Repository navigation
feat(tri): gates prs — a conflicting PR loses its path-filtered CI (#2492) - #2546
Merged
Merged
Conversation
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
This was referenced Aug 23, 2026
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
#2543 left one pull request's missing CI undiagnosed. Checking every open pull request found the mechanism — and then the mechanism corrected itself.
Three more pull requests have the same shape
Two of them change
bootstrap/src/compiler.rsand got three checks:check-linked-issue, GitGuardian, NotebookLM. Every one path-less.That file's gate carries this comment, written to prevent exactly this:
The path was added. The gate is defeated by a merge conflict instead.
The mechanism
A pull request that is CONFLICTING when an event fires cannot have its merge diff computed, so
paths:filters cannot be evaluated and only path-less workflows run. The checks that remain are green — they are the ones that never look at the diff — which reads exactly like a passing pull request.And the rule I first wrote was wrong
The correlation looked exact: four conflicting pull requests with 3, 3, 9 and 7 checks; everything else 21 to 35. I wrote "a CONFLICTING pull request loses most of its checks" into the tool's own documentation.
An hour later two of those four reported 21 and 26. They had been mergeable when their events fired, kept those results, and only conflicted afterwards. A conflict does not retract past runs.
So the detectable shape is not the state. It is conflicting and a check list far below its siblings' — the command now computes a reference from the non-conflicting pull requests rather than asserting from the state alone. Three flagged, two correctly excluded.
Caught by re-running the command on fresher data, not by thinking harder. That is the argument for putting a finding into a tool rather than a document: a sentence cannot disagree with tomorrow's data; a command can.
And it crashed on its first run
Slicing a title by byte index, in the middle of an em dash — in a pull request this campaign had opened. Truncate by chars.
Refs #2492