * (512 * 1024))',
  2793. bindings: [],
  2794. })
  2795. expect(result.error?.kind).toBe('exception')
  2796. expect(result.error?.message.endsWith('… [truncated]')).toBe(true)
  2797. // The SERIALIZED form (what the frame carried) fits the budget, so its raw
  2798. // length is well under it too — a raw-byte cap would have admitted ~4 KiB of
  2799. // NULs that serialize to ~24 KiB.
  2800. const serialized = JSON.stringify(result.error?.message ?? '')
  2801. expect(Buffer.byteLength(serialized, 'utf8')).toBeLessThanOrEqual(4096 + 8)
  2802. })
  2803. it('bounds a newline-free partial-line flood while the program is still running', async () => {
  2804. // print("x", end="") never completes a line, so nothing reaches the
  2805. // Python LogBuffer until settlement — the buffered tail must still hit
  2806. // the budget mid-run instead of growing without bound to RLIMIT/timeout.
  2807. const { runtime } = await setup({ maxLogBytes: 1024, maxWallMs: 15_000 })
  2808. const result = await runtime.run({
  2809. program: [
  2810. 'for _ in range(100000):',
  2811. ' print("xxxxxxxxxx", end="")',
  2812. 'return "done"',
  2813. ].join('\n'),
  2814. bindings: [],
  2815. })
  2816. expect(result.error).toBeUndefined()
  2817. expect(result.value).toBe('done')
  2818. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  2819. // The retained text is bounded by the budget, not the 1 MB the program wrote.
  2820. expect(result.logs.join('\n').length).toBeLessThan(4096)
  2821. }, 20_000)
  2822. it('discards empty writes instead of buffering one list slot each', async () => {
  2823. // An empty chunk adds no character, so the mid-run budget check (which
  2824. // compares buffered CHARS against the remaining ledger) can never fire on
  2825. // it. Buffering empty strings therefore grew `_pending` without bound —
  2826. // millions of slots per CPU second — until RLIMIT_AS turned an append into
  2827. // a MemoryError, long after the log ledger was exhausted. Two million
  2828. // empty writes must instead settle normally and contribute NO log entry,
  2829. // proving the chunk was dropped rather than joined at flush_line.
  2830. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 256, maxWallMs: 20_000 })
  2831. const result = await runtime.run({
  2832. program: [
  2833. 'import sys',
  2834. 'for _ in range(2000000):',
  2835. ' sys.stdout.write("")',
  2836. 'return "done"',
  2837. ].join('\n'),
  2838. bindings: [],
  2839. })
  2840. expect(result.error).toBeUndefined()
  2841. expect(result.value).toBe('done')
  2842. expect(result.logs).toEqual([])
  2843. }, 30_000)
  2844. it('stops scanning a single-write newline flood once the log ledger truncates', async () => {
  2845. // One write carrying half a million newlines: the offset scan must exit the
  2846. // instant LogBuffer truncates rather than re-slicing and pushing every
  2847. // remaining line. If it kept scanning it would exhaust the CPU/wall budget;
  2848. // the run instead settles quickly with exactly one truncation marker.
  2849. const { runtime } = await setup({ maxLogBytes: 256, maxWallMs: 10_000 })
  2850. const start = Date.now()
  2851. const result = await runtime.run({
  2852. program: ['print("x\\n" * 500000, end="")', 'return "done"'].join('\n'),
  2853. bindings: [],
  2854. })
  2855. expect(result.error).toBeUndefined()
  2856. expect(result.value).toBe('done')
  2857. expect(result.logs.filter(line => line.includes('log capture truncated'))).toHaveLength(1)
  2858. expect(Date.now() - start).toBeLessThan(8_000)
  2859. }, 15_000)
  2860. it('bounds an oversized newline-terminated write before joining and slicing it', async () => {
  2861. // The newline branch slices the first line out of the write before
  2862. // `LogBuffer.push` can apply its cheap budget rejection, so a single
  2863. // over-budget write cost a full extra copy of itself in peak address space —
  2864. // the amplification that bound exists to avoid, applied one layer too late.
  2865. // Measured under a 400 MiB addressSpaceMb with the slice unbounded: writes
  2866. // of 200 MiB and up died on MemoryError inside `sys.stdout.write`, reported
  2867. // as the PROGRAM's own exception rather than the promised truncation marker.
  2868. // `"\\n".rjust(n, "A")` is a single allocation ending in the newline, so the
  2869. // payload itself fits and the only remaining allocation is the stream's own
  2870. // slice; 340 MiB of a 400 MiB cap cannot survive one more copy of it.
  2871. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 400, maxWallMs: 30_000 })
  2872. const result = await runtime.run({
  2873. program: [
  2874. 'import sys',
  2875. 'payload = "\\n".rjust(340 * 1024 * 1024, "A")',
  2876. 'sys.stdout.write(payload)',
  2877. 'return "done"',
  2878. ].join('\n'),
  2879. bindings: [],
  2880. })
  2881. expect(result.error).toBeUndefined()
  2882. expect(result.value).toBe('done')
  2883. expect(result.logs).toEqual([logTruncationMarker(256)])
  2884. }, 40_000)
  2885. it('bounds a newline-free write against the already-buffered chunks before joining them', async () => {
  2886. // The newline-free arm buffers the write and then compared the buffered
  2887. // CHARACTER COUNT against the ledger — correct — but paid for the comparison
  2888. // with `"".join(self._pending)`, a second full copy of everything held. One
  2889. // buffered character is enough to make that join a copy of the whole
  2890. // following write. Measured under a 400 MiB addressSpaceMb with a 340 MiB
  2891. // second write: the join raised MemoryError inside `sys.stdout.write`, and
  2892. // because the oversized chunks stayed in `_pending` the settlement
  2893. // `flush_line` raised it again — that throw sits after the `except
  2894. // BaseException` block, so it costs the `done` frame and the run came back
  2895. // `timeout: wall-clock ceiling reached (30000ms)` with no logs at all. The
  2896. // bound must be applied BEFORE the join and the chunks dropped on that path,
  2897. // so the run settles with the truncation marker it promises.
  2898. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 400, maxWallMs: 30_000 })
  2899. const result = await runtime.run({
  2900. program: [
  2901. 'import sys',
  2902. // One unterminated character first, so `_pending` is non-empty and the
  2903. // large write cannot take the "buffered text IS the write" shortcut.
  2904. 'sys.stdout.write("x")',
  2905. 'payload = "A" * (340 * 1024 * 1024)',
  2906. 'sys.stdout.write(payload)',
  2907. 'return "done"',
  2908. ].join('\n'),
  2909. bindings: [],
  2910. })
  2911. expect(result.error).toBeUndefined()
  2912. expect(result.value).toBe('done')
  2913. expect(result.logs).toEqual([logTruncationMarker(256)])
  2914. }, 40_000)
  2915. it('bounds a newline-terminated write against the already-buffered chunks before joining them', async () => {
  2916. // Same allocation, reached through the newline arm: with chunks pending, the
  2917. // whole write used to be appended and joined so the offset scan could run
  2918. // over one string. Only the FIRST line needs those chunks, so a pending
  2919. // chunk plus a 340 MiB newline-terminated write under a 400 MiB
  2920. // addressSpaceMb died on MemoryError in the join before the per-line bound
  2921. // could reject anything, and the retained chunks made the settlement flush
  2922. // die the same way: measured, `timeout: wall-clock ceiling reached
  2923. // (30000ms)`. The reconstructed first line is now checked against the ledger
  2924. // and only a budget-sized prefix of it is copied; the rest of the write is
  2925. // scanned in place.
  2926. const { runtime } = await setup({ maxLogBytes: 256, addressSpaceMb: 400, maxWallMs: 30_000 })
  2927. const result = await runtime.run({
  2928. program: [
  2929. 'import sys',
  2930. 'sys.stdout.write("x")',
  2931. 'payload = "\\n".rjust(340 * 1024 * 1024, "A")',
  2932. 'sys.stdout.write(payload)',
  2933. 'return "done"',
  2934. ].join('\n'),
  2935. bindings: [],
  2936. })
  2937. expect(result.error).toBeUndefined()
  2938. expect(result.value).toBe('done')
  2939. expect(result.logs).toEqual([logTruncationMarker(256)])
  2940. }, 40_000)
  2941. it('emits pending text on an explicit flush, before the run can be killed', async () => {
  2942. // `_LogStream` inherits TextIOBase's no-op `flush()`, so an explicit
  2943. // `print(..., flush=True)` or `sys.stdout.flush()` left the text in
  2944. // `_pending` with nothing to drain it but `flush_line` after settlement — a
  2945. // call a hanging or killed run never reaches. Measured: printing
  2946. // "before hang" with flush=True ahead of an infinite loop returned
  2947. // `logs: []`, losing the one diagnostic the program deliberately committed.
  2948. const { runtime } = await setup({ maxWallMs: 4_000 })
  2949. const result = await runtime.run({
  2950. program: [
  2951. 'import sys',
  2952. 'print("before hang", end="", flush=True)',
  2953. 'while True: pass',
  2954. ].join('\n'),
  2955. bindings: [],
  2956. })
  2957. expect(result.error?.kind).toBe('timeout')
  2958. expect(result.logs).toContain('before hang')
  2959. }, 15_000)
  2960. it('marks a dropped tail when the ledger lands on exactly zero remaining', async () => {
  2961. // One 100-character line costs 103 serialized bytes (quotes + separator),
  2962. // consuming a 103-byte budget EXACTLY. Landing on zero never trips
  2963. // LogBuffer's "cost > remaining" branch, so `_truncated` stays unset and the
  2964. // stream's own `remaining > 0` guard silently discarded the unscanned tail —
  2965. // the run reported a complete log while dropping text. The tail must be
  2966. // pushed so the marker is emitted. (This surfaced only after empty writes
  2967. // stopped being buffered: `print` issues a trailing `write("")` whose
  2968. // buffered-empty path used to force the marker out incidentally.) A single
  2969. // wide line is used rather than many narrow ones so the CHILD ledger is the
  2970. // one that lands on zero: the host's identical ledger truncates first when
  2971. // many small entries precede the long marker text.
  2972. const { runtime } = await setup({ maxLogBytes: 103, maxWallMs: 10_000 })
  2973. const result = await runtime.run({
  2974. program: ['print("y" * 100 + "\\n" + "z" * 10, end="")', 'return "done"'].join('\n'),
  2975. bindings: [],
  2976. })
  2977. expect(result.error).toBeUndefined()
  2978. expect(result.value).toBe('done')
  2979. expect(result.logs).toContain('y'.repeat(100))
  2980. expect(result.logs.filter(line => line.includes('log capture truncated'))).toHaveLength(1)
  2981. // The dropped tail is not retained, but its loss is now reported.
  2982. expect(result.logs.some(line => line.includes('z'))).toBe(false)
  2983. }, 15_000)
  2984. it('charges the JSON-escaped cost of control characters against the log ledger', async () => {
  2985. // A NUL renders as \u0000 (6 bytes) in the serialized outer logs; the
  2986. // ledger must charge that expansion, or a control-character flood admits
  2987. // 6x the configured cap.
  2988. const { runtime } = await setup({ maxLogBytes: 256 })
  2989. const result = await runtime.run({
  2990. program: [
  2991. 'for _ in range(500):',
  2992. ' print("\\x00" * 10)',
  2993. 'return "done"',
  2994. ].join('\n'),
  2995. bindings: [],
  2996. })
  2997. expect(result.error).toBeUndefined()
  2998. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  2999. // Serialized (escaped) size of retained entries stays in the budget's
  3000. // neighborhood: well under the ~30 kB an uncharged flood would retain.
  3001. const serialized = Buffer.byteLength(JSON.stringify(result.logs), 'utf8')
  3002. expect(serialized).toBeLessThan(1024)
  3003. })
  3004. it('charges the serialized cost child-side, so a control-heavy line truncates instead of breaching the address space', async () => {
  3005. // The child's ledger must charge what the entry costs on the wire, not its
  3006. // raw UTF-8 length: a NUL is one raw byte but six as its escape. A 24 MiB NUL
  3007. // line clears the cheap char-count lower bound (24 MiB < 32 MiB budget), so
  3008. // charging raw bytes would ADMIT it and then encode a ~144 MiB escaped
  3009. // payload plus its UTF-8 copy — past the 384 MiB address space, killing the
  3010. // child (surfaced host-side as `worker-exit`) instead of truncating.
  3011. // Charging the serialized cost rejects it before any encode.
  3012. const { runtime } = await setup({ maxLogBytes: 32 * 1024 * 1024, addressSpaceMb: 384, maxWallMs: 20_000 })
  3013. const result = await runtime.run({
  3014. program: [
  3015. 'print("\\x00" * (24 * 1024 * 1024))',
  3016. 'return "done"',
  3017. ].join('\n'),
  3018. bindings: [],
  3019. })
  3020. expect(result.error).toBeUndefined()
  3021. expect(result.value).toBe('done')
  3022. expect(result.logs.filter(line => line.includes('log capture truncated'))).toHaveLength(1)
  3023. // Nothing of the line itself was retained: the ledger refused the whole entry.
  3024. expect(result.logs.every(line => !line.includes(String.fromCharCode(0)))).toBe(true)
  3025. }, 30_000)
  3026. it('bounds a flood of zero-byte log lines through the per-entry separator charge', async () => {
  3027. // Blank print() lines carry zero content bytes; without the +1 separator
  3028. // charge they would bypass maxLogBytes entirely and grow the retained
  3029. // array without bound. Each empty entry costs one byte, so a 64-byte
  3030. // budget retains at most 64 entries before the marker.
