Community
Participate
Working Groups
Created attachment 199138 [details] Workspace log showing exceptions Build ID: Eclipse 3.7 (Indigo) 1-time occurrence, see attached workspace log, here is my understanding of what happened: - This was an existing fairly complex Indigo install with several dropins - I added another *.link file to dropins (adding egit) and launched - During launch, following exception was logged (p2 race condition?) org.osgi.framework.BundleException: State change in progress for bundle - Eclipse tried to bring up the Problems View - But that failed due to org.eclipse.swt.SWTException: Invalid thread access - Which caused the whole Eclipse Launch to terminate. On relaunching, everything was fine so the config itself looked OK. I'm reporting this since I've seen similar problems with the framework trying to activate the Problems View prematurely on other occasions too.
Any idea who tries to start the UI bundle from a non-UI thread?
In case the bundle is started in the non-UI thread we can use asyncExec instead of timerExec.
Created attachment 199165 [details] Fix
Why not asyncExec in line 392 too ? The runnable will timerExec itself repeatedly until the Workbench is up. In other cases I've seen a warning logged about "problems view has not been created yet", so potentially creating it earlier may be a good idea...
(In reply to comment #4) > Why not asyncExec in line 392 too ? Because the stuff will be in the display thread/loop after the first asyncExec ;-)
Ping, could this go into 3.7.1 rc 1 ? The sooner it's in the more often our automated testsuite will exercise it :)
(In reply to comment #6) > Ping, could this go into 3.7.1 rc 1 ? > The sooner it's in the more often our automated testsuite will exercise it :) Yes. As you know I had other priorities ;-).
Eric, OK for 3.7.1?
Looks good... Committed in >20110809. Applied the patch. Dani, could you do the VERIFY ?
(In reply to comment #9) > Looks good... > > Committed in >20110809. Applied the patch. See bug 351752 comment 8 f. > Dani, could you do the VERIFY ? Yes, will do.
Verified for 3.7.1. in M20110810-0800. I'm reopening this bug until we either have separate bugs for 3.8, 4.1.1 and 4.2 or we have comments in this bug confirming the commit into those branches.
(In reply to comment #11) > I'm reopening this bug until we either have separate bugs for 3.8, 4.1.1 and > 4.2 or we have comments in this bug confirming the commit into those branches. Opened bug 354490 against 3.8 and bug 354491 against 4.x. 4.1.1 and 4.2 has not split yet.
(In reply to comment #11) > Verified for 3.7.1. in M20110810-0800. Resetting to VERIFIED. Thank you for verifying the fix, Dani.
