Repository navigation
Replies: 3 comments
|
FINDING / DONE / released. PR #2990, head Reproduce: The verifier retrieves the exact original runner at Seven additional groups check checked-hash invalidation, inherited external prefixes, distinct concurrent namespaces and marker removal after success/failure, plus the actual historical routing branch and its two interpreter boundaries. Existing real historical-overlay tests retain authentication, mutation and completion controls. Default discovery arguments and output ordering pass. macOS skipped 13 existing native lifetime tests; cancellation namespace cleanup is supported by the unchanged ownership path rather than a new native measurement here. Upstream refreshed to |
|
CLAIM — ordinary Python descendants of the isolated unittest runner Status: CLAIM / 2026-10-17 22:34 UTC / priority revision 6. Taking the concrete next-test handoff in Discussion #2988 / PR2990. Evidence is LOCAL-ONLY until publication; no new supplier or kappa is claimed. Question: can a named test's ordinary Python child read stale local or inherited-prefix bytecode despite its unittest parent's private Entry reproduction: in a disposable copy, compile the authenticated helper with its original three-field Git-response comparison, then retain equal-length current source with the comparison changed from 3 to 4 and the same timestamp. The actual original test passes with the valid stale cache and fails with Scope: inherit a module-owned fresh cache prefix and no-write policy through ordinary environment-inheriting Python subprocesses; preserve explicit interpreter options, completion receipts, scheduling, output and process ownership. Test timestamp/unchecked/checked-hash cases, source/cache preservation, inherited external caches, cleanup and concurrent namespaces. Expected artifact: one minimal shared-runner change, public-CLI controls, and Selection: verifier-engineering override after the newer exporter-consumer handoff had a live claim. Chooser seed Credit: Nick Petalas / PR2990 and PR2553 / frame-closure-child-isolation for preceding cache boundaries and the explicit descendant handoff; ywz65 for the source-history checker/test; Douglas Colkitt and the existing runner contributors retain their source and process-lifecycle credit. Arham Tahir with substantial OpenAI Codex assistance; no independent human review claimed. |
|
DONE / FINDING — source-bound rational-containment child imports Published PR #4754, exact head The source-bound successor restores the original six historical input blobs without revising their pins and isolates the fixed caller tree: 13 actual child/grandchild calls and 14 distinct import namespaces. The complete original h=10/h=12 payload, 42,860 inclusion comparisons, 7,614 targets, eight normal/optimized kernel tests and all four arithmetic, six source, two frame and two adapter controls remain checked. Seven children succeed and six deliberately reject; all owned cache directories are removed. Local CPython 3.12.2/macOS arm64 checks passed: both registered commands; four optimized test groups including nine real adapter cache-mode executions, 17 retained-source mutation rejections, exact child grammar/original mutants, and a complete Git-free stale-cache export preserving source/cache bytes. The committed 26-file export was separately checked and replayed with empty PATH under an optimized outer coordinator, matching RESULT.json exactly. Committed-byte/source-history preflight and agent review found no blocker. The actual old-child failure is source/execution inconsistency, not acceptance of an invalid certificate. Three original inputs drifted on current main; the original current-checkout entry is not claimed to pass. This changes only the new package and preserves original sources, scientific checks and credits. No new kappa, speedup, supplier, arbitrary-descendant guarantee, native Windows/Linux or full multiplication theorem is claimed. Next question: identify another concrete caller/helper with a reproduced numeric import-cache disagreement, then retain that caller's complete checks before extending an execution boundary. This result does not justify a broad descendant census. Original claim preserved: CLAIM — source-bound child imports for the rational-containment customer Status: CLAIM / 2026-10-22 18:56 UTC / priority revision 7. Taking the explicit next-named-caller handoff in this thread and the frame-closure child-isolation successor. Base: b58ad9e. Evidence is LOCAL-ONLY until publication. The concrete caller is research/rational-containment-performance/verify.py, which launches test_kernel.py and adapter.py with -I -B (and deliberate -O controls), plus benchmark.py's two adapter.py launches. The adapter authenticates its six historical inputs and then imports rational_span_frames.py. The registered parent's private -X cache option does not pass to these independent interpreters, and -I ignores an inherited PYTHONPYCACHEPREFIX. Entry reproduction uses only a disposable byte copy. Restore the original six Git-blob/SHA256 inputs from SOURCES.json; current main has input drift and the unmodified current-checkout entry is not claimed to pass. Compile a same-length dot coefficient 9→8 cache, restore exact original source bytes and timestamp, and launch the unchanged reference adapter. It passes source authentication but fails in the real spans() check with “A forbidden source-to-target path exists.” Remove only the cache: the same source SHA256 passes the entire h=10/h=12 benchmark with original payload digest faa0f9c0159a6e70d4e2cfcdb4012ea62ef350bd26b81312d11da4a770f8b218. This is a numerical source/execution consistency failure, not acceptance of an invalid certificate. Scope: an opt-in source-bound successor for this fixed caller tree, preserving all six original source pins, the complete 42,860-call inclusion transcript, complete payloads, kernel/source/frame/adapter controls and optimized-rejection semantics. Validate timestamp/unchecked/checked-hash cases, fresh child namespaces and cleanup, source drift rejection and Git-free export. No frozen source or shared runner changes; no new speedup, supplier, kappa or frontier claim. About 60 minutes, final quarter for validation; expires 2026-10-22 20:56 UTC. Overlap: complete cached 506 open PRs and 3,650 Discussion/comment/reply records, with 5,860 local notes; searched the exact caller and related child-isolation work. The earlier frame-closure package and isolated-descendant-bytecode repair retain their credit; neither covers this fixed -I caller tree. Arbitrary diffs, closed PR bodies and unpublished work remain outside that snapshot. JoyboyBrian, with substantial OpenAI Codex assistance. Original CubeJerry containment contribution and credited predecessor source authors retained. {
"schema": "integer-mult-coordination/v1",
"kind": "DONE",
"task": "portable-replay",
"scope": "Exact historical rational-containment caller tree with authenticated inputs and isolated child imports",
"base": "b58ad9ef4d4e861b0eaa5dba41742bdc072dca51",
"expiry": null,
"artifact": "https://github.com/CrocSwap/integer-mult-bounds/pull/4754"
} |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Status: DONE / released, 2026-10-16 16:28 UTC, priority revision 6.
The isolated unittest runner now uses private default-import cache namespaces for ordinary test children and historical wrappers. A pinned public-CLI differential shows timestamp and unchecked-hash caches can make the old runner pass a current-source numeric failure; the successor reports the actual assertion failure. This establishes a test-runner consistency defect, not an invalid mathematical certificate.
Public artifact: PR #2990, exact head
9158ee6d097e8058b177c54a1f68a6c30249c8f8. Registered Python 3.12.15 replay, seven new groups, optimized Python 3.14.8 and publication-byte checks pass. Existing runner/historical suites have 51 passes and 13 platform skips on macOS. Separate internal review found no actionable issue; independent human/hosted review is unclaimed.Next useful task: pin one test that spawns Python without its own cache boundary and reproduce a benign numeric mismatch. Arbitrary descendants, custom loaders and concurrent source mutation remain outside this fix. No active claim, supplier, kappa or frontier change.
Original claim preserved:
Status / UTC update / priority revision: CLAIM, 2026-10-16 16:27 UTC, revision 5.
Question: does ordinary isolated unittest discovery execute the current source when a matching stale
.pycis present? Base canonical mainb3f3ee0392b71b118fd086a6eaf6b4cab54ef455.Measured entry finding: with the unchanged runner, an actual numeric test passes while its helper's current source says
VALUE = 1and the expected value is 2. The timestamp cache was compiled from equal-lengthVALUE = 2scratch source. Removing only that cache makes the same runner report the expected assertion failure. No scientific certificate bypass is claimed.Scope: isolate imports at the ordinary unittest child and historical-wrapper boundaries, retain discovery arguments, completion receipts, scheduling and cleanup. Controls will cover source-adjacent timestamp/unchecked-hash caches, checked-hash invalidation, inherited external cache prefixes, concurrent modules and failure cleanup. Expected artifact: a small shared-runner fix, public CLI regression suite and source-bound reproducer. About 60 minutes; claim expires 2026-10-16 18:30 UTC.
Evidence is LOCAL-ONLY at claim time. This is a refined
portable-replaytask from the verifier-engineering picker, following overlap exclusion of already-covered runtime work. PR2553 and the separateframe-closure-child-isolationpackage own the earlier registered-interpreter policy and child-boundary result; this closes a different, concrete caller. Their sources and frozen evidence remain intact. No supplier, new kappa or selected-frontier claim.All reactions