This Bugzilla instance is a read-only archive of historic NetBeans bug reports. To report a bug in NetBeans please follow the project's instructions for reporting issues.
[NBdev-206]+ settings[NBdev-201] rh70, jdk1.3.1 ================================ Probably occures only on linux. If I mount CVS FS and press finish button of Mounting Wizard (choosing comand-line client support) IDE freez. Martin Entlicher says: "Probably problem of core - Repository collision of JAva parsing and add filesystem in core" I'm attaching FTD
Created attachment 1516 [details] FTD
Maybe the issue #12690 occures of this bug
Downgrading the priority because it is not fully reproductable.
JavaParser is waiting in AWT thread for something, so assigning to java module. But I am not sure so feel free to reassign.
"Java source parsing" daemon prio=1 tid=0x4cbb9280 nid=0x3316 waiting for monitor entry [0xbcbff000..0xbcbff8b0] at org.openide.filesystems.Repository.getFileSystems(Repository.java:263) at org.openide.filesystems.Repository.fileSystems(Repository.java:271) at org.openide.filesystems.FileSystemCapability.fileSystems(FileSystemCapability.java:70) ---- at org.netbeans.modules.debugger.support.LineBreakpointEvent.access$100(LineBreakpointEvent.java:61) at org.netbeans.modules.debugger.support.LineBreakpointEvent$FileSystemRepositoryListener.fileSystemAddedOrRemoved(LineBreakpointEvent.java:132) at org.netbeans.modules.debugger.support.LineBreakpointEvent$FileSystemRepositoryListener.fileSystemRemoved(LineBreakpointEvent.java:113) at org.openide.util.WeakListener$Repository.fileSystemRemoved(WeakListener.java:481) at org.openide.filesystems.Repository.fireFileSystem(Repository.java:466) ---- It seems that the event is dispatched while holding lock on the Repository object. The parser then cannot enumerate filesystems to obtain classpath for the project and the debugger (AWT thread which holds the filesystems lock) waits for the parser. "AWT-EventQueue-0" prio=1 tid=0x81cbc88 nid=0x3303 waiting on monitor [0xbe7fe000..0xbe7ff8b0] [snip]
Update: The repository object is quite innocent :-) at org.netbeans.modules.vcscore.wizard.mountcvs.CvsMountFS.addNotify(CvsMountFS.java:80) at org.openide.filesystems.Repository.addFileSystem(Repository.java:170) addNotify is called while holding the lock. addNotify calls removeFileSystem, which is probably a bad thing. What do you think ?
Adding developer of CVS to CC of this bug
fireFileSystem is called from removeFileSystem that holds lock. There already once was intention to prevent fire events without locks from removeFileSystem. But by mistake on line 202 stayed original firing of event. Sorry for that. So I`m assigning bug on me.
Yes, this will solve the deadlock. However in the mean time I have changed CvsMountFS.addNotify(CvsMountFS.java:80) to remove itself later in a request processor. This should also fix this deadlock.
Fixed in main trunk. Method removeFileSystem fires event outside of synchronized block.
In which build number? I hope it is in [NBdev-208]
if it is in build no. #208 I must reopen this bug. Ide freez ater I finish mounting wizard and try to start some work. It is hard to reproduce and I don't know if it isn't a coincidence. Because .... I give you my procedure of steps. My work procedure: ================== 1.take [NBdev-208] build and unzip it and run it. I ran it with my previous settings from NBdev branche. Tha last run build was [NBdev-207] with this settings 2. I open some files in editor (new 1 + 10 has been opened atomaticaly form my settings) 3. In the newly opened file I set a few breakpoints 4. Tools->Option-Editor->Java editor I customize color of the editor 5. File->Mount FS-> another type -> CVS and close this dialog !!!.In this time I recognice a dialog with Exception occured in Request Proccessor !!! a CVS wizard was started 6. I went trought this wizard, selected empty working and select local repository and Finsih it. 7. Seems that FS was mounted, but I'm not sure, cause the dialog with E hided my Explorer. And If I press Show detail on it. Its window has been resized but NO details has been shown. And an attempt to move with this window discovered that ide freez. I created another FuulThreadDump for you and also attach E. in Requered Proccessor (if it was stored in ide.log)
Created attachment 1547 [details] FTD II.
Created attachment 1548 [details] Still repeating E.in my IDE.log
Because NPE, which was cyclable writting to my ide.log make's my HDD full. I must increase the priority to P1 Maybe it isn't related to this issue and more over there was an issue #10723 that if you try mount CVS FS via File->Mount Filesystm... the E. was repeatedly written to your ide.log and made your HDD full and cause crash of your all apps. and OS.
At first glance there seems to be problem outside of filesystems. Perhaps in implemenattion of children in VcsFileSystem. Reassigned to vcscore.
Reproduction of an E. in Request Proccessor is quite easy even on windows 2000. I'm attaching it. IMHO, It seems that NPE which filled my ide.log & HDD is some individual issue:-/ So, what's your suggestion?
Created attachment 1550 [details] E. in RequedProccessor
ok. I fixed the last NPE. it was caused by the code in CvsMountFs. This filesystem (which is used just to mount javacvsfs or cvsfs) did not implement the AbstractFileSystem's info, change, attr interfaces. now it does (in a trivial way, just to prevent the NPE.)
ehmm... The one befor last Exception which caused overflow of my HDD isn't fixed,yet? am I right? pls, include the fix of NPE (under attachment named"Still repeating E.in my IDE.log (text/plain)") thanks P.S. sorry for confusing you with this lots of attachment which they aren't under any relationship. But all happend to me at once acction:-]
one more stuff: mind the second attachment "FTD II" where is FullThreadDump when ide freez. Probably caused by overloading ide.log and my HDD.
I fixed the NullPointerException in VcsFileSystem (attachment id=1548: "Still repeating E.in my IDE.log"). Fixed in dev build #210. This should finally fix the whole issue => marking as fixed.
The fix in Repository class i verified in the trunk rev 1.32.
*** Issue 12725 has been marked as a duplicate of this issue. ***
OK on [NBdev-211]
Resolved for 3.4.x or earlier, no new info since then -> closing.