Skip to content

feat(sieve): C9 first evidence -- gen-c output determinism, 12 specs identical (Closes #6517) - #6754

Merged
gHashTag merged 1 commit into
masterfrom
sieve/c9-reproducible
Oct 6, 2026
Merged

gHashTag merged 1 commit into
masterfrom
sieve/c9-reproducible

Conversation

@gHashTag

@gHashTag gHashTag commented Oct 6, 2026

Copy link
Copy Markdown
Owner

What

Sieve filter C9 (reproducible builds) is UNKNOWN for t27c in specs/compiler/theory/sieve.t27 (epic #6488). This PR records the first piece of evidence collectible with t27c alone: output determinism of t27c gen-c.

The run

  • Binary: ./target/release/t27c, t27c 0.4.0, built from master d5b046b00 (cargo build --release -p t27c, 2026-10-06). Host: Darwin 25.5.0 arm64.
  • Method: t27c gen-c <spec> > out.1.c then > out.2.c, then cmp. All 12 specs named by sieve C9: record gen-c output determinism of t27c on 12 specs #6517 exist on that commit, 0 skipped.
  • Result: 12 of 12 byte-identical. Sizes recorded in the spec (sieve 14156 B .. gf16 110604 B).
  • Negative control: gen-c of specs/queen/priority.t27 vs specs/compiler/theory/sieve.t27, same binary: cmp exits 1 (differ at char 109, line 2) -- the 12 identicals are not a cmp artifact.

What this is not

Output determinism of ONE t27c binary run twice on the same host. Not a bit-identical build of t27c itself: that needs two independent builds of t27c from the same source, and it stays open (DDC #6508 is the stronger form). The spec says so explicitly.

Checks (all run on this branch)

  • t27c typecheck specs/compiler/theory/reproducible.t27 -> Typecheck OK (0 errors, 0 warnings)
  • t27c gen-c of the new spec twice + cmp -> exit 0
  • t27c test-report -> 2 tests, 2 pass, 0 FAIL, 100%
  • grep -c '^test ' -> 2
  • git diff --name-only origin/master...HEAD -> only specs/compiler/theory/reproducible.t27
  • gitleaks: no leaks found

Only .t27 files in the diff (owner hard rule 2026-10-05).

Closes #6517
Refs #6488

🤖 Generated with Claude Code

…identical (Closes #6517, Refs #6488)

Record the run in specs/compiler/theory/reproducible.t27: one t27c 0.4.0
binary built from master d5b046b, gen-c twice per spec on the 12 specs
named by the issue, cmp byte-identical on all 12, negative control pair
(priority vs sieve) differs at char 109. All 12 specs exist, 0 skipped.
This is output determinism of one binary, not a bit-identical build of
t27c itself; that stays open (DDC #6508 is the stronger form).

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
@github-actions

github-actions Bot commented Oct 6, 2026

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-10-06 12:06:45 UTC

Summary

Status Count
Total Open PRs 50
PRs with Failing Checks 40
PRs with All Checks Green 10
READY 0
FAILING 40
PENDING 0
NO CHECKS YET 0

These columns do not partition: 0 + 40 + 0 + 0 = 40, and there are 50 open PRs. A PR is being counted twice or not at all.

Seal Status

  • ⚠️ STALE -- sha256(compiler.rs)=3c78f3c7ffb7 != 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).

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.

sieve C9: record gen-c output determinism of t27c on 12 specs

1 participant