Bugzilla – Full Text Bug Listing
| Summary: | Infinite loop in org.eclipse.acceleo.common.utils.CompactLinkedHashSet#deleteIndex | ||||||
|---|---|---|---|---|---|---|---|
| Product: | [Modeling] Acceleo | Reporter: | Nicolas Bros <nicolas.bros> | ||||
| Component: | Core | Assignee: | Project Inbox <acceleo-inbox> | ||||
| Status: | CLOSED FIXED | QA Contact: | |||||
| Severity: | normal | ||||||
| Priority: | P3 | CC: | laurent.goubet, modisco.web-inbox | ||||
| Version: | 3.0.0 | ||||||
| Target Milestone: | --- | ||||||
| Hardware: | PC | ||||||
| OS: | Linux | ||||||
| Whiteboard: | |||||||
| Bug Depends on: | |||||||
| Bug Blocks: | 341193 | ||||||
| Attachments: |
|
||||||
This is with: org.eclipse.acceleo.common_3.1.0.v20110323-1016.jar Note that the test in question doesn't use Acceleo explicitly. The test is deleting a project, which causes a workspace change notification, which Acceleo acts on by (trying to) remove an element from the set. I can't understand how this loop could be infinite ... Could you tell me which of your unit test fails, and how I can reproduce the problem if there are mandatory steps other than simply launching the offending suite? (In reply to comment #2) > I can't understand how this loop could be infinite ... Could you tell me which > of your unit test fails The test is here: https://dev.eclipse.org/svnroot/modeling/org.eclipse.mdt.modisco/tests/trunk/org.eclipse.modisco.jee.jsp.discoverer.tests/src/org/eclipse/modisco/jee/jsp/discoverer/tests/compatibility/TestHtml.java > and how I can reproduce the problem if there are > mandatory steps other than simply launching the offending suite? I couldn't identify which conditions triggered this problem. This happens often (but not at each run) on Hudson, but I couldn't reproduce locally. Maybe we could try to run the tests on Hudson with a version of the acceleo.common plug-in with some debug code to print information about what happens in the method and in the loop? "couldn't reproduce locally" and "random" are two things I really don't like to see on bugzillas :p. I've just commited a fix on HEAD that should remove the possibility of a failure (that the only possibility I see for an infinite loop there). Could you try and see if you manage to reproduce? (In reply to comment #4) > "couldn't reproduce locally" and "random" are two things I really don't like to > see on bugzillas :p. I know, me neither :) > I've just commited a fix on HEAD that should remove the possibility of a > failure (that the only possibility I see for an infinite loop there). Could you > try and see if you manage to reproduce? Do you have a nightly update site from which I can get the fixed binaries? I looked in: http://download.eclipse.org/modeling/m2t/acceleo/updates/nightly/3.1/ But it seems the last version dates back from 2011-03-21 I just promoted a new nightly with this change, you can grab it from the nightly update site you mentionned in your previous comment :). Thank you. I will report back here after a few builds to confirm the problem is solved. I haven't seen this error again since I switched to the nightly build. So, I'm marking this bug as fixed. Thanks for the update. I hope this is indeed fixed and not hidden, but with no way to validate, I'll just pray this doesn't pop out again some time in the future. Closing our "resolved" and "verified" bugs. |
Created attachment 192344 [details] stacktrace Some MoDisco unit tests are failing due to a timeout. It appears that the tests get stuck in an infinite loop in Acceleo code: I think that the while loop in CompactLinkedHashSet#deleteIndex never meets its condition. Note that I couldn't reproduce this locally. It only happens randomly on a Hudson instance on Eclipse servers. See the attached stack trace.