Community
Participate
Working Groups
Build ID: I20080502-0100 Steps To Reproduce: 1. Start eclipse 2. Go to Preferences -> Network, select System Proxy Configuration 3. Then go to the CVS repository view (the eclipse source) and open the HEAD branch. Most times this will hang, regardless of the Gnome proxy settings. More information: This is caused by the latest patch that was checked in for bug 228738 (https://bugs.eclipse.org/bugs/attachment.cgi?id=97612). I had previously diagnosed and fixed this issue with the patch here: https://bugs.eclipse.org/bugs/attachment.cgi?id=97572 But the subsequent patch (that was actually checked in) removed my fix in order to get the Gnome configuration in a lazy manner. The fix is to make sure the Gnome configuration is accessed (using the native code) from the UI thread.
Oh, and I'm running Fedora 8 on an x86_64 system, though I'm using 32-bit Java and Eclipse.
Hi, I saw you change this to team, but the reason is happens in the CVS view is because CVS does the network access to the repository while not on the UI thread, and that causes the Gnome initialization (gconf calls) to be made on a thread other than the UI thread. This has nothing to do with the team component and was caused by the system proxy detection stuff. Have a look at this comment for more details: https://bugs.eclipse.org/bugs/show_bug.cgi?id=228738#c7
Francis, the proxy manager is owned by the team component. That's why Oleg reassigned it to Team. As I understand the problem is related to lazy initialization introduced by one of the patches in bug 228738. Moreover I see that it is reproducible. We will return to the previous version then. If bug 228738 is reopened, this bug is its duplicate, right? However I think that we should leave 228738 closed, since the work related to it is already in cvs. This bug is about special case related to threads.
(In reply to comment #3) > Francis, the proxy manager is owned by the team component. That's why Oleg > reassigned it to Team. > > As I understand the problem is related to lazy initialization introduced by one > of the patches in bug 228738. Moreover I see that it is reproducible. We will > return to the previous version then. > > If bug 228738 is reopened, this bug is its duplicate, right? However I think > that we should leave 228738 closed, since the work related to it is already in > cvs. This bug is about special case related to threads. > I agree, I just closed 228738. Do you want me to do a patch to make this right?
Sure. I will apply it on Monday.
(In reply to comment #5) > Sure. I will apply it on Monday. > I was not able to get to the patch for this and the other UI issue this weekend, but I can do it on Monday (my time). Sorry about the delay.
Created attachment 98752 [details] Patch Fixes the initialization to be back on the main thread. Tested several times (you need to restart eclipse each time and test with CVS) and it works correctly.
Adding Tomek as RC1 committer/reviewer.
Created attachment 99070 [details] Szymon's changes in Patch
Francis, could you please check if the latest changes works fine for you?
(In reply to comment #10) > Francis, could you please check if the latest changes works fine for you? > Just a minor comment, I see you are relying on the flag to not make calls to the native code if the library is not loaded (which is fine, and better). However, if you are doing that then the UnsatisfiedLinkException when accessing the gnome libraries should be logged as an error since it's not expected (and could indicate some kind of bug). But as I said this is a minor point.
Right. Good remark.
Created attachment 99239 [details] Patch v03
Reviewed and released to HEAD with some minor modifications.
Verified.