GNOME Bugzilla – Bug 113800
evolution segment fault if gnopernicus start
Last modified: 2009-08-15 18:40:50 UTC
start gnopernicus, Then start evolution 1.3.92 with accessiblity enabled, i.e. GNOME_ACCESSIBILITY=1, evolution crash after flash window disappear and the main window begin to show.
Just to find the reason, I comment out spi_atk_emit_eventv() in spi_atk_bridge_window_event_listener(), then the evolution can be start up. Any idea?
I am using gnopernicus trunk, redhat 9
I am on redhat 8 with a garnome 0.21.0. gnopernicus is cvs header (2003-05-26 checked out). In this case, not only evolution but also gedit crashed. Please see a stack on evolution crash:
+ Trace 37233
It seems that windows related event processing cause the problem. In srl_process_event_real_window(), comment out the code as below, evolution start up, and focus event can be read out. Index: SRLow.c =================================================================== RCS file: /cvs/gnome/gnopernicus/srlow/libsrlow/SRLow.c,v retrieving revision 1.51 diff -u -4 -r1.51 SRLow.c --- SRLow.c 21 May 2003 13:51:57 -0000 1.51 +++ SRLow.c 28 May 2003 06:23:13 -0000 @@ -646,8 +646,9 @@ if (sro_get_from_accessible_event (event->acc, event->type, &obj)) { SREvent *ev; + /* ev = sre_new (); if (ev) { ev->type = SR_EVENT_WINDOW; @@ -658,8 +659,9 @@ notify_clients (ev, 0); sre_release_reference (ev); } sro_release_reference (obj); + */ } return TRUE; } Any Idea?
In my opinion, this is not a gnopernicus bug. It seems to be a at-spi or atk bug. Gnopernicus is a separate process. It ask information from other processes. If that process crashes, then something is wrong there, not in gnopernicus.
Created attachment 16899 [details] [review] a simple test program
Previous attached files is a simple test program. It should be copied in at-spi/test directory, then recompiled and launched. If, while this test is running, the applications mentioned above has the same behaviour, then the applications do something wrong. Please run this test and confirm me if apps have same behaviour with test program.
I have tested the program. I first launch it, then start evolution with accessiblity. evolution started and run well. Below is the output of the test program, 806b428p window:create for Ximian Evolution() role frame with details 0 0 806b270p window:restore for Ximian Evolution() role frame with details 0 0 806b240p window:deactivate for Ximian Evolution() role frame with details 0 0 806b428p window:destroy for Ximian Evolution() role frame with details 0 0 806b270p window:create for Summary - Ximian Evolution() role frame with details 0 0 806b270p window:restore for () role window with details 0 0 806b428p window:restore for () role window with details 0 0 806b7b0p window:restore for Summary - Ximian Evolution() role frame with details 0 0 806b3c8p window:maximize for Summary - Ximian Evolution() role frame with details 0 0 806b3c8p window:restore for () role window with details 0 0 806b270p window:deactivate for INBOX (358) - Ximian Evolution() role frame with details 0 0
note that the test program attached leaks three strings for every event processed...
Created attachment 16939 [details] [review] another test file
Please run gnopernicus with new SRLow.c file and send me back the results.
Evolution still crash, after the flash window show evolution flash disappear, output as below, ********************** * SCREEN READER CORE * ********************** gnopernicus-Message: Callbacks are NOT supported by all drivers gnopernicus-Message: speech initialization succeded (srcore:3835): gnopernicus-WARNING **: Voice V0 Kevin - Festival GNOME Speech Driver is NOT supporting speech markers gnopernicus will use a "idle" callback and will NOT force sometimes it call start: event window:create for Ximian Evolution end (srcore:3835): gnopernicus-WARNING **: Voice V0 Kevin - Festival GNOME Speech Driver is NOT supporting speech markers start: event window:destroy for Ximian Evolution end (srcore:3835): gnopernicus-WARNING **: Voice V0 Kevin - Festival GNOME Speech Driver is NOT supporting speech markers start: event window:create for Summary - Ximian Evolution end start: event window:restore for (null) end start: event window:restore for (null) end start: event window:restore for Summary - Ximian Evolution end
I noticed a comment that gedit also crashed. Does gedit or evolution crash when using version 1.3.4 of at-spi? If so, I think that this bug should be transferred to at-spi/atk-bridge.
Because this crash occurs in onother process, this is not a gnopernicus bug. Please report this bug to evolution component.
Apologies for spam; adding a11y keyword so these bugs are picked up by our historical bug tracking script. Find/delete these emails by searching for "calum fixing a11y script".