Repository navigation
Keep default Unity artifact processing within the existing two-minute cap #689
Description
Resolution
Resolved in the current draft PR #619 source at c9343908. The independently admitted study, local worker/failure controls, exact five-file integration and fresh supported qualification satisfy this issue's declared acceptance. Final evidence and scope. Shipping #685, compiler #687 and release readiness remain separate and open. This closure does not claim a merged or published release.
Current qualification
The independently frozen default-only study qualifies, and its five production files are integrated at 57da0228. Fresh supported Main validation passes all eight standard workflows, all four default Unity versions and both performance scopes. Fourteen archives retain 120 raw members; all six licensed cleanup gates pass. Default processing takes 36/66/66/66 seconds under the unchanged 120-second cap. Consolidated evidence.
Current c9343908 also passes all eight standard workflows, all four default Unity versions and both performance scopes. Its processing times are 38/66/66/66 seconds under the same 120-second cap. Shipping #685 remains a separate inconclusive acceptance study. The original failed parent and the prospective protocol below remain retained. PR #619 remains draft for remaining release gates.
Observed failure
PR619 f213317a410b58d1cd2190e082df045e72557fbe, Unity Tests run40558596655, Unity2021 job122381302513: all three default modes and their verifiers pass (Editor1013PASS/15existing skips, Mono27793PASS, IL2CPP27789PASS). The artifact-processing action runs from18:27:47.618 to18:29:47.468 and reaches its existing120-second cap. All uploads skip, all four license/lock cleanup steps succeed. There are no raw artifacts for this job, so console counts do not qualify current-head correctness. Keep the original failed job; no retry or timeout increase.
Prospective default-scope investigation
Question: can the three independent ordinary artifact roots complete inside the same action cap when the existing unchanged CLI processes them concurrently? The already prepared #685 six-file shipping candidate0692 is INCONCLUSIVE/STOPPED for its full shipping acceptance because of #687; that verdict and its20-cell/native-payload gates remain unchanged. This new default failure justifies a separately declared default-only evaluation, not a favorable reinterpretation of the shipping study.
Prepare an isolated descendant of f213. Reuse only the optional bounded four-worker action and real action-child tests, keeping sequential behavior as the default. The ordinary Unity workflow may opt into concurrency for its three explicit uploaded roots only. Shipping, incremental and consumer selections retain the original single broad root and sequential mode. Processor code, matching rules, file-size/other budgets, source/tests/profile inputs, both action/job caps and all sealing/upload/error/cleanup gates stay byte-identical.
Before native admission: prove complete default path coverage, disjoint roots, all-worker completion and failure if any child fails; retain missing/duplicate/nested-path controls and sequential behavior. Compare full output inventories/bytes on isolated copies of retained ordinary data; label local timing as local/previously processed rather than reconstruction of the absent failed packet. Run practical workflow/PowerShell/format/static checks and full script tests for the changed composite action. Freeze source, case inventories and analysis before new data.
ONE research dispatch on the actually failing2021 endpoint, complete default Editor/Mono/IL2CPP scopes, may follow after existing f213 Main jobs are terminal with cleanup confirmed. The pre-data order amendment defers #680 until default artifact retention and fresh MainCI qualify; neither target study had run when this order changed. Expected1028/27793/27789 cases; all372configuration regressions/controls and prior outcomes must qualify. Require original120-second action cap, all three raw archives/member hashes, complete source/input/process identity and all four cleanup steps. Stop as inconclusive on missing artifacts or another defect; reject if a single scope exceeds the unchanged cap or failure can pass. No identical retry, artificial waits, processor changes or post-data threshold changes.
If this independent default-only qualification succeeds, integrate only its declared files into existing draft619 and require all four supported default versions plus standard checks at the new Main head. The old #685 shipping candidate/full acceptance and #687 stay unresolved; no shipping, cost, throughput, parity or release-readiness claim follows. Evidence: .artifacts/session482/default-artifact-processing/.
Activity
wallstop commentedon Oct 27, 2026
Before implementation, the smallest default-only routing shape is fixed: keep the existing
paths: .artifacts/unitysequential invocation byte-identical. Add one optional newline-separated concurrent-path input (empty by default). Only ordinary selections populate it with the three explicit per-version EditMode/PlayMode/standalone roots; shipping/incremental/consumer selections populate no concurrent paths and execute the original broad-root CLI invocation. This avoids changing shipping path coverage or adopting #685's unqualified shipping workflow. The existing bounded worker algorithm and actual-child cases may be reused, with controls for the optional-input routing and exact ordinary upload-root coverage. No processor or matching-module changes.wallstop commentedon Oct 27, 2026
The original f213 Unity2022 job122381302558 also reaches the unchanged120-second artifact action timeout (120091ms), after Editor1013PASS/15existing skips, Mono27793PASS and IL2CPP27789PASS are reported and verified. It uploads nothing; preserve this second UNQUALIFIED job. Main6.3/6.5 continue without retry. The separate earlier c7f research2021 packet processed in80seconds on its selected runner; that passing result does not explain the two Main timeouts.
The default-only action/workflow candidate's real child controls and20workflow checks pass. On TWO isolated copies of previously processed qualified68fe2021data, sequential and concurrent action execution preserve all13files and bytes exactly. Local times11.97357s/6.294005s are ARM64/previously processed-input observations, not Windows cap compliance or a matched reconstruction of the absent f213files. Full script tests are active. A direct JS formatter expanded an unrelated pre-existing assertion line and failed the unchanged line-budget check; that formatting will be removed after the active process ends, then relevant checks repeated. No budget, timeout or processing-rule change.
wallstop commentedon Oct 27, 2026
Local candidate qualification is complete. Production-only commit
57da0228a612642fd8f13c109c552023d03920aachanges exactly five action/workflow/test files. Research descendant6581c39e7d8ed4f4e75ebc9d7414251d527f313fadds only the declared2021matrix projection; that projection must not integrate. Runtime, Editor, all Unity tests, Samples, processor and matching module have exact f213Git identities. Empty concurrent input preserves the original sequential CLI/broad-root behavior for shipping/incremental/consumer selections.All1,464Node/twoPython tests pass, zero failures/skips,457.68s. The final affected PowerShell harness passes independently in226.05s after removing only the unrelated formatter expansion; the actual locked Babel parser proves full-tested and final JS AST equality (excluding positions/comments/parser metadata). Validation, spelling, workflow checks, actionlint, actual YAML config and normal hooks pass. No budget increase. Local source/evidence receiptSHA
8a47b664e3124b5e4a571d293cf19b34fcaf049f09bdb633cb28c8f5449185e5; initial formatter/config/parser-offset preparation errors remain recorded. Thirteen retained-corpus files remain byte-identical across sequential/concurrent/original data; local timing limits remain.Frozen native admissionSHA
d10f5a5c43c33cab13b2c8cc9c8987a63bd3e2a88b15e86b4985f2c5497fbf31:1,888physical source hashes, analysis and prior2021XML hashes, exact case totals1028/27793/27789, original372configuration classifier rules and unchanged120-second cap. ONEdefault2021dispatch only after all f213Main jobs AND the already admitted #680temporal job finish and confirm cleanup. No native observations yet. Missing artifacts, cap/control/source/assertion/cleanup defects stop inconclusive; no identical retry. Only if qualified, copy the five production files from57da into draft619 and require fresh all-four-version MainCI. Old685shipping acceptance remainsINCONCLUSIVE/open.wallstop commentedon Oct 27, 2026
PRE-DATA dispatch-order amendment: neither6581default candidate nor07ebtemporal candidate has been dispatched. Actual f2132021/2022/6.3 now all fail artifact preparation and retain no raw archives. The #680temporal experiment therefore defers until this transport defect is qualified; its source/cases/classifiers stay frozen for a later independently admitted descendant. The default2021candidate6581may run ONCE after all currentf213Main jobs are terminal with cleanup confirmed, without waiting for the deferred temporal job. Source/analysis/baseline hashes,120-second cap, three raw archives,372configuration checks and all correctness/cleanup gates remain unchanged. This changes only pre-data execution order to avoid a known broken retention path; it does not alter #685's stopped shipping verdict or reduce this study's acceptance.
wallstop commentedon Oct 27, 2026
The ONEadmitted default2021dispatch is accepted once (HTTP204,19:19:56UTC) as run40568130784, exact frozen source
6581c39e7d8ed4f4e75ebc9d7414251d527f313f. Before dispatch, all8f213Mainworkflows were terminal, all6licensed jobs confirmed all4cleanup steps, live PRhead remained f213, candidate had zero prior dispatch runs, and all source/analysis/baseline hashes matched. The pre-data order amendment is applied; #680temporal remains deferred/not executed. Native result pending; no redispatch or Main integration.wallstop commentedon Oct 27, 2026
ONE-run verdict: QUALIFIED_DEFAULT_ARTIFACT_CAP_AND_FULL_SCOPE at6581c39e, run40568130784/job122414870878ELI. The original120-second action completes in 38seconds; all3archives/13members retain exact metadata/digests/extracted bytes. Editor1013PASS/15same skips; Mono27793PASS; IL2CPP27789PASS. All prior case/outcome inventories and372configuration cases qualify, including48original cases with10valid cycles each. Process settings and all4cleanup steps pass. Final frozen-classifier receiptSHA
2b4903bf536e98ee10b31c978e9bbe5d408bf20280e5c39ab5c6e183f8b47393under.artifacts/session482/default-artifact-processing/.This is cap/correctness qualification on the selected ELI endpoint, not a matched DAD speedup estimate or attribution of its historical timeout rate. The original f213all-four failures remain retained. StudyCOMPLETE/STOPPED; no identical retry. The production workflow matches every parsed research node except the declared three exclusions; only five production files from57da0228may integrate into draft619, with all four supported default versions still requiring fresh MainCI. Shipping/incremental/consumer remain on their original sequential path; old685fullshippingacceptance staysINCONCLUSIVE/open. No processor, Runtime, Unity-test, cap, assertion, policy, native-cost or release-readiness change.
wallstop commentedon Oct 27, 2026
Fresh production Main57da now qualifies the2021and2022default scopes with six archives/26raw members. Exact prior outcomes and372configuration cases pass in both Runtime modes for each version, all process/cleanup gates pass.2021job122420228187ELI processes in36seconds;2022job122420228290DAD processes in66seconds, both under the unchanged120-second cap. ReceiptsSHA
89db794601934e96db06ad20008cbe6518ccc7094e446424e1438efe6dc10651and3410860092542cf89dea84e3059e7491ebf879366e8aed4db661ea7d88d6e791. This includes the previously failing host, but is not a paired timing/causal speedup estimate.6.3/6.5and remaining performance/standard work continue; full Main qualification and issue closure remain pending.wallstop commentedon Oct 27, 2026
Acceptance is complete for this default-only defect. Reverified all eleven local evidence hashes and all five integrated source blobs, the one-shot native study, and all four supported default versions/full standard checks at current
c9343908. All three raw scopes remain retained per version, processing stays within the original120seconds, and cleanup succeeds. Final current-source qualification. Closing as resolved in draft619 source. The failed parent remains retained; #685 shipping and #687 compiler acceptance remain open. No merged release, cost, throughput or readiness claim.