Skip to content

Delete pre_build's two dead failure-tolerant sites (audit step 1)#158

Merged
Jammy2211 merged 1 commit into
mainfrom
feature/pre-build-dead-lines
Jul 16, 2026
Merged

Delete pre_build's two dead failure-tolerant sites (audit step 1)#158
Jammy2211 merged 1 commit into
mainfrom
feature/pre-build-dead-lines

Conversation

@Jammy2211

Copy link
Copy Markdown
Collaborator

Summary

Step 1 of the design merged in #157 (docs/pre_build_failure_audit.md; issue #156, campaign #155): delete the two dead failure-tolerant sites in pre_build.sh, measured no-ops before deletion:

  • the root-glob staging line (git add -- *.py *.md … || true) — exit 128 / staged nothing in 13/13 repos (six of its globs match nothing in any repo, poisoning the whole pathspec);
  • the README version-bump block + its readme_pkg parameter and all 13 call-site args — its sed edit was never staged (dead glob line), the runner-side bump was removed under feat: release stamping slim — wheels+tags only, no daily bump commits (R3-core) #120, and the pins it targeted are proven stale on origin. The banner printed a false claim on every release; README-pin ownership is a Phase 4 decision (release_version_sync_back_to_main.md).

Doc syncs ride along: the pre_build skill's step description and stale "README version-bump is handled inside pre_build.sh" note; docs/internals.md's stale "dataset force-added with git add -f" claim (pre-#150) and root-level-files row (now owned by the runner).

Behaviour on main is unchanged by construction: both deleted sites provably staged/produced nothing that ever landed. The only local difference is that READMEs no longer get dirty-edited-then-dropped on each release.

Steps 2 (own the runner's git add -A sweep) and 3 (shrink pre_build to its genuine product) follow as separate planned PRs.

Shipped under the human-authorized corrective-PR exception for Heart RED (standing authorization this session; "merge and continue" instruction 2026-07-16). PR-open only; merge stays human.

API Changes

None — release-script internals and docs only. run_workspace loses its unused 5th positional arg (readme_pkg); all call sites updated in the same file.

Test Plan

  • bash -n pre_build.sh — syntax OK.
  • Full PyAutoBuild suite: 113 passed (includes the test_pre_build_skill drift guard, which parses pre_build.sh's run_workspace lines and $PYAUTOBASE deps).
  • No-op proof carried over from the merged audit: per-repo git add -n exit-code matrix (13/13 exit 128 on the glob line); origin-verified stale pins showing the sed's output never landed.

Generated by the PyAutoLabs agent workflow.

Per the audit merged in #157 (docs/pre_build_failure_audit.md): the root-glob
staging line was a measured no-op in all 13 repos (unmatched globs poison the
whole pathspec, exit 128 swallowed by || true), and the README version bump's
sed output was never staged while the runner-side bump was removed under #120
— the banner printed a false claim on every release. Both sites deleted along
with the unused readme_pkg parameter and its 13 call-site args; skill doc and
internals.md synced (including the stale pre-#150 "git add -f dataset" claim).
Behaviour on main unchanged by construction: neither site produced anything
that ever landed. README-pin ownership moves to campaign Phase 4 (#155).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
@Jammy2211 Jammy2211 added the pending-release Merged PR awaiting inclusion in the next release build label Jul 16, 2026
@Jammy2211 Jammy2211 merged commit a6c9ec8 into main Jul 16, 2026
@Jammy2211 Jammy2211 deleted the feature/pre-build-dead-lines branch July 16, 2026 13:08
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

pending-release Merged PR awaiting inclusion in the next release build

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant