-
-
Notifications
You must be signed in to change notification settings - Fork 4.6k
run-examples-macos-on-metal job is failing #24291
Copy link
Copy link
Open
Labels
A-Build-SystemRelated to build systems or continuous integrationRelated to build systems or continuous integrationC-BugAn unexpected or incorrect behaviorAn unexpected or incorrect behaviorP-CriticalThis must be fixed immediately or contributors or users will be severely impactedThis must be fixed immediately or contributors or users will be severely impactedS-Ready-For-ImplementationThis issue is ready for an implementation PR. Go for it!This issue is ready for an implementation PR. Go for it!
Metadata
Metadata
Assignees
Labels
A-Build-SystemRelated to build systems or continuous integrationRelated to build systems or continuous integrationC-BugAn unexpected or incorrect behaviorAn unexpected or incorrect behaviorP-CriticalThis must be fixed immediately or contributors or users will be severely impactedThis must be fixed immediately or contributors or users will be severely impactedS-Ready-For-ImplementationThis issue is ready for an implementation PR. Go for it!This issue is ready for an implementation PR. Go for it!
See https://github.com/bevyengine/bevy/actions/runs/25822568493/job/75867370625?pr=24240 for an example of the failure, although it's quite widespread.
See
bevy/.github/workflows/example-run.yml
Line 22 in a399db0
The logs show that this is calling
rustup-initrather thancargo. This appears to be because we've somehow gotten that stuck in our cache. We should be able to clear this as a bandaid, but the bigger problem is that there's a problem in https://github.com/bevyengine/bevy/blob/a399db032194403995eb3b8ed14b3c348e5d36e3/.github/workflows/update-caches.ymlMy analysis here is that there are two bugs here:
update-caches.yml, to avoid this sort of problem. We could maybe do better than this, but https://github.com/Swatinem/rust-cache/blob/master/README.md#known-issues suggests to me that this is non-trivial to do correctly.1 should fix the current problem, at least when combined with a cache flush. I think 2 is probably worth doing too, but that should be a separate PR that looks at all of our usages of this pattern.