Skip to content

spec(shootout): move the measured AOT baseline row out of the bench module - #13128

Merged
cursor[bot] merged 2 commits into
masterfrom
lane/corpus-cause-c
Nov 9, 2026
Merged

cursor[bot] merged 2 commits into
masterfrom
lane/corpus-cause-c

Conversation

@gHashTag

@gHashTag gHashTag commented Nov 2, 2026 •

Copy link
Copy Markdown
Collaborator

What

On #13121 the corpus ratchet names two parse-no-discard rows that are not in that PR's scope:

  • specs/tri/shootout/acceptance_matrix.t27
  • specs/tri/shootout/codegen_audit.t27

Both come from the same line 24, use tri::shootout::bench::MEASURED_T27B_BASELINE;, added by #13099. bench lexes as a keyword, so the use path stops at shootout:: and the parser drops : : MEASURED_T27B_BASELINE ;, 4 tokens per file (#11101, spec half #13113). No spelling of a path through a module named bench parses, so the row has to move.

This PR does not bless, does not raise a cap and does not touch either ledger.

Change

  • New specs/tri/shootout/measured_aot.t27. It is the one home of LanguageResult and MEASURED_T27B_BASELINE. Both are moved verbatim from bench.t27, with their measurement comment, and one test is added: calls per second times the runtime gives back the 200M logical calls, within 0.01 percent.
  • bench.t27, codegen_audit.t27, acceptance_matrix.t27 now read the row through use tri::shootout::measured_aot::MEASURED_T27B_BASELINE;. No assert changed.
  • The type moves with the constant, so the row keeps a single definition (FR-003 of spec(tri/shootout): baseline at a keyword use path #13113).

Checked locally

  • t27c parse-complete --specs-dir specs/tri/shootout: before, 13 files scanned and 2 with discards, 8 tokens; after, 14 files scanned and 0 discarded.
  • t27c use-edges: all three importers resolve to specs/tri/shootout/measured_aot.t27.
  • t27c check reports Typecheck OK on all four files. It reported OK before the change too, so this is not evidence that the import resolves.
  • No seal exists for the shootout specs, so there is nothing to re-save.

Not done

What this PR may turn red

  • t27b-native-ratchet, UNLISTED. The ratchet is red when the reference passes a spec the ledger does not name (unlisted in specs/tri/t27b/steward.t27). On master this check already lists 2 such rows from spec(t27b): L1 clamp gate, if-conversion conditions, noisy-row rule (Refs #12877) #13099: l1_clamp.t27 and ifconv_plan.t27. If the reference passes the new measured_aot.t27, it becomes a third row. A ledger row for it is a steward decision and is not in this PR.
  • bench.t27 keeps its t27b pass row. Whether t27b still passes it through the new import is not measured yet.

Refs #13113
Refs #12877

🤖 Generated with Claude Code


Note

Low Risk
Spec-only refactor to fix parser/import paths; no runtime, auth, or data-handling changes.

Overview
Fixes parse failures where use tri::shootout::bench::MEASURED_T27B_BASELINE drops tokens because bench is a keyword (#13113).

Adds specs/tri/shootout/measured_aot.t27 as the single home for LanguageResult and MEASURED_T27B_BASELINE, moved verbatim from bench.t27, plus a test that calls_per_sec × runtime recovers ~200M logical calls within 0.01%.

bench.t27, codegen_audit.t27, and acceptance_matrix.t27 now import the baseline via tri::shootout::measured_aot::; benchmark harness logic and existing asserts are unchanged aside from the removed duplicate type/constant definitions in bench.t27.

Reviewed by Cursor Bugbot for commit 4c95a20. Bugbot is set up for automated code reviews on this repo. Configure here.

…odule

`bench` lexes as a keyword, so `use tri::shootout::bench::MEASURED_T27B_BASELINE;`
drops `: : MEASURED_T27B_BASELINE ;` at parse in acceptance_matrix.t27 and
codegen_audit.t27 (cause C of the corpus ratchet red, added by #13099).

- New specs/tri/shootout/measured_aot.t27 holds LanguageResult and
  MEASURED_T27B_BASELINE, moved verbatim from bench.t27, plus one test.
- bench.t27, codegen_audit.t27 and acceptance_matrix.t27 import the row
  from measured_aot. No assert changed.

parse-complete on specs/tri/shootout: 2 discards (8 tokens) before, 0 after.
use-edges: all three importers resolve to measured_aot.t27.
Tests not run on the Mac.

Refs #13113
Refs #12877

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@github-actions

github-actions Bot commented Nov 2, 2026

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-11-02 13:54:18 UTC

Summary

Status Count
Total Open PRs 47
PRs with Failing Checks 44
PRs with All Checks Green 3
READY 2
FAILING 44
PENDING 0
NO CHECKS YET 0

These columns do not partition: 2 + 44 + 0 + 0 = 46, and there are 47 open PRs. A PR is being counted twice or not at all.

Seal Status

  • ⚠️ STALE -- sha256(compiler.rs)=e75b498e5e79 != manifest seal=87e5cbd3ad94.
    The committed NMSE numbers were certified against an older compiler.rs.
    Run scripts/reseal-check.sh locally for the two-step reseal command (advisory; not a merge gate).

This was referenced Nov 2, 2026
Co-authored-by: Dmitrii Vasilev <gHashTag@users.noreply.github.com>
@github-actions

github-actions Bot commented Nov 9, 2026

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-11-09 21:01:41 UTC

Summary

Status Count
Total Open PRs 50
PRs with Failing Checks 47
PRs with All Checks Green 3
READY 0
FAILING 47
PENDING 0
NO CHECKS YET 0

These columns do not partition: 0 + 47 + 0 + 0 = 47, and there are 50 open PRs. A PR is being counted twice or not at all.

Seal Status

  • ⚠️ STALE -- sha256(compiler.rs)=b882be1e4933 != manifest seal=87e5cbd3ad94.
    The committed NMSE numbers were certified against an older compiler.rs.
    Run scripts/reseal-check.sh locally for the two-step reseal command (advisory; not a merge gate).

@cursor
cursor Bot merged commit 1901f98 into master Nov 9, 2026
46 of 53 checks passed
@cursor
cursor Bot deleted the lane/corpus-cause-c branch November 9, 2026 21:15
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants