blog: the gate was right and nothing stopped (en+ru) - #866
Merged
Conversation
A gate that guards against retracted numbers re-entering live documents caught a violation on the pull request that introduced it -- right file, right lines, exit 1, three times, once per line -- and the pull request merged anyway, because it is not among the four required checks. It stayed red on the main branch through two further merges. The violation was mine, in a note about that same gate: it quoted three of the retracted figures as examples of what the registry protects. The post names this as the THIRD variant of a reach failure, beside the two the gate audit had already found: "it never runs" (a self-test present in the tree and invoked by nothing) and "it runs on the wrong diff" (a paths: filter omitting the gate's own script). This one is the least safe while looking the safest -- it produces a full, correct, actionable failure report into a stream nobody is required to read. It argues that "not required" is worse than "not there": a gate that has never run is honestly empty and has trained nobody, while a gate that fails without consequence teaches that red is a colour some checks are. Two other gates in the same repository are red on main permanently, one with no green run in the last hundred. Ends on the fix that was refused. The gate offers an exemption list for exactly this case and three lines would have turned the branch green in seconds; the prose was rewritten instead, because every exemption is a small permanent hole. And on the decision NOT taken: whether the gate joins the required set is an owner call, since a branch ruleset is a security setting and quietly widening what blocks other people's merges is not a repair. Verified locally: tsc clean on the touched files, vite build succeeds, the post is in the bundle, no literal markup outside code, EN and RU audits pass.
github-actions Bot
added a commit
that referenced
this pull request
Aug 22, 2026
blog: the gate was right and nothing stopped (en+ru) (#866) A gate that guards against retracted numbers re-entering live documents caught a violation on the pull request that introduced it -- right file, right lines, exit 1, three times, once per line -- and the pull request merged anyway, because it is not among the four required checks. It stayed red on the main branch through two further merges. The violation was mine, in a note about that same gate: it quoted three of the retracted figures as examples of what the registry protects. The post names this as the THIRD variant of a reach failure, beside the two the gate audit had already found: "it never runs" (a self-test present in the tree and invoked by nothing) and "it runs on the wrong diff" (a paths: filter omitting the gate's own script). This one is the least safe while looking the safest -- it produces a full, correct, actionable failure report into a stream nobody is required to read. It argues that "not required" is worse than "not there": a gate that has never run is honestly empty and has trained nobody, while a gate that fails without consequence teaches that red is a colour some checks are. Two other gates in the same repository are red on main permanently, one with no green run in the last hundred. Ends on the fix that was refused. The gate offers an exemption list for exactly this case and three lines would have turned the branch green in seconds; the prose was rewritten instead, because every exemption is a small permanent hole. And on the decision NOT taken: whether the gate joins the required set is an owner call, since a branch ruleset is a security setting and quietly widening what blocks other people's merges is not a repair. Verified locally: tsc clean on the touched files, vite build succeeds, the post is in the bundle, no literal markup outside code, EN and RU audits pass.
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.
Sixth post of the honesty campaign.
A gate guarding against retracted numbers re-entering live documents caught a violation on the pull request that introduced it — right file, right lines, exit 1, three times, once per line — and the PR merged anyway, because it is not among the four required checks. It stayed red on main through two further merges.
The violation was mine, in a note about that same gate: it quoted three of the retracted figures as examples of what the registry protects.
The post names this as the third variant of a reach failure, beside the two the audit had already found:
paths:filter omitting the gate's own script;The first two are absences you can grep for. This one produces a full, correct, actionable failure report into a stream nobody is required to read — which is why "not required" is worse than "not there": a gate that never ran is honestly empty and has trained nobody, while a gate that fails without consequence teaches that red is a colour some checks are.
Ends on the fix that was refused (three exemption lines would have turned the branch green in seconds; the prose was rewritten instead) and the decision not taken (a branch ruleset is a security setting — filed for the owner, alongside two gates that are red on main permanently).
Verified locally: tsc clean, build succeeds, post in the bundle, no literal markup outside code, EN and RU audits pass.
🤖 Generated with Claude Code