[Fengzdadi]

Which issue or RFC does this PR close?

No issue was filed. This fixes a CI failure introduced by #2248: since bad4c0c, upgrade-acceptance / Prepare upgrade artifacts and matrix fails on master and on pull requests, and Upgrade release gate fails after it.

Rationale for this change

#2248 made scripts/parallel_upgrade_release_gate.py identify the checkout by PC_UPGRADE_SOURCE_SHA, falling back to GITHUB_SHA. .github/workflows/upgrade-release.yml exports PC_UPGRADE_SOURCE_SHA at workflow level, so the variable is set in the step that runs the release gate tests.

test_parallel_workers_reject_changed_preparation_inputs[checkout] models a worker on another checkout by changing only GITHUB_SHA. While PC_UPGRADE_SOURCE_SHA is set, the identity the gate reads does not change, so the case fails with DID NOT RAISE GateError.

Only that job fails because only the upgrade workflow sets the variable. Locally and in the tests (3.12, unit) job it is unset, the gate falls back to GITHUB_SHA, and the case passes. The gate behaves as #2248 intends; the test depended on the environment of the job that ran it.

What changes are included in this PR?

• The parallel_reports fixture unsets PC_UPGRADE_SOURCE_SHA before preparing, so the parallel gate tests model an unpinned checkout in every environment. The checkout case still changes GITHUB_SHA, which is now the identity the gate reads wherever the test runs, so the GITHUB_SHA fallback stays covered.
• The pinned path stays covered by test_manual_release_evidence_tracks_target_checkout_not_workflow_commit from ci(release): allow explicit publication without upgrade acceptance #2248, which sets both variables explicitly.
• The gate script and workflows are unchanged.

Are there any user-facing changes?

No. This changes one test fixture.

How was this change tested?

• Reproduced the failure on unmodified master with GITHUB_SHA and PC_UPGRADE_SOURCE_SHA set to the same commit, as in a push or pull request run: uv run pytest -q tests/test_upgrade_release_gate.py -k test_parallel_workers_reject_changed_preparation_inputs fails the checkout case.
• With this change, the three cases pass with neither variable set, with both set to the same commit, with PC_UPGRADE_SOURCE_SHA set to a different commit than GITHUB_SHA (a pinned source_ref), and with only PC_UPGRADE_SOURCE_SHA set.
• Mutations of scripts/parallel_upgrade_release_gate.py, reverted afterwards:
  • without the source SHA comparison, the checkout case fails with and without PC_UPGRADE_SOURCE_SHA set;
  • without the GITHUB_SHA fallback, the checkout case fails;
  • when the gate ignores PC_UPGRADE_SOURCE_SHA, test_manual_release_evidence_tracks_target_checkout_not_workflow_commit fails.
• uv run pytest -q tests/test_upgrade_release_gate.py tests/test_upgrade_publication.py tests/test_upgrade_schema_equivalence.py, the files the job runs, passes with and without the variables set. On macOS, 6 supervisor tests skip because they need Linux subreaper support.
• make check
• I did not run the upgrade workflow itself. The Prepare upgrade artifacts and matrix job on this PR is the direct check.

AI usage statement

Claude Code (Claude Opus 5.5) was used to diagnose the failure from the CI logs, write the change, run the checks above, and draft this description. I reviewed the change.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota.

An organization owner can choose the repository organization as the billing source for reviews requested by members with a paid Copilot license. Learn about Copilot code review billing.

[hidb4ai]

LGTM