Repository navigation
Replies: 17 comments
|
FINDING / RELEASED — contribution 1/15: Handle malformed overlap cache sections without losing usable evidence. Malformed cache roots and source records now yield explicit missing-coverage warnings. A usable source survives a bad neighboring source, and offline evidence is not rewritten. Validation: 35 overlap tests pass normally and under Python 3.12 -O; original main crash reproduced; git diff --check. Base |
|
FINDING / RELEASED — contribution 1/15: Handle malformed overlap cache sections without losing usable evidence. Malformed cache roots and source records now yield explicit missing-coverage warnings. A usable source survives a bad neighboring source, and offline evidence is not rewritten. Validation: 100 overlap consumer tests pass on Python 3.12 normally and with -O; original main crash reproduced; git diff --check. Base |
|
FINDING / RELEASED — contribution 2/15: Bound overlap refreshes when GraphQL cursors stop advancing. Every previously unguarded GraphQL connection now stops on non-progress. Failure leaves the preceding complete cache source available, with an explicit refresh warning. Validation: 98 overlap-consumer tests on Python 3.12; 3 new tests under -O; git diff --check. Base |
|
FINDING / RELEASED — contribution 3/15: Exclude directory aliases from local overlap evidence. Local note search now rejects parent aliases and non-files through the existing path validator. A replaced tracked package directory produces a visible unsearched-HEAD-path warning. Validation: 99 overlap-consumer tests on Python 3.12; 4 new alias tests under -O; git diff --check. Base |
|
FINDING / RELEASED — contribution 4/15: Reject ambiguous coordination JSON before posting or reading claims. Contradictory duplicate keys in optional coordination metadata now produce an explicit error. The same parser protects both posted metadata files and cached Discussion blocks. Validation: 13 coordination tests; 4 new tests under -O; no-post failure control; git diff --check. Base |
|
FINDING / RELEASED — contribution 5/15: Require unambiguous, typed metadata in research export receipts. Offline receipts no longer pass through Python Boolean/integer equality or duplicate JSON keys. Valid export format and 40/64-character Git labels remain supported. Validation: 31 export tests; 5 new tests under -O; git diff --check. Base |
|
FINDING / RELEASED — contribution 6/15: Add an opt-in executable-mode check before package publication. The optional publication check now detects byte-identical scripts whose Git execute bit differs from the working file, even when core.filemode hides the change. Validation: 15 publication-preflight tests; 4 mode tests under -O; actual Git and CLI controls; git diff --check. Base |
|
FINDING / RELEASED — contribution 7/15: Explain recursive invoice comparisons with exact tariff witnesses. Contributors can now inspect exact tariffs and old/new costs when concave order is inconclusive. The report includes target credits, byte hashes and the explicit limit that mixed signs do not establish a power crossing. Validation: 15 invoice tests on Python 3.12; 625 signed profiles; 5 new tests under -O; isolated optimized CLI; git diff --check. Base |
|
FINDING / RELEASED — contribution 8/15: Certify recursive invoice comparisons at exact rational powers. An exact rational-power comparator now decides strict invoice signs that the concave-order screen can leave open. It uses only integer rounding and states when precision is insufficient. Validation: 8 tests normally and under -O; 216 exact power inequalities; 512-rank limit control; isolated optimized CLI; git diff --check. Base |
|
FINDING / RELEASED — contribution 9/15: Combine recursive invoices with exact multiplicities and source receipts. A contributor can assemble repeated component bills without losing productive target debits. The generated invoice feeds the existing comparator, while a separate receipt binds every source and output byte. Validation: 15 invoice tests on Python 3.12; 5 new tests under -O; byte-exact receipt and no-partial-output CLI controls; git diff --check. Base |
|
FINDING / RELEASED — contribution 10/15: Reject linked output paths before creating research drafts. The draft generator now validates its output namespace before reading sources or creating temporary files. A linked research directory cannot redirect the new package outside the checkout. Validation: 6 scaffold tests plus onboarding tests on Python 3.12; 3 alias tests under -O; external-file preservation controls; git diff --check. Base |
|
FINDING / RELEASED — contribution 11/15: Keep Unicode overlap excerpts centered on the actual match. Overlap excerpts now show the original matched text even when Unicode lowercase changes string length. ASCII excerpts and hit ranking are unchanged. Validation: 100 overlap-consumer tests on Python 3.12; 5 new tests under -O; original omitted-match example reproduced; git diff --check. Base |
|
FINDING / RELEASED — contribution 12/15: Preserve exact Git-path query case and whitespace across hosts. Git-path overlap queries now preserve case and spaces consistently. Empty input fails explicitly instead of producing an empty apparent search. Validation: 100 overlap-consumer tests; 5 new tests under -O; simulated Windows normalization and no-network invalid-query control; git diff --check. Base |
|
FINDING / RELEASED — contribution 13/15: Preserve overlap evidence when cache publication fails. Cache write failures now preserve the same previous evidence in the report and on disk, while removing partial staging files. A warning about old data now describes the actual returned snapshot. Validation: 100 overlap-consumer tests; 5 new tests under -O; partial-write, disk-full and replace failure controls; git diff --check. Base |
|
FINDING / RELEASED — contribution 14/15: Keep option-like paths literal in sparse materialization commands. Generated sparse-checkout commands now separate options from repository paths, so option-like directory names cannot change the command mode or trigger stdin parsing. Validation: 16 sparse-baseline tests; 2 new tests normally and under -O; real option-like directories; git diff --check. Base |
|
FINDING / RELEASED — contribution 15/15: Check declared replay inputs in offline research exports. Reviewers can now check that an offline export actually contains the files named by its bundled replay manifest. The optional check distinguishes missing declarations from valid byte-integrity receipts. Validation: 32 export tests on Python 3.12; 6 new tests under -O; coherent missing-input archive and Git-free CLI controls; git diff --check. Base |
|
DONE — PR #5761 repairs the priority-8 assignment-test failures with exact frozen legacy fixtures and all22 live-card checks. All12 focused and16 related tests pass; hosted full-plan status is pending. Canonical evidence and limits. The released Discussion5722 CI handoff is addressed; no production routing or frontier change. Scope released. |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
DONE / RELEASED — 15 independently reviewable contributor-tool contributions.
The donor requested smaller useful PRs for future contributors. These changes improve overlap checks, publication and replay integrity, and exact invoice comparison. They do not claim a new κ. No investigation scope remains reserved.
Each contribution started from the canonical main returned by a fresh checkpoint. All 15 source bases are
df0137efd35e11b4445734bf612bdd4c7ebc330c. Priorities and overlap were checked before each one. The contribution checkpoints used revision 7. The final refresh read revision 8 atde5ec2e7a6074068452f45b1a191b5c72832df4a; it keeps verification tooling and full signed-invoice comparison in scope. The selected record blob remains581f8d87008354eb5b869f7ed9de1317ad837ce8.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 15 PRs are now marked merged by the maintainer's separate queue integration. This contributor session did not merge or close any PR. Current main
c297a40e7732bc9452bbf745f9aad43455b1e330contains all of the changed code and tests. The only difference in the combined contribution file set is additional invoice-guide text from other contributors.The maintainer recorded shared-tool checks in queue A and queue B. Those receipts cover 164 and 221 affected-consumer tests respectively; queue B records five skips. They do not certify a new scientific claim.
Local verification: each PR passed its relevant tests and
git diff --check. All 15 patches applied together with no conflict in an isolated temporary index. Their 69 new tests passed together under normal and optimized Python 3.12. The same 69 tests passed again on the integrated current main with four byte-preserving test-file renames for CI discovery. The contributions change only tools, tests and documentation; they do not changeresearch/,certificates/orupstream/.The initial hosted plans for PRs #5723 through #5743 failed the same two existing current-status navigation checks (
reviewandsource_manifest); their package replay jobs were skipped. The repair in PR #5703 is now integrated. The hosted plan for PR #5744 passed; its replay job was skipped because no research package changed. The earlier failed jobs are historical results and are not relabelled as passing.CI registration is completed in the separate follow-up PR #5749, head
e3117af2cd7d1e5e473ee1ff73c94ebbe8daa16f. The maintainer integrated the original PRs before these final renames. This follow-up changes four names, with no content changes, to put their 22 tests under the existingtest_research_*.pydiscovery rule. It is not counted as a sixteenth independent research result.The full initial overlap cache covered PR bodies, Discussions and local notes. Before each contribution, fresh open-PR heads, file lists and recent Discussion activity supplemented that cache. This coverage is bounded and does not prove novelty. Claims and per-PR findings are retained below.
Limits and next use
The invoice tools operate on declared bills. A candidate still needs correct ownership, complete charges, source binding and an operator proof before it can support a κ claim. The rational-power tool can return an unresolved interval; its finite precision is not an equality proof. Composition does not establish physical compatibility between components.
Native Windows execution was not tested. POSIX mode checks are opt-in. The full mathematical supplier replay, Lean and Docker were not run for these tool changes. The selected conditional κ is unchanged.
Future contributors can combine the selected invoice exporter with these comparison tools and the common-base all-exponent comparison. Those are distinct contributions by their credited authors. A new candidate still needs a justified complete bill.
Prepared for Ali Romman with substantial OpenAI Codex assistance. Preserve the original comparator, research tooling, export tooling and mathematical source credits in each file and linked PR. The Discussion records author evidence; it does not replace independent mathematical review.
Additional local coverage: the full shared discovery command on the follow-up completed 526 tests with 5 failures, 16 errors and 29 skips (432.374 seconds). The logs show unavailable historical Git blobs, omitted sparse-checkout inputs, a denied
pscall, an inventory CLI timeout and a sparse current-assignment comparison failure. This run is not a full local pass; it is separate from the passing 69 contribution tests. The follow-up's hosted full-checkout plan is tracked in its PR.Hosted verification at head
e3117af2cd7d1e5e473ee1ff73c94ebbe8daa16f: the plan log records all 69 new contribution tests as passing. The full run completed 533 tests with 8 failures, 2 errors and 22 skips. Every failure/error is in the existingtest_research_current_assignment.pyafter upstream changed the catalogue to revision 8. The old tests still require 16 rewritten cards and legacy receipt fields; the revision-8 cards already use the current entry command. Two failures (0 != 16and the missing-helper expectation) reproduce from exact unchanged mainde5ec2e7a6074068452f45b1a191b5c72832df4a, without this rename patch. The plan is red and the replay job is skipped; the new tests passing does not make the full plan pass.The isolated inventory-coverage recheck passes all 5 tests (41.958 seconds); this resolves the transient 30-second timeout from the broad local run. Other broad-run sparse-checkout and sandbox limitations remain as recorded.
Remaining handoff: update the current-assignment tests for the live revision-8 catalogue while retaining explicit frozen legacy-routing fixtures, then rerun the shared plan. This is outside the four-file rename patch and does not require a frontier change.
All reactions