eclipse-platform / eclipse.platform.releng.aggregator

Aggregated repository for Eclipse Java IDE
https://www.eclipse.org/eclipse/
Eclipse Public License 2.0
30 stars 74 forks source link

Temporarily use 2024-06 API-Tools for API checks in verification builds #2374

Closed HannesWell closed 2 weeks ago

HannesWell commented 2 weeks ago

Revert to use previous release's API-Tools to hopefully unblock verification builds that are currently suffering from https://github.com/eclipse-platform/eclipse.platform.releng.aggregator/pull/2360#issuecomment-2364207760.

If this doesn't help, the problem is probably not directly within PDE's API-Tools but in the integration with Tycho.

HannesWell commented 2 weeks ago

If this does not help, the only candidates on the Tycho site I see are

HannesWell commented 2 weeks ago

Ok, so with this applied and the api-checks profile enabled it still happens that the tycho-apitools:verify mojo hangs and I therefore assume https://github.com/eclipse-tycho/tycho/pull/4105 has something to do with it. But since I also saw false positive errors/warnings I assume that the underlying PDE Api-Tools als cause some troubles. Therefore I suspect that this could still fix some problems in verification builds. At least it worked to check a few plug-ins and also worked with tycho 4.0.8 and I didn't saw a change that should be incompatible to the old api-tools version. So I hope that in the worst case, this simply does not help.

laeubi commented 2 weeks ago

Ok, so with this applied and the api-checks profile enabled it still happens that the tycho-apitools:verify mojo hangs

Is there a way to reproduce the problem e.g. can one build a single project to see the hang or does it only happen randomly?

HannesWell commented 2 weeks ago

Ok, so with this applied and the api-checks profile enabled it still happens that the tycho-apitools:verify mojo hangs

Is there a way to reproduce the problem e.g. can one build a single project to see the hang or does it only happen randomly?

I didn't see a systematic schema so I think it's random. But when I run the aggregator build it usually hangs after a few projects have been built.

HannesWell commented 2 weeks ago

Btw. if we don't find another immediate remediation, couldn't we simply use only the apitools plug-in in version 4.0.8?

iloveeclipse commented 2 weeks ago

Could it be, that jenkins builds are failing now because something is not in-sync with the pom config? See https://ci.eclipse.org/platform/job/eclipse.platform.ui/job/PR-2288/4/console for recent build failure for PR that was green before rebase: https://github.com/eclipse-platform/eclipse.platform.ui/pull/2288

HannesWell commented 1 week ago

Could it be, that jenkins builds are failing now because something is not in-sync with the pom config? See https://ci.eclipse.org/platform/job/eclipse.platform.ui/job/PR-2288/4/console for recent build failure for PR that was green before rebase: eclipse-platform/eclipse.platform.ui#2288

I cannot really say anything with certainty. I assume this didn't make it worse, but I cannot guarantee it. But I have started another attempt for a work-around to unblock verification builds:

iloveeclipse commented 1 week ago

Could it be, that jenkins builds are failing now because something is not in-sync with the pom config? See https://ci.eclipse.org/platform/job/eclipse.platform.ui/job/PR-2288/4/console for recent build failure for PR that was green before rebase: eclipse-platform/eclipse.platform.ui#2288

I cannot really say anything with certainty. I assume this didn't make it worse, but I cannot guarantee it. But I have started another attempt for a work-around to unblock verification builds:

The funny thing is that PR build failed 3 times with different issues (two times due API errors) and after that it succeeded without any change in the code. So I assume we have (had?) some infrastructure issues.

HannesWell commented 1 week ago

The funny thing is that PR build failed 3 times with different issues (two times due API errors) and after that it succeeded without any change in the code. So I assume we have (had?) some infrastructure issues.

That's also possible, but as far as I can tell #2380 improved the situation for now.