Retire the expected failure on the C++ macro-expanded call - #166
Retire the expected failure on the C++ macro-expanded call#166DavidBakerEffendi wants to merge 1 commit into
Conversation
The annotation recorded that reverse navigation from the PARITY_CALL macro argument did not resolve to the expanded_label definition. Bifrost v0.10.5 resolves it to the authored include/parity.h:45:20, so the claim is stale and the case is now an ordinary pass. Measured over the 110-case balanced core: under v0.10.1 the usage-to- declaration lookup returned no_definition and the case reported expected_failure; under v0.10.5 it returns ok at the expected target and the runner reports improved, which is exactly the signal that an expectation has outlived the defect it described. The authored contract does not change. The declaration, expected usage, and lookup are untouched -- only the assertion about what Bifrost cannot do goes away. The review notes keep the retired reason and the evidence for retiring it, so the record survives rather than the annotation being deleted silently. Verified: validate passes 61 case files, and cpp-lsp-parity now runs 6 passed, 0 improved, 0 expected failures, with the pre-existing cpp-parity-using-alias-constructor gap unchanged. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
Closing — this approach is forbidden by the promotion tier, and CI was right to
That is not an inconvenience to work around. From
The cohort was frozen in #124. The selection being unpublished does not make it I proposed this in #164 without checking whether the case file was hash-bound, |
1 participant
Closes #164.
Bifrost v0.10.5 navigates
cpp-parity-function-like-macro-expanded-callcorrectly, so the annotation asserting it cannot is stale.
expected_failureimprovedno_definition, no targetok→include/parity.h:45:2045:20is exactly theexpectedDeclarationthe case authors.improvedisthe runner's signal for a case that passes while annotated as an expected
failure — the corpus was already flagging this rather than hiding it.
What changes
Only the
expectedFailureblock. The declaration, expected usage, and lookupare untouched — the authored contract is unchanged, and what goes away is the
claim about what Bifrost cannot do.
The retired reason and the evidence for retiring it are recorded in the case's
verification.notes, so the history survives in the corpus rather than beingdeleted with the annotation. This is reviewed ground truth; it should say why
it moved.
Testing
validatepasses 61 case files.cpp-lsp-parity.yamlagainst v0.10.5(
d775e3b4) now reports 6 passed, 0 improved, 0 expected failures, withthe pre-existing
cpp-parity-using-alias-constructorgap unchanged.Measurement context in #160. Note that was a native host run — reports
self-label
host_specific— so the reference-container run in the freeze isthe authority.
🤖 Generated with Claude Code