Apache OpenOffice (AOO) Bugzilla – Issue 16270
pthread_mutex_unlock: mutex is locked by a different thread
Last modified: 2003-07-29 18:45:14 UTC
Mutexes may only be unlocked by the thread that locked them. However, it looks like OpenOffice breaks this rule in the following circumstance: In one shell window, run soffice without arguments under valgrind. In a second shell window, run "soffice foo.txt", where foo.txt is any document. The first instance of openoffice will emit several interesting errors (or at least it did on my dual CPU Red Hat 8 box) of the sort that imply nasty, hard-to-repeat problems may follow. A log is attached.
I forgot to mention that this may cause openoffice to crash. Haven't seen it myself, but I think there's at least one bug in Issuezilla that might be another example of this...
Created attachment 7288 [details] valgrind log of main soffice when you use a 2nd soffice to load a document
forgot to set target...
Dan, I guess you forgot to set the _correct_ target. You don't want to say, this is a release showstopper, do you?
Depends. Opening just once instance of OpenOffice and then closing it works fine. However, this bug is probably triggered any time you have OpenOffice open and double-click on another file in Konqueror or the Finder or whatever. Is that a special case? If so, then this bug is P3, not P2. It's a nasty little programming error of the brown-paper-bag sort if indeed it's real. I'd like to set the target to 1.1rc2, but I don't see that choice :-) Go ahead and reset the target to 1.1 if you think this can't be looked at before 1.1rc1's release. Thanks...
verified programming bug, will fix
my issue now
fixed in sw020
QA here seems to have a hard time reproducing the problem at all (on multi processor linux machines). Dan or anyone else experiencing this problem, could you try to build the desktop project with /source/app/officeipcthread.cxx rev. 1.33.2.1 /source/app/officeipcthread.hxx rev. 1.12.46.1 and copy the resulting unxlngi5(.pro)/bin/soffice to your program directory and see if the problem goes away. If you have problem building that yourself, tell me which version you are using and I'll prepare a binary with the fix for you to test.
please verify on sw020, script from Dan will be provided by me
ATR: checked in sw020, looks good, bug fixed.
.
..
Created attachment 7375 [details] Shell script used to reproduce the problem
*** Issue 16006 has been marked as a duplicate of this issue. ***
ATR: Checked again in latest OpenOffice build (m13s1), fixed.
I tested Issue: 16006 on 1.1rc2. It's work fine.
Pichi, did you test with Valgrind? You can't see the problem otherwise, at least on Linux.
No. I tested bug described in issue 16006 on the same OS (RH9). I don't know if this crash in cause of wrong pthread_mutex_unlock as described here. But some behavior indicate, it can be this problem. It dosn't crash in 1.1rc2, but crash in 1.1rc. If m13s1 is merged in 1.1rc2 it may be corected.
Oh, if you were able to reproduce issue 16006 without valgrind, then you're probably right. I guess the NPTL pthreads in RH9 are touchier than the Linuxthreads pthreads in RH8.