Repository navigation
Repair the one call site RUNNER-WIDTH-1 left behind, which reds every PR's floor - #11871
Merged
Merged
Conversation
… PR's floor #11762 replaced the Int-valued gunbc_runner_slots_per_host with the RunnerWidthResolution answer and DELETED the Int, which is the right replacement-migration move: two answers to one question is the fork DESIGN 3 exists to prevent. It migrated every production call site and missed one witness. dag/test/claim/runner/runner_host_file_converge_witness_test.dag still imports and calls the deleted name, so the required floor's declarations phase refuses IMPORT-MEMBER-ABSENT on main itself. That refusal is not about any diff: it reds the floor lane for EVERY open pull request, and the witnesses lane fails derivatively behind it. The repair is the shape build_cache_endpoint_observe_test already uses, not a new one -- match the resolution and take the derived slots. The unresolved arm answers `0 - 1` rather than `0` deliberately, and that is the whole lesson of the receipt this migration was named for: zero is the IDENTITY of the projections downstream of a width, so an unresolved width rendered as zero is read by every consumer as its own grounded answer. A negative sentinel cannot be mistaken for one. Only the type's variants are imported; the RunnerWidthResolution type name is not referenced here, so it is not imported. Evidence: the declarations refusal names this file, this line and this member; the floor log for run 35516936949 carries exactly one declarations FAIL and this is it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This is a main-side break, not a defect in any open PR. The required
floorlane refuses on main itself, andwitnessesfails derivatively behind it, so every open pull request is currently blocked by it.#11762 replaced the
Int-valuedgunbc_runner_slots_per_hostwithRunnerWidthResolutionand deleted theInt— correct under DESIGN §3's replacement-migration rule, since a survivingIntwould be the attractor. It migrated every production call site and missed one witness:The repair uses the shape
build_cache_endpoint_observe_testalready migrated to, rather than inventing one. The unresolved arm answers0 - 1rather than0on purpose — that is the point of the RUNNER-WIDTH-1 receipt this migration is named for: zero is the identity of the projections downstream of a width, so an unresolved width rendered as zero reads to every consumer as its own grounded answer. A negative sentinel cannot.Only the variants are imported;
RunnerWidthResolutionitself is not referenced, so it is not imported.Scope check: the module
gunbc.runner_slot_allocationis fine — it moved file but kept its module name, so its ~70 importers are unaffected. Exactly one file referenced the deleted symbol, at two sites, both in this diff.Verification: the floor log for run 35516936949 carries exactly one
declarations FAILand this is it. I could not re-run the resolve locally to green — this session'sgunbcbinary predates thehmac_sha256_hexinterpreter primitive and reports it as a source error in files untouched here — so CI is the check.cc: #11845 is open against this same file and is blocked by this same red.
🤖 Generated with Claude Code