Apache OpenOffice (AOO) Bugzilla – Issue 15436
Crashes in file picker dialog when run under gdb
Last modified: 2004-03-22 09:36:01 UTC
This is pretty annoying -- it makes it hard to get symbolic stack tracebacks of other people's bugs if reproducing the bug requires using "Export". Just try to export as pdf; when you type the first char in the file picker, the program aborts. Only happens when run under gdb. (FWIW, I'm using Red Hat 8.) So I guess it's a reverse Heisenbug...
Created attachment 6750 [details] Stack dump of app crash after first keystroke in Export As file picker
TM->PL: Please have a look, thanks !
You are looking at the wrong thread, the file picker is just happily waiting for input (waits in select) while another thread probably crashed (caught SIGABRT). As i cannot reproduce the problem could you please append the stack of the crashed thread (see which threads are available with "info threads", select a thread with "thread <n>").
pl -> tm: please try to reproduce, i can't. pl -> dan: can you still reproduce this and perhaps produce a stack trace of the thread that actually crashes ?
Sorry chum, that's the only thread left. "info thread" is mute at that point. I suppose I'll have to build from source one of these days and track it down further myself. Unless you'd fancy a gdbserver session...
I can't reproduce on Suse 8.0 with OOo 1.1RC.
Ah, well, once of these days I'll build OOo from source, and maybe fix it myself if it's still happening. Thanks for looking.
Hey, I have some more info! With rc2, when I do this, I get the following stack trace from OOo itself before it pops me back to the gdb prompt. Maybe this stack trace is more enlightening? I have a copy of rc1 compiled with line numbers; I could try to get it to fail there, too, if you like. Fatal exception: Signal 5 Stack: /opt/OpenOffice.org1.1rc2/program/libsal.so.3[0x40baee02] /opt/OpenOffice.org1.1rc2/program/libsal.so.3[0x40baef8c] /opt/OpenOffice.org1.1rc2/program/libsal.so.3[0x40baf052] /lib/libpthread.so.0[0x4112291d] /lib/libc.so.6[0x412bdc88] /lib/libc.so.6[0x413a0a4e] /lib/ld-linux.so.2(_dl_catch_error+0x86)[0x4000a426] /lib/libc.so.6(_dl_open+0x5f)[0x413a069f] /lib/libdl.so.2[0x41112f19] /lib/ld-linux.so.2(_dl_catch_error+0x86)[0x4000a426] /lib/libdl.so.2[0x411132c4] /lib/libdl.so.2(dlopen+0x34)[0x41112ec4] /opt/OpenOffice.org1.1rc2/program/libsal.so.3[0x40ba5089] /opt/OpenOffice.org1.1rc2/program/libsal.so.3(osl_loadModule+0x9d)[0x40ba501d] /opt/OpenOffice.org1.1rc2/program/libcppuhelpergcc3.so.3(_ZN4cppu29loadSharedLibComponentFactoryERKN3rtl8OUStringES3_S3_RKN3com3sun4st ar3uno9ReferenceINS6_4lang20XMultiServiceFactoryEEERKNS8_INS6_8registry12XRegistryKeyEEE+0x17f)[0x40b55287] /opt/OpenOffice.org1.1rc2/program/shlibloader.uno.so[0x41e1cf29] /opt/OpenOffice.org1.1rc2/program/libcppuhelpergcc3.so.3[0x40b4a1a1] /opt/OpenOffice.org1.1rc2/program/libcppuhelpergcc3.so.3[0x40b4923c] /opt/OpenOffice.org1.1rc2/program/libcppuhelpergcc3.so.3[0x40b481f2] /opt/OpenOffice.org1.1rc2/program/libcppuhelpergcc3.so.3[0x40b48eab] /opt/OpenOffice.org1.1rc2/program/servicemgr.uno.so[0x4205abd2] /opt/OpenOffice.org1.1rc2/program/servicemgr.uno.so[0x4205b50f] /opt/OpenOffice.org1.1rc2/program/libsvt645li.so(_ZN20SvtMatchContext_Impl10ReadFolderERK6StringS2_h+0x8dc)[0x407e8f46] /opt/OpenOffice.org1.1rc2/program/libsvt645li.so(_ZN20SvtMatchContext_Impl3runEv+0x442)[0x407ea530] /opt/OpenOffice.org1.1rc2/program/libvos3gcc3.so(_ZN3vos27_cpp_OThread_WorkerFunctionEPv+0x14)[0x40b11f66] /opt/OpenOffice.org1.1rc2/program/libvos3gcc3.so(_Z23_OThread_WorkerFunctionPv+0x1f)[0x40b11f4b] /opt/OpenOffice.org1.1rc2/program/libsal.so.3[0x40ba453d] /lib/libpthread.so.0[0x4111bbd0] /lib/libc.so.6(__clone+0x3a)[0x41370f9a] (no debugging symbols found)...(no debugging symbols found)... Program received signal SIGABRT, Aborted.
TM->Dan: Can you reproduce this in RC3 too ? If you do so, please submit a stack trace. Thanks !
Looks like it might have been a gdb or glibc bug in RH8. I did a fresh install of Red Hat 9 and OOo1.1.1rc3, and could not reproduce the problem. Thanks for looking into it -- sorry I wasted your time! (Incidentally, I noticed that a new thread is created *for each keystroke* in that dialog. Somebody was smoking crack when they wrote that! :-)
.