gates: the two required contexts name their own subject, not the wrong one - #3062
Merged
Conversation
…g one Closes #3061 Both are in docs/BRANCH-PROTECTION.md's required list, so each wrong message lands on a check nobody can merge past. NOW Sync Gate. `${PR_BASE_SHA:?}` catches unset and empty, not a non-empty SHA naming an object this clone does not have -- a shallow fetch, a force-push that orphaned the base, a rerun after the branch moved. `git diff` exits 128, and the `|| true` that exists to absorb grep's no-match absorbs that identically, so the gate printed "SYNC REQUIRED: this PR/push adds no docs/now/ entry" about a comparison it never made. Reproduced with a bad base SHA. It now verifies both revisions with `git cat-file -e <sha>^{commit}` and exits 2 when one is absent, naming the variable and the value; resolvable revisions produce exactly the verdict they did before. The push arm is fixed the same way. Issue Gate. workflow_dispatch carries no `pull_request` object, so PR_TITLE and PR_BODY render as empty strings, grep matches nothing, and the step printed "L1 TRACEABILITY violation: No issue reference found in PR title/body" against a pull request that does not exist. Not hypothetical: tools/check_now_entry_shape.py:185 already records the dispatch run that produced it. It now reads PR_NUMBER and, when there is none, says it examined nothing and exits 0 -- POSIX 1003.3 UNTESTED, no subject, which is different from an instrument that failed, and the message says it is not a pass for any PR. Found by a read-only fan-out over workflows, hooks, python gates and the Rust --gate commands: 41 candidates, 15 confirmed, 9 refuted, 0 agent deaths. These two are the only ones marked blocks-a-merge; the other thirteen are in #3061's closing note. Four mutants, four kills, and two are caught only by the controls: refusing every revision kills the resolvable-pair control, and taking the no-PR branch unconditionally kills both real-PR controls. The test extracts the Issue Gate's `run:` body from the workflow rather than restating it, so editing the YAML and not the test is caught -- and it asserts the extraction matched, because an extractor that stops matching would assert nothing at all. The control for the NOW gate compares HEAD against itself rather than HEAD~1: it needs no history, so it survives a fetch-depth 1 checkout instead of silently skipping there. A control that vanishes with the runner's clone depth is not one.
gHashTag
enabled auto-merge (squash)
September 3, 2026 20:44
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
This was referenced Sep 3, 2026
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
PR DashboardGenerated at: 2026-09-03 22:00:32 UTC
Summary
Seal Status
|
This was referenced Sep 3, 2026
Contributor
PR DashboardGenerated at: 2026-09-03 22:10:02 UTC
Summary
Seal Status
|
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
PR DashboardGenerated at: 2026-09-03 22:20:47 UTC
Summary
Seal Status
|
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
PR DashboardGenerated at: 2026-09-03 22:34:38 UTC
Summary
Seal Status
|
Both sides append a step at the same anchor in untrusted-input-gate.yml, so the conflict is structural rather than semantic and both steps are always right. This is the fourth time; the cheap fix is an ordering rule for the step list, noted rather than done here.
…b.com/gHashTag/t27 into w54-required-gates-name-their-subject
Contributor
PR DashboardGenerated at: 2026-09-03 23:06:10 UTC
Summary
Seal Status
|
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
This branch owns no skill section; master's copy carried the three conflict markers from #3072, so the repaired copy is used rather than propagating them.
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
PR DashboardGenerated at: 2026-09-03 23:15:14 UTC
Summary
Seal Status
|
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
PR DashboardGenerated at: 2026-09-03 23:30:26 UTC
Summary
Seal Status
|
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
PR DashboardGenerated at: 2026-09-03 23:55:59 UTC
Summary
Seal Status
|
…b.com/gHashTag/t27 into w54-required-gates-name-their-subject
Contributor
PR DashboardGenerated at: 2026-09-03 23:59:22 UTC
Summary
Seal Status
|
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #3061
Two of the five required contexts in
docs/BRANCH-PROTECTION.mdprinted a verdict about something they never examined. A wrong message on a required check is the worst place for one: it is the check a contributor cannot merge past, and it sends them to fix the wrong thing.1. NOW Sync Gate — a revision it cannot resolve read as a missing entry
${...:?}catches unset and empty, not a non-empty SHA naming an object this clone does not have — a shallow fetch, a force-push that orphaned the base, a rerun after the branch moved.git diffexits 128, and the|| truethat exists to absorb grep's no-match absorbs that identically.After: both revisions are verified with
git cat-file -e <sha>^{commit}; an absent one exits 2 and names the variable and the value. Resolvable revisions produce exactly the verdict they did before. Thepusharm is fixed the same way.2. Issue Gate — no pull request read as an unreferenced pull request
workflow_dispatchcarries nopull_requestobject, soPR_TITLEandPR_BODYrender as empty strings, grep matches nothing, and the step printedagainst a pull request that does not exist.
IS_BOTdoes not rescue it — a human dispatcher makes every disjunct false. Not hypothetical:tools/check_now_entry_shape.py:185already records the dispatch run that produced it.After: the step reads
PR_NUMBER, and with none it says it examined nothing and exits 0. POSIX 1003.3 calls that UNTESTED — no subject, which is different from an instrument that failed — and the message says it is not a pass for any PR.Evidence
PR_BASE_SHAand the value, no "SYNC REQUIRED"pushwith an absentPUSH_AFTERpull_requestCloses #7✅ Issue gate passedThe test extracts the Issue Gate's
run:body from the workflow rather than restating it, so editing the YAML and not the test is caught — and it asserts the extraction matched, because an extractor that stops matching would assert nothing at all.The NOW gate's control compares HEAD against itself rather than
HEAD~1: it needs no history, so it survives afetch-depth: 1checkout instead of silently skipping there. A control that vanishes with the runner's clone depth is not one.The test runs in
untrusted-input-gate.ymlrather than in either gate's own workflow, so a break in the test cannot turn a required context red for a reason unrelated to the change under test.Where these came from
A read-only fan-out over workflows, hooks, python gates and the Rust
--gatecommands: 41 candidates, 15 confirmed, 9 refuted, 0 agent deaths. These two are the only ones marked blocks-a-merge.Among the thirteen left is a fourth quiet shape that
tri gates quiet(#3049) does not classify —if grep -n PATTERN A B 2>/dev/null; then … else echo OK, where grep's exit 2 (cannot open) merges with exit 1 (no match) and theelseprints success.coq-kernel.yml:121is its only instance in 49 workflow files, and the reproduction ran GNU grep 3.11 with one operand absent and gotOK: no Admitted, exit 0. Listed in #3061 rather than fixed here.