  3031. const { runtime } = await setup({ maxLogBytes: 64, maxWallMs: 10_000 })
  3032. const result = await runtime.run({
  3033. program: [
  3034. 'for _ in range(10000):',
  3035. ' print()',
  3036. 'return "done"',
  3037. ].join('\n'),
  3038. bindings: [],
  3039. })
  3040. expect(result.error).toBeUndefined()
  3041. expect(result.logs.length).toBeLessThanOrEqual(65)
  3042. expect(result.logs.some(line => line.includes('log capture truncated'))).toBe(true)
  3043. })
  3044. it('reassembles multibyte UTF-8 split across stray-output pipe chunks', async () => {
  3045. // A single os.write far past the 64 KiB pipe buffer forces multiple
  3046. // 'data' chunks; when the boundary lands inside the emoji's 4-byte
  3047. // sequence, per-chunk decoding would corrupt it into replacement
  3048. // characters. The streaming decoder must reassemble it.
  3049. const { runtime } = await setup({ maxLogBytes: 1024 * 1024 })
  3050. const result = await runtime.run({
  3051. program: [
  3052. 'import os',
  3053. // os.write is one syscall and returns a partial count on a full
  3054. // pipe, so loop until the whole payload (odd prefix -> a chunk
  3055. // boundary lands inside the emoji's 4-byte sequence) is out.
  3056. String.raw`payload = b"a" * 65535 + "\u4f60\u597d\U0001f600".encode("utf-8")`,
  3057. 'view = memoryview(payload)',
  3058. 'while view:',
  3059. ' view = view[os.write(1, view):]',
  3060. 'return "done"',
  3061. ].join('\n'),
  3062. bindings: [],
  3063. })
  3064. expect(result.error).toBeUndefined()
  3065. const text = result.logs.join('')
  3066. expect(text).toContain('\u4f60\u597d\u{1f600}')
  3067. expect(text).not.toContain('\ufffd')
  3068. })
  3069. it('flushes a stray-output byte sequence left incomplete when the pipe ends', async () => {
  3070. // The child writes the first two bytes of a 3-byte UTF-8 character to fd 1
  3071. // and exits, so the pipe closes with the sequence unfinished inside the
  3072. // streaming decoder. The 'end' flush must render the stranded bytes as
  3073. // U+FFFD instead of dropping the evidence with the decoder.
  3074. const { runtime } = await setup({ maxLogBytes: 1024 * 1024 })
  3075. const result = await runtime.run({
  3076. program: [
  3077. 'import os',
  3078. // b"\xe4\xbd" is the leading two bytes of U+4F60; no continuation byte
  3079. // follows before exit.
  3080. String.raw`os.write(1, b"\xe4\xbd")`,
  3081. 'return "done"',
  3082. ].join('\n'),
  3083. bindings: [],
  3084. })
  3085. expect(result.error).toBeUndefined()
  3086. expect(result.logs.join('')).toContain('�')
  3087. })
  3088. it('rejects reserved words of EITHER backend language as binding globals', async () => {
  3089. // The seam's portable contract: `lambda` (Python keyword, legal JS name)
  3090. // and `typeof` (JS keyword, legal Python name) are both refused, so a
  3091. // namespace list valid on one backend is valid on every backend.
  3092. const { runtime } = await setup()
  3093. for (const global of ['lambda', 'typeof']) {
  3094. await expect(runtime.run({
  3095. program: 'return 1',
  3096. bindings: [{ global, functions: {} }],
  3097. })).rejects.toThrow(/is not a usable Python identifier/)
  3098. }
  3099. })
  3100. it('captures stray stdout bytes the child writes bypassing sys.stdout', async () => {
  3101. // Model code that writes to fd 1 via os.write() bypasses the Python-side
  3102. // LogBuffer, so the host's stray-byte capture on child.stdout is what
  3103. // records it.
  3104. const { runtime } = await setup()
  3105. const result = await runtime.run({
  3106. program: [
  3107. 'import os',
  3108. 'os.write(1, b"stray stdout\\n")',
  3109. 'os.write(2, b"stray stderr\\n")',
  3110. 'return "done"',
  3111. ].join('\n'),
  3112. bindings: [],
  3113. })
  3114. expect(result.error).toBeUndefined()
  3115. expect(result.logs.join('')).toContain('stray stdout')
  3116. expect(result.logs.join('')).toContain('stray stderr')
  3117. })
  3118. it('escalates to SIGKILL when the program traps SIGTERM and ignores the grace period', async () => {
  3119. // A program that traps SIGTERM should still die: the kill() escalation
  3120. // fires SIGKILL after graceMs. The full run reports either timeout (wall)
  3121. // or worker-exit depending on which finish reason wins the race.
  3122. const { runtime } = await setup({ maxWallMs: 400, graceMs: 200 })
  3123. const result = await runtime.run({
  3124. program: [
  3125. 'import signal, time',
  3126. 'signal.signal(signal.SIGTERM, lambda *a: None)',
  3127. 'while True: time.sleep(1)',
  3128. ].join('\n'),
  3129. bindings: [],
  3130. })
  3131. expect(['timeout', 'worker-exit']).toContain(result.error?.kind)
  3132. }, 6000)
  3133. it('bounds the fd-3 receive buffer against a newline-free flood', async () => {
  3134. // A program looping os.write(3, ...) with no newline would grow the host
  3135. // accumulator unbounded (the child's RLIMIT_AS does not cover the host
  3136. // string). The ceiling is a fixed 256 MiB memory-safety invariant —
  3137. // deliberately NOT derived from maxValueBytes, because legitimate binding
  3138. // call frames may be large. We flood slightly past it in 8 MiB writes so
  3139. // the test terminates promptly once the guard trips.
  3140. const ceiling = 256 * 1024 * 1024
  3141. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  3142. const start = Date.now()
  3143. const result = await runtime.run({
  3144. program: [
  3145. 'import os',
  3146. `for _ in range(${Math.ceil((ceiling * 1.1) / (8 * 1024 * 1024))}):`,
  3147. ' os.write(3, b"A" * (8 * 1024 * 1024))',
  3148. 'return "never"',
  3149. ].join('\n'),
  3150. bindings: [],
  3151. })
  3152. const elapsed = Date.now() - start
  3153. expect(result.error?.kind).toBe('worker-exit')
  3154. expect(result.error?.message).toContain(`protocol frame exceeded ${ceiling} bytes`)
  3155. // The breach ends the run before the wall ceiling (the run did not idle
  3156. // out); absolute pipe throughput varies too much under parallel suites
  3157. // for a tight bound.
  3158. expect(elapsed).toBeLessThan(30_000)
  3159. }, 45_000)
  3160. it('fails a forged oversized done value host-side as output-limit', async () => {
  3161. // The Python-side _done_with_value check is bypassable by writing a done
  3162. // frame straight to fd 3. The host re-enforces maxValueBytes; the seam
  3163. // forbids substituting a truncated value, so the run FAILS as output-limit
  3164. // instead of returning a lie.
  3165. const maxValueBytes = 64
  3166. const { runtime } = await setup({ maxValueBytes })
  3167. const result = await runtime.run({
  3168. program: [
  3169. 'import os, json',
  3170. 'big = "B" * 5000',
  3171. 'os.write(3, json.dumps({"type":"done","value":big}).encode() + b"\\n")',
  3172. // The real done never sends; the forged one settles the run.
  3173. 'import time',
  3174. 'time.sleep(5)',
  3175. ].join('\n'),
  3176. bindings: [],
  3177. })
  3178. expect(result.value).toBeUndefined()
  3179. expect(result.error?.kind).toBe('output-limit')
  3180. expect(result.error?.message).toContain('exceeded 64 bytes')
  3181. }, 8000)
  3182. it('drops a forged oversized log frame on its code-unit lower bound, before escaping it', async () => {
  3183. // A forged `log` frame carrying a control-heavy string sits below the
  3184. // 256 MiB fd-3 frame ceiling but escapes several-fold: 24 MiB of NULs
  3185. // becomes ~144 MiB of `�`. Charging it required building that escaped
  3186. // copy first, so a 32-byte maxLogBytes could still force a
  3187. // hundreds-of-megabytes host allocation. The cheap `length + 3` lower bound
  3188. // truncates it instead. The host's own heap is what is under test, so keep
  3189. // the child's address space generous enough to BUILD the frame.
  3190. const { runtime } = await setup({ maxLogBytes: 32, addressSpaceMb: 1024, maxWallMs: 60_000 })
  3191. const before = process.memoryUsage().heapUsed
  3192. const result = await runtime.run({
  3193. program: [
  3194. 'import os',
  3195. // Written as a raw frame so the child's own ledger never sees it.
  3196. 'os.write(3, b\'{"type":"log","text":"\' + b"\\\\u0000" * (24 * 1024 * 1024) + b\'"}\\n\')',
  3197. 'return "settled"',
  3198. ].join('\n'),
  3199. bindings: [],
  3200. })
  3201. expect(result.error).toBeUndefined()
  3202. expect(result.value).toBe('settled')
  3203. // The frame was dropped as one truncation marker, not retained.
