Summary: | Deadlock between ApiBaseline.doDispose() and ApiBaselineManager.initializeStateCache() | ||
---|---|---|---|
Product: | [Eclipse Project] PDE | Reporter: | Andrey Loskutov <loskutov> |
Component: | API Tools | Assignee: | Andrey Loskutov <loskutov> |
Status: | RESOLVED FIXED | QA Contact: | |
Severity: | normal | ||
Priority: | P3 | CC: | wellmann.hannes1 |
Version: | 4.23 | Keywords: | regression |
Target Milestone: | 4.23 M2 | ||
Hardware: | PC | ||
OS: | Windows 10 | ||
See Also: |
https://git.eclipse.org/r/c/pde/eclipse.pde.ui/+/190028 https://git.eclipse.org/c/pde/eclipse.pde.ui.git/commit/?id=bf8d0b62bd7809d456d0b728b9ed4609990b7579 |
||
Whiteboard: | |||
Bug Depends on: | |||
Bug Blocks: | 576736 |
Description
Andrey Loskutov
2022-01-25 16:45:31 EST
New Gerrit change created: https://git.eclipse.org/r/c/pde/eclipse.pde.ui/+/190028 Could this also cause deadlocks during test execution in the pde.ui build? I observed it twice that the build timed-out at the following location: """ Running org.eclipse.pde.api.tools.tests.ApiToolsPluginTestSuite reflectNestedClassUseDollar=true due to isJRE9Plus Build timed out (after 39 minutes). Marking the build as aborted. Terminating xvnc. """ A retrigger let the build pass. (In reply to Hannes Wellmann from comment #2) > Could this also cause deadlocks during test execution in the pde.ui build? > > I observed it twice that the build timed-out at the following location: > """ > Running org.eclipse.pde.api.tools.tests.ApiToolsPluginTestSuite > reflectNestedClassUseDollar=true due to isJRE9Plus > Build timed out (after 39 minutes). Marking the build as aborted. > Terminating xvnc. > """ > > A retrigger let the build pass. Shouldn't, because during tests we don't run API analysis as job parallel to the build, but without a stack hard to say. If there *are* parallel PDE tasks, for sure, the code is a mine field and isn't seriously meant to be used in MT environment (despite a lot of synchronized methods that actually cause most of the issues in MT environment). (In reply to Andrey Loskutov from comment #3) > > Shouldn't, because during tests we don't run API analysis as job parallel to > the build, but without a stack hard to say. > > If there *are* parallel PDE tasks, for sure, the code is a mine field and > isn't seriously meant to be used in MT environment (despite a lot of > synchronized methods that actually cause most of the issues in MT > environment). Understand. Unfortunately I cannot contribute much more information. I just observed the deadlock in the tests recently two times. Do you know if this happened before? Just in case you can get more information from the console, this was one build: https://ci.eclipse.org/pde/job/eclipse.pde.ui-Gerrit/4078/ (In reply to Hannes Wellmann from comment #4) > Understand. Unfortunately I cannot contribute much more information. I just > observed the deadlock in the tests recently two times. Do you know if this > happened before? No, not really. > Just in case you can get more information from the console, this was one > build: > https://ci.eclipse.org/pde/job/eclipse.pde.ui-Gerrit/4078/ Unfortunately I don't see any thread dumps there (like in JDT tests), so I've added something similar via bug 578391. Feel free to add freeze report to more tests, especially if you know which were affected. Gerrit change https://git.eclipse.org/r/c/pde/eclipse.pde.ui/+/190028 was merged to [master]. Commit: http://git.eclipse.org/c/pde/eclipse.pde.ui.git/commit/?id=bf8d0b62bd7809d456d0b728b9ed4609990b7579 |