Issue 116803 - Crash when closing stylist with closer
Summary: Crash when closing stylist with closer
Status: CLOSED FIXED
Alias: None
Product: Draw
Classification: Application
Component: ui (show other issues)
Version: DEV300m99
Hardware: All Linux, all
: P2 Trivial (vote)
Target Milestone: 3.4.1
Assignee: Mathias_Bauer
QA Contact: issues@graphics
URL:
Keywords: regression
: 117077 (view as issue list)
Depends on:
Blocks:
 
Reported: 2011-02-04 12:43 UTC by wolframgarten
Modified: 2013-03-01 20:23 UTC (History)
3 users (show)

See Also:
Issue Type: DEFECT
Latest Confirmation in: ---
Developer Difficulty: ---


Attachments
compiles sfx2/source/control/dispatch.cxx without compiler optimization on linux (1.09 KB, patch)
2011-04-11 12:28 UTC, clippka
no flags Details | Diff

Note You need to log in before you can comment on or make changes to this issue.
Description wolframgarten 2011-02-04 12:43:56 UTC
Open a new draw or impress, call the stylist using F11. Right click on default,
select modify. Close the dialog with cancel. Close the stylist by clicking on
the closer "X". Office crashes. Only reproducible unde rlinux, works in Writer.
Comment 1 wolframgarten 2011-03-01 12:49:39 UTC
*** Issue 117077 has been marked as a duplicate of this issue. ***
Comment 2 wolframgarten 2011-04-07 11:29:32 UTC
Works in 3.2.1. Adding keyword regression.
Comment 3 clippka 2011-04-07 15:38:16 UTC
this is an annoying bug. The crash is actually in sfx module. But as soon as I build sfx with debug information, the crash is gone.

I'm trying non pro build now to maybe get more information.

Also I suspect it may be caused by the broken sd string resource for the navigator which I fixed in cws sd2gbuild. Will check that if non pro does not give me more information
Comment 4 clippka 2011-04-07 16:34:15 UTC
linux non pro with just sfx lib linked with debug gives me SfxDispatcher::PostMsgHandler as the point of crash
Comment 5 clippka 2011-04-07 17:03:54 UTC
the string resource problem in sd was not the cause. Now back to good old printf debuging...
Comment 6 clippka 2011-04-07 17:53:46 UTC
OSL_TRACE showed me that we do not actually enter SfxDispatcher::PostMsgHandler

valgrind gives me this

==22573== Invalid read of size 4
==22573==    at 0x4A2D75A: SfxDispatcher::PostMsgHandler(SfxRequest*) (in /cws/bugfixes/DEV300/oracle/openoffice.org/basis3.4/program/libsfxli.so)
==22573==  Address 0x185f7c4 is not stack'd, malloc'd or (recently) free'd
==22573==
Comment 7 clippka 2011-04-11 11:08:46 UTC
my fprintf debugging shows me that the crash is at the end of SfxDispatcher::PostMsgHandler in the line

    delete pReq;

I confirmed that pReq is a valid pointer to a constructed SfxRequest that has not yet been deleted.

Valgrind shows me that the invalid read is at an address that looks unrelated to the SfxRequest, f.e. if pReq == 0xa6021c0, the Invalid read is at address 0x1d949d4.

This crash also happens after I replace FuTemplate::DoExecute with a simple return statement.

Will continue to reduce code in sd module to see if I can pinpoint the source of the problem.

Another option could be to compile dispatch.cxx without optimization but that may hide a real memory corruption.
Comment 8 clippka 2011-04-11 11:20:02 UTC
to summarize. Opening the stylist in impress, then select one style (can be presentation or graphic style) and select one of the options in the context menu, select cancel on the next dialog and then close the stylist, the office crashes.

Now this crash also happens if the slot called from the context menu does nothing in sd, f.e. just a return in DrawViewShell::FuTemporary.

So I'm currently out of ideas how this crash can be caused by the sd module, will try if the noopt compiling of dispatch.cxx changes anything.
Comment 9 clippka 2011-04-11 12:28:43 UTC
Created attachment 76343 [details]
compiles sfx2/source/control/dispatch.cxx without compiler optimization on linux

compiling sfx2/source/control/dispatch.cxx without optimization fixes this issue. Since there are no crashes on other plattforms, including 64 bit unxlngi I suspect it is safe to assume that this actually is a compiler problem.
Comment 10 clippka 2011-04-11 12:29:17 UTC
cl->mba: as discussed, I propose the attached patch. Please delegate as you see fit.
Comment 11 Mathias_Bauer 2011-04-28 16:50:52 UTC
thanks for the investigation; I applied the patch