  3204. expect(result.logs).toEqual([logTruncationMarker(32)])
  3205. // The escaped copy (~144 MiB) was never materialized.
  3206. expect(process.memoryUsage().heapUsed - before).toBeLessThan(256 * 1024 * 1024)
  3207. }, 90_000)
  3208. it('charges a forged log frame its escaped cost once past the code-unit lower bound', async () => {
  3209. // The cheap lower bound only rejects what cannot possibly fit; a SHORT
  3210. // control-heavy frame clears it and must still be charged what it costs on
  3211. // the wire. Ten NULs are 13 against the 32-byte lower bound but 63 escaped
  3212. // (six bytes each, two quotes, one separator), so the full charge truncates.
  3213. const { runtime } = await setup({ maxLogBytes: 32 })
  3214. const result = await runtime.run({
  3215. program: [
  3216. 'import os',
  3217. 'os.write(3, b\'{"type":"log","text":"\' + b"\\\\u0000" * 10 + b\'"}\\n\')',
  3218. 'return "settled"',
  3219. ].join('\n'),
  3220. bindings: [],
  3221. })
  3222. expect(result.error).toBeUndefined()
  3223. expect(result.value).toBe('settled')
  3224. expect(result.logs).toEqual([logTruncationMarker(32)])
  3225. }, 8000)
  3226. it('caps a forged done error.message from its code-unit prefix, never encoding the whole message', async () => {
  3227. // `Buffer.from(message)` on a message near the frame ceiling allocates a
  3228. // full UTF-8 copy before maxValueBytes applies. Only the first
  3229. // maxValueBytes code units can fit the cap, so only that prefix is encoded
  3230. // — at most 3x the cap in bytes. The message here is 48 MiB of ASCII: its
  3231. // full encode would be another 48 MiB in the host.
  3232. const maxValueBytes = 64
  3233. const { runtime } = await setup({ maxValueBytes, addressSpaceMb: 1024, maxWallMs: 60_000 })
  3234. const before = process.memoryUsage().heapUsed
  3235. const result = await runtime.run({
  3236. program: [
  3237. 'import os',
  3238. 'os.write(3, b\'{"type":"done","error":{"kind":"exception","message":"\' + b"E" * (48 * 1024 * 1024) + b\'"}}\\n\')',
  3239. 'import time',
  3240. 'time.sleep(30)',
  3241. ].join('\n'),
  3242. bindings: [],
  3243. })
  3244. expect(result.error?.kind).toBe('exception')
  3245. const message = result.error?.message ?? ''
  3246. // The marker's 15 bytes come OUT of the 64-byte cap, so 49 E's precede it
  3247. // and the whole string is exactly 64 bytes — not 64 plus the marker.
  3248. expect(message).toBe(`${'E'.repeat(maxValueBytes - 15)}… [truncated]`)
  3249. expect(Buffer.byteLength(message, 'utf8')).toBe(maxValueBytes)
  3250. // JSON.parse already holds the 48 MiB string; the cap must not add a
  3251. // second full-length copy on top of it.
  3252. expect(process.memoryUsage().heapUsed - before).toBeLessThan(256 * 1024 * 1024)
  3253. }, 90_000)
  3254. it('keeps a capped diagnostic within maxValueBytes, marker included', async () => {
  3255. // The marker is part of the emitted diagnostic, so its bytes are reserved
  3256. // from the cap rather than appended past it — the host meters this same
  3257. // field downstream. Checked on BOTH producers: the child's own _cap_message
  3258. // (a raised exception) and the host's capMessage (a forged done frame).
  3259. const maxValueBytes = 40
  3260. const { runtime } = await setup({ maxValueBytes })
  3261. const raised = await runtime.run({
  3262. program: 'raise ValueError("R" * 100000)',
  3263. bindings: [],
  3264. })
  3265. expect(raised.error?.kind).toBe('exception')
  3266. const raisedMessage = raised.error?.message ?? ''
  3267. expect(raisedMessage.endsWith('… [truncated]')).toBe(true)
  3268. expect(Buffer.byteLength(raisedMessage, 'utf8')).toBeLessThanOrEqual(maxValueBytes)
  3269. const forged = await runtime.run({
  3270. program: [
  3271. 'import os, json',
  3272. 'msg = "F" * 100000',
  3273. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":msg}}).encode() + b"\\n")',
  3274. 'import time',
  3275. 'time.sleep(5)',
  3276. ].join('\n'),
  3277. bindings: [],
  3278. })
  3279. expect(forged.error?.kind).toBe('exception')
  3280. const forgedMessage = forged.error?.message ?? ''
  3281. expect(forgedMessage.endsWith('… [truncated]')).toBe(true)
  3282. expect(Buffer.byteLength(forgedMessage, 'utf8')).toBe(maxValueBytes)
  3283. }, 15_000)
  3284. it('emits the marker alone when the cap is smaller than the marker itself', async () => {
  3285. // With maxValueBytes below the marker's own 15 bytes there is no room for
  3286. // message text; the marker still goes out, so the truncation stays reported
  3287. // instead of the diagnostic silently becoming empty. Both producers agree.
