Repository navigation
Conversation
About 5% of dev server starts crash when several Playwright workers start servers in the same project directory. React Router's dev plugin deletes and regenerates `.react-router/types` on every start, without awaiting the work, so an overlapping start fails with ENOTEMPTY (or ENOENT) and the unhandled rejection exits the CLI with code 1. In CI this failed whole spec files, and recipes, which get only one Playwright retry, could fail the job outright. `DevServer.start()` now retries a crash that happens before the server reports ready, up to three attempts, logging the CLI's final output line on each retry so the frequency and cause stay visible in CI logs. Only early exits are retried. Startup timeouts already cost two minutes and Shopify CLI login/link errors cannot be fixed by trying again, so those still fail immediately. Starting three servers at a time in one project locally: 8 of 165 starts crashed through the CLI directly, all in `.react-router/types`. Co-authored-by: AI (Pi/Claude Opus 5.5) <noreply@pi.dev>
Passing E2E runs take 9-14 minutes, but each retry of a failing test can add a minute or more (consent specs get four attempts at 60s each). A run with a handful of failures overran the 20-minute limit, GitHub reported the job as cancelled rather than failed, and the Playwright report upload (`if: always() && !cancelled()`) was skipped. The real test failures were only visible by digging through the raw log. Raising the limit to 30 minutes lets a failing run finish, report as a failure, and upload its report. Co-authored-by: AI (Pi/Claude Opus 5.5) <noreply@pi.dev>
The retry log used the CLI's last line of output, but the CLI prints its error followed by a stack trace, so on the real skeleton the log read "at watch (.../vite.js:1392:3)" instead of the ENOENT/ENOTEMPTY error that explains the crash. The summary now uses the last line that names an error, falling back to the last line when none does. Co-authored-by: AI (Pi/Claude Opus 5.5) <noreply@pi.dev>
The dev server fixture tests log retries to the same CI output that is searched for real startup crashes. Their simulated error was a copy of the real React Router ENOTEMPTY error, so every run would have appeared to contain several real crashes. The simulated error now says it is simulated. Co-authored-by: AI (Pi/Claude Opus 5.5) <noreply@pi.dev>
kdaviduik
added this pull request to stack #4270
October 24, 2026 03:40
kdaviduik
changed the title
Contributor
|
Oxygen deployed a preview of your
Learn more about Hydrogen's GitHub integration. |
This branch has not been deployed
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.
WHY are these changes introduced?
Stacked on #4268, which surfaced why E2E dev servers crash during startup.
About 5% of dev server starts crash when several Playwright workers start servers in the same project directory. React Router's dev plugin deletes and regenerates
.react-router/typeson every start, without retrying when the directory changes underneath it. An overlapping start fails withENOTEMPTYorENOENT, and the CLI exits with code 1. Playwright retries absorb most of these, butrecipesgets only one retry, which is how the merge queue run for #4233 failed.Separately, passing E2E runs take 9 to 14 minutes against a 20-minute job limit. A run with a few failing tests overruns it, GitHub reports the job as cancelled rather than failed, and the Playwright report upload (
if: always() && !cancelled()) is skipped, which hides the failing tests.WHAT is this pull request doing?
DevServer.start()retries a crash that happens before the server is ready, up to 3 attempts. Each retry logs the CLI's error line, so the frequency and cause stay visible in CI logs.Starting three dev servers at a time in
templates/skeletonlocally:.react-router/typesDevServerwith this retryOut of scope
Risk