Skip to content

fix(ci): avoid signing metadata pipe failures - #58

Merged
Dunky13 merged 1 commit into
mainfrom
fix/intel-core-ci-timeout
Oct 30, 2026
Merged

Dunky13 merged 1 commit into
mainfrom
fix/intel-core-ci-timeout

Conversation

@Dunky13

@Dunky13 Dunky13 commented Oct 30, 2026 •

Copy link
Copy Markdown
Owner

Summary

Fix the remaining current-main CI failure: quiet grep can close its input before printf finishes, so pipefail rejects valid signing metadata. Feed the same team, Developer ID authority, timestamp and hardened-runtime checks with Bash here-strings.

- printf metadata | grep -q predicate
+ grep -q predicate <<< metadata

#57's final-head CI passed all 20 jobs. Its tree equals main at 367836835eeb419fc3d523ebb9231c17690ee334. Main run 41105282352 subsequently passed every runtime partition, both native jobs and universal packaging, but failed ARM tooling shard 2 with printf: write error: Broken pipe in TestMacOSResumeRetainsExactArtifactsAndIDsAcrossAttempts. This follow-up repairs that separate failure. The historical Intel core timeout needs no additional timeout or coverage change.

Evidence

  • Before: the new real-helper regression rejects valid code and disk-image metadata on the base helper under Bash 3.2 and pipefail.
    After: both valid cases pass; all four missing-field cases fail. Existing wrong-Intel-team, timestamp/runtime, artifact custody and publication-refusal tests remain passing.
  • Local: both complete tooling race shards PASS (276.918s / 202.890s); focused installer, ownership and real-helper lifeline race tests PASS; full native suite PASS, including 118 supervisor checks and both architecture builds. Tooling vet, source boundary, ShellCheck, Bash syntax, formatting, whitespace and signature/DCO policy PASS. Two release-app Homebrew acceptance tests skipped because no final notarized app was supplied.
  • Independent: Codex gpt-6.1-sol, XHigh, read-only Standards and Spec validation found no actionable findings on bc457b5b158f38763c11c3a5bb62627352ab2b5a, tree d0f6f9f0b1cf9cf37945a0133fdc152a67d6487c, full diff SHA-256 83b537ebec9bd64e7532286a5da4271cf605fb0e114c30c530e50b97ef32a889. Independently passed 16 focused tooling race tests, 30 startup/ownership tests across three packages, static/policy checks and an isolated base/candidate contrast.
  • Hosted: run 41109536907, attempt 2 is successful on exact head bc457b5b158f38763c11c3a5bb62627352ab2b5a: all 20 effective checks pass, including ci-required. Attempt 1 had a post-SIGKILL cold authentication timeout in Intel HTTPS. The unchanged complete lifecycle race test passed locally, then the targeted same-head Intel HTTPS retry and dependent aggregate passed. No source, deadline, bcrypt, race or coverage change was made for that retry. Cold race-instrumented bcrypt remains a leading explanation, not a proven hosted cause.

Reproduction and continuation details: CI signature verification handoff.

Merge Danger

Door: two-way, a narrow shell-input change with no state migration.

Blast Radius: signing metadata verification. Matching predicates, per-slice codesign verification, timeouts, required matrices, ownership checks and release gates are preserved. Real Service Management, installed-app/Intel acceptance, Developer ID signing and Apple acceptance remain separate gates. No merge, release, fresh Apple submission or deployment is included.

Signed-off-by: Marc Went <marc@went.io>
@Dunky13
Dunky13 merged commit 7b4b27c into main Oct 30, 2026
38 of 40 checks passed
@Dunky13
Dunky13 deleted the fix/intel-core-ci-timeout branch October 30, 2026 19:49
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.

1 participant