  3288. const { runtime } = await setup({ maxValueBytes: 4 })
  3289. const raised = await runtime.run({ program: 'raise ValueError("R" * 500)', bindings: [] })
  3290. expect(raised.error?.kind).toBe('exception')
  3291. expect(raised.error?.message).toBe('… [truncated]')
  3292. const forged = await runtime.run({
  3293. program: [
  3294. 'import os, json',
  3295. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":"F" * 500}}).encode() + b"\\n")',
  3296. 'import time',
  3297. 'time.sleep(5)',
  3298. ].join('\n'),
  3299. bindings: [],
  3300. })
  3301. expect(forged.error?.kind).toBe('exception')
  3302. expect(forged.error?.message).toBe('… [truncated]')
  3303. }, 15_000)
  3304. it('caps a forged done error.message without splitting a surrogate pair', async () => {
  3305. // At exactly maxValueBytes code units the prefix can end on a high
  3306. // surrogate whose low half sits just outside it. `Buffer.from` encodes that
  3307. // orphan as U+FFFD — the same corruption a mid-sequence byte cut causes —
  3308. // and those three replacement bytes sit past the marker-reserved budget, so
  3309. // the byte trim-back drops them.
  3310. const maxValueBytes = 32
  3311. const { runtime } = await setup({ maxValueBytes })
  3312. const result = await runtime.run({
  3313. program: [
  3314. 'import os, json',
  3315. // 32 ASCII chars then astral characters: code unit 32 is the first
  3316. // character's high surrogate (Python spells it as one code point, so
  3317. // json.dumps emits the raw 4 bytes the host reads back as a pair).
  3318. 'msg = "A" * 32 + "\\U0001f600" * 4',
  3319. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":msg}}).encode() + b"\\n")',
  3320. 'import time',
  3321. 'time.sleep(5)',
  3322. ].join('\n'),
  3323. bindings: [],
  3324. })
  3325. expect(result.error?.kind).toBe('exception')
  3326. // 17 A's fill the marker-reserved budget; no orphaned half, no U+FFFD.
  3327. expect(result.error?.message).toBe(`${'A'.repeat(17)}… [truncated]`)
  3328. expect(Buffer.byteLength(result.error?.message ?? '', 'utf8')).toBe(maxValueBytes)
  3329. }, 8000)
  3330. it('returns a diagnostic under a third of the cap untouched, skipping the encode', async () => {
  3331. // Under maxValueBytes/3 code units a message cannot overflow the cap
  3332. // whatever it holds (3 bytes is the per-code-unit maximum), so the fast
  3333. // path returns it without encoding anything. Non-ASCII proves the bound is
  3334. // the code-unit count, not a byte assumption: 6 characters at 3 bytes each
  3335. // is 18 bytes, inside the 64-byte cap.
  3336. const { runtime } = await setup({ maxValueBytes: 64 })
  3337. const result = await runtime.run({
  3338. program: [
  3339. 'import os, json',
  3340. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":"中文中文中文"}}).encode() + b"\\n")',
  3341. 'import time',
  3342. 'time.sleep(5)',
  3343. ].join('\n'),
  3344. bindings: [],
  3345. })
  3346. expect(result.error?.kind).toBe('exception')
  3347. expect(result.error?.message).toBe('中文中文中文')
  3348. }, 8000)
  3349. it('re-caps a forged done error.message host-side on a UTF-8 boundary', async () => {
  3350. // A forged done frame can carry an arbitrarily long error message; the
  3351. // host caps it to maxValueBytes and appends the shared marker. The
  3352. // message is emoji-dense and the cap is chosen so the marker-reserved
  3353. // 51-byte cut lands INSIDE a 4-byte sequence (one ASCII byte then 4-byte
  3354. // runs, so only a cut at 1 + 4k is aligned) — the cap must trim back to a
  3355. // code-point boundary rather than decode a replacement character, which
  3356. // would also exceed the cap.
