Community
Participate
Working Groups
Build Identifier: 20110615-0604 The SVN code has done this to me several times now. 1. Try to do a SVN check-in 2. It gets stuck early in the thermometer (sporadic, reason unknown) 3. Wait several minutes. 4. Give up and press cancel 5. The SVN plug-in blocks on "canceling", which never completes. 6. IO operations in Eclipse are blocked by this task—which can't be cancelled. It is now impossible to save any open files, initiate other activities or even quit Eclipse. This is a "CRITICAL" data-loss issue because it is now impossible to save any open files. The IDE is effectively hung: you have quite likely lost work in your files, and you will definitely lose IDE state since you can't even quit. I'm not sure if the issue in step (2) is caused by a bug in Subversive, a problem with my corporate SVN server, or a dropped packet. Step (3) seems like a problem, in that there is no timeout on subversion progress. Step (5) is the big issue here. It makes this plugin unsafe to use. I've seen this from other Eclipse plug-ins before, and it's a terrible problem. There should *always* be a hard-cancel available that unblocks other tasks _no_matter_what_. There is no legitimate reason why a failed SVN commit should keep me from saving a file that wasn't even part of the same project. Reproducible: Sometimes
Hello, Could you please: 1) specify Subversive and Subversive connectors version 2) attach a thread dump, captured when threads are blocked. There is a guideline about how to capture a stack trace when deadlock is occured: http://wiki.eclipse.org/How_to_report_a_deadlock
Created attachment 203549 [details] Frozen Eclipse stack dump
I (unfortunately) managed to experience another instance where SVN blocked my entire JVM. It wasn't exactly the same use case, but the symptoms were the same: some SVN activity sits forever, and ignores any effort to cancel it. This brings down the entire IDE, which must be killed. This time it corrupted my workbench, so I had to reconfigure my perspectives. This is really a *terrible* bug. I've attached the jstack. My installed versions: Subversive SVN Connectors 2.2.2.I20110602-1700 org.polarion.eclipse.team.svn.connector.feature.group Polarion Software Subversive SVN Integration for the Mylyn Project (Optional) (Incubation) 0.7.9.I20110602-1700 org.eclipse.team.svn.mylyn.feature.group Eclipse.org Subversive SVN JDT Ignore Extensions (Optional) (Incubation) 0.7.9.I20110602-1700 org.eclipse.team.svn.resource.ignore.rules.jdt.feature.group Eclipse.org Subversive SVN Team Provider (Incubation) 0.7.9.I20110602-1700 org.eclipse.team.svn.feature.group Eclipse.org SVNKit 1.3.5 Implementation (Optional) 2.2.2.I20110602-1700 org.polarion.eclipse.team.svn.connector.svnkit16.feature.group Polarion Software
Hello, I don't succeed to launch Eclipse Helios : the splash screen halts when trying the load the Subversive GUI (org.eclipse.team.svn.ui). I installed Subversion yesterday, but I've been able to start Eclipse several times without any problem. It seems the problem occurs because I switch off my system yesterday when leaving the office, and boot it today. I can't tell you the exact release I use, but I installed the last release of Helios one week ago, and grab Subversion from the Helios depository. I chose the last releases of connectors and others. I am on CentOS 5.5, and I use Java 1.6.0.27. Please don't tell me to use Indigo : it has some problems with CDT.
Hello, I don't succeed to launch Eclipse Helios : the splash screen halts when trying the load the Subversive GUI (org.eclipse.team.svn.ui). I installed Subversion yesterday, but I've been able to start Eclipse several times without any problem. It seems the problem occurs because I switch off my system yesterday when leaving the office, and boot it today. I can't tell you the exact release I use, but I installed the last release of Helios one week ago, and grab Subversion from the Helios depository. I chose the last releases of connectors and others. I am on CentOS 5.5, and I use Java 1.6.0.27. Please don't tell me to use Indigo : it has some problems with CDT. Here is the trace with eclipse -vm java -vmargs -XX:+HeapDumpOnOutOfMemoryError : Exception in thread "Thread-0" org.eclipse.swt.SWTException: Invalid thread access at org.eclipse.swt.SWT.error(SWT.java:4083) at org.eclipse.swt.SWT.error(SWT.java:3998) at org.eclipse.swt.SWT.error(SWT.java:3969) at org.eclipse.swt.widgets.Widget.error(Widget.java:466) at org.eclipse.swt.widgets.Widget.checkWidget(Widget.java:404) at org.eclipse.swt.widgets.Control.isVisible(Control.java:3175) at org.eclipse.swt.widgets.ProgressBar.timerProc(ProgressBar.java:276) at org.eclipse.swt.widgets.Display.windowTimerProc(Display.java:4378) at org.eclipse.equinox.launcher.JNIBridge._takedown_splash(Native Method) at org.eclipse.equinox.launcher.JNIBridge.takeDownSplash(JNIBridge.java:167) at org.eclipse.equinox.launcher.Main.takeDownSplash(Main.java:2043) at org.eclipse.equinox.launcher.Main$SplashHandler.run(Main.java:111)
Oodini's bug is a totally different use case.
(In reply to comment #2) > Created attachment 203549 [details] > Frozen Eclipse stack dump Hello Sarah, Thank you for the provided information, it is really helpful, although I can't say I understand now where it should be fixed. It seems that issue is related to the SVN Kit implementation of the svn+ssh protocol (I assume it is the protocol you're using to connect with the SVN server, right?), but it doesn't seem I can find any issues in the SVN Kit code, that could lead to the deadlock as it was happened in your case. On the other hand I'm not the SVN Kit library author, so it's expectable for me to not have deep knowledge of its actual workflows. So, for now I've filled corresponding report in the SVN Kit tracker: http://issues.tmatesoft.com/issue/SVNKIT-131
Thank you Alexander, And yeah, we're using svn+ssh to connect. Fortunately this doesn't seem to happen to me too often, maybe once a month. My guess is it may have to do with a slow response from the server side, as our server sometimes hangs. The last time it happened, I seem to recall that I got tired of waiting after maybe 20 sec and put it in the background. However I think the first time, when I files the bug, I left it in the foreground for several minutes. It would be nice if it were possible to kill a plug-in, or at least have some kind of force-save option to write out your files and close down the IDE without corrupting your workspace. It's really annoying when you have unsaved work and you can't save it because some process is making your save wait in queue. Anyway, at least it's progress :-) I'll watch what happens with SVNKit.
BTW it's funny how you can yell at the QA guys all the time for not attaching a proper stack trace and/or log file with a bug and then go do exactly the same thing when you're the frustrated consumer. Oh well :-)
It seems the issue is solved in the SVN Kit 1.3.6 and now we have SVN Kit 1.3.6 support (see bug #363000). The changes will be included into the next "early access" build.
