| Summary: | Stack traces in console in nightly builds | ||||||
|---|---|---|---|---|---|---|---|
| Product: | [Eclipse Project] Equinox | Reporter: | DJ Houghton <dj.houghton> | ||||
| Component: | p2 | Assignee: | Tobias Oberlies <t-oberlies> | ||||
| Status: | VERIFIED FIXED | QA Contact: | |||||
| Severity: | normal | ||||||
| Priority: | P3 | CC: | irbull, pascal, t-oberlies, tjwatson | ||||
| Version: | 3.7 | ||||||
| Target Milestone: | 3.7 M7 | ||||||
| Hardware: | PC | ||||||
| OS: | All | ||||||
| Whiteboard: | |||||||
| Attachments: |
|
||||||
|
Description
DJ Houghton
Still failing in last night's build. Do we know how serious this is? If it failed in the nightly then we can assume it will likely fail in the integration build. Is the build still good or should we send a note to eclipse-dev list? I've changed the summary, the current test failures are covered by bug 341169. Pascal, do you know why these stack traces don't seem to be causing any test failures? Created attachment 192751 [details]
Patch to handle empty array equivalent to null
This was in fact a regression from the publisher refactorings. The attached patch fixes the problem. I'm not yet committer, so feel free to apply it immediately.
I was a bit surprised that to see the tests pass despite the ArrayIndexOutOfBoundsException. Obviously someone happily catches all kinds of exceptions. I don't dare to adress this no - who knows who relies on this broken behaviour...
(In reply to comment #4) > I was a bit surprised that to see the tests pass despite the > ArrayIndexOutOfBoundsException. Obviously someone happily catches all kinds of > exceptions. I don't dare to adress this no - who knows who relies on this > broken behaviour... No, I don't think we should address this now, but it might be good to file a bug to capture this. EquinoxExecutableAction#fullBrandExecutables is the code which catches all exceptions and ignores them. Is this a new class or one that was re-factored or.... ? Comment on attachment 192751 [details]
Patch to handle empty array equivalent to null
I've released the patch.
(In reply to comment #6) > EquinoxExecutableAction#fullBrandExecutables is the code which catches all > exceptions and ignores them. Is this a new class or one that was re-factored > or.... ? This file came from the org.eclipse.equinox.p2.publisher bundle. Sorry, noone else to blame for this ;-) "It was like that when I got here", eh? :-) I've created bug 342198 to investigate if we should really be catching and ignoring all Exceptions in that publisher code. There were no more exceptions from the BrandingIron in the nightly test run: http://download.eclipse.org/eclipse/downloads/drops/N20110407-2000/testresults/consolelogs/linux.gtk-6.0_consolelog.txt We are still seeing this message in the error log. Is this a problem? [java] Error - 6 icon(s) not replaced in /tmp/p2.brandingIron3084785278937756120/eclipse.exe using /tmp/LauncherName8776581383509947268.ico I have opened bug 342489 to address the above error messages being printed to the console. |