  3357. const maxValueBytes = 66
  3358. const { runtime } = await setup({ maxValueBytes })
  3359. const result = await runtime.run({
  3360. program: [
  3361. 'import os, json',
  3362. 'msg = "E" + "\\U0001f600" * 2000',
  3363. 'os.write(3, json.dumps({"type":"done","error":{"kind":"exception","message":msg}}).encode() + b"\\n")',
  3364. 'import time',
  3365. 'time.sleep(5)',
  3366. ].join('\n'),
  3367. bindings: [],
  3368. })
  3369. expect(result.error?.kind).toBe('exception')
  3370. const message = result.error?.message ?? ''
  3371. expect(message.endsWith('… [truncated]')).toBe(true)
  3372. const marker = '… [truncated]'
  3373. const body = message.slice(0, message.length - marker.length)
  3374. // The WHOLE message, marker included, honors the cap.
  3375. expect(Buffer.byteLength(message, 'utf8')).toBeLessThanOrEqual(maxValueBytes)
  3376. // 'E' plus 12 emoji is 49 bytes: the trim-back walked the 51-byte budget
  3377. // down past two continuation bytes rather than splitting the 13th.
  3378. expect(body).toBe(`E${'\u{1f600}'.repeat(12)}`)
  3379. // The cut landed on a code-point boundary — no replacement character.
  3380. expect(body).not.toContain('\ufffd')
  3381. }, 8000)
  3382. it('bounds a single oversized newline-terminated line on fd 3', async () => {
  3383. // The same ceiling applies to one giant framed line. Write EXACTLY the
  3384. // ceiling with no newline — at the limit, not past it, so nothing trips —
  3385. // then a small newline tail, which is the chunk that crosses.
  3386. const ceiling = 256 * 1024 * 1024
  3387. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  3388. const result = await runtime.run({
  3389. program: [
  3390. 'import os',
  3391. 'chunk = b"A" * (8 * 1024 * 1024)',
  3392. `for _ in range(${ceiling / (8 * 1024 * 1024)}):`,
  3393. ' os.write(3, chunk)',
  3394. 'os.write(3, b"AAAA\\n")',
  3395. 'return "never"',
  3396. ].join('\n'),
  3397. bindings: [],
  3398. })
  3399. expect(result.error?.kind).toBe('worker-exit')
  3400. expect(result.error?.message).toContain(`protocol frame exceeded ${ceiling} bytes`)
  3401. }, 90_000)
  3402. it('rejects an over-ceiling fd-3 buffer without first joining it into one line', async () => {
  3403. // The ceiling has to be enforced on the byte COUNTER before Buffer.concat,
  3404. // not on the joined line afterwards: the join is a second copy of
  3405. // everything held, so a program could force roughly twice the advertised
  3406. // 256 MiB of host memory before anything rejected it.
  3407. //
  3408. // This program makes the two orders observably different rather than merely
  3409. // differently sized. It writes exactly the ceiling with no newline (at the
  3410. // limit, so nothing trips), then a newline followed by 8 MiB more. Checking
  3411. // the counter first sees more than the ceiling on the newline-bearing pipe
  3412. // chunk and rejects. Checking the joined line instead produced a FIRST LINE
  3413. // of exactly the ceiling — inside the per-line bound, so it passed as a junk
  3414. // frame — and left an 8 MiB residual well under the bound, so the breach was
  3415. // never reported: measured, the run settled as
  3416. // `python exited (code=0, signal=null) before completing` after the host had
  3417. // held the ceiling AND copied it, which is the doubling this check prevents.
  3418. const ceiling = 256 * 1024 * 1024
  3419. const { runtime } = await setup({ maxWallMs: 60_000, addressSpaceMb: 2048 })
  3420. const result = await runtime.run({
  3421. program: [
  3422. 'import os',
  3423. 'chunk = b"A" * (8 * 1024 * 1024)',
  3424. `for _ in range(${ceiling / (8 * 1024 * 1024)}):`,
  3425. ' os.write(3, chunk)',
  3426. // One drain loop: a single os.write past the pipe buffer returns short,
  3427. // and a truncated tail would change which bytes cross the ceiling.
  3428. 'view = memoryview(b"\\n" + b"B" * (8 * 1024 * 1024))',
  3429. 'while view:',
  3430. ' view = view[os.write(3, view):]',
  3431. 'return "never"',
  3432. ].join('\n'),
  3433. bindings: [],
  3434. })
  3435. expect(result.value).toBeUndefined()
  3436. expect(result.error?.kind).toBe('worker-exit')
  3437. expect(result.error?.message).toContain(`protocol frame exceeded ${ceiling} bytes`)
  3438. }, 120_000)
  3439. })