Crash after changing canvas size fields by click

Bug #168173 reported by M-miamoto
2
Affects Status Importance Assigned to Milestone
Inkscape
Fix Released
Critical
Unassigned

Bug Description

When changing a document's canvas size, Inkscape 0.44 +dev nightly (25 Oct)
crashes after a width value is entered and, without hitting enter, the
height field is clicked upon.

Revision history for this message
Bryce Harrington (bryce) wrote :

Originator: NO

Could a second person verify this occurs on Windows? I cannot reproduce
the issue on Linux.

Revision history for this message
Johan Engelen (johanengelen) wrote :

Originator: NO

Cannot reproduce on WinXP, Inkscape 0.45+devel (01-01-2007).
(I assumed changing the document's canvas is done in the "Document
Properties" dialog ( -> Page -> Format -> Custom size))

Could you please give the exact steps to reproduce the crash (maybe
provide an SVG too)?

Revision history for this message
gustav_b (gustav-b) wrote :

Originator: NO

File Added: inkscape-effects_with_flowcharts.svgz

Revision history for this message
gustav_b (gustav-b) wrote :

Originator: NO

File Added: 1624076-warnings+bt

Revision history for this message
gustav_b (gustav-b) wrote :

Originator: NO

I can reproduce this /sometimes/ using svn head (on linux with gtk 2.10.6
/ gtkmm 2.12.2).

I've managed to trigger it in some cases using the following steps:
1. Start Inkscape and open the attached
inkscape-effects_with_flowcharts.svgz
2. Ctrl+shift+d for document properties
3. Change the document width field from 1500 to 1400
4. Click the height field
...sometimes I just get some warnings, and sometimes it also crashes with
the attached backtrace.

(I occasionally manage to crash it by repeting step 3 and 4 with new width
values)

Could this be caused by a problem with our custom signal_value_changed for
the Scalar widget and unfortunate timing with the default signal for
focus-out on the Gtk::Spinbutton? A timing problem would explain the
randomness of this bug...

Revision history for this message
M-miamoto (m-miamoto) wrote :

Originator: YES

Yes gustav_b, those are the exact steps i have followed to trigger the
crash (besides dimensions). The crash occurs every time for me on WinXP.
And good point about the custom signal_value_changed.

johanengelen: 0.45+devel (1-1-07)? Where is that available? The bug may
have been inadvertently fixed if you can't reproduce the crash in WinXP.

Revision history for this message
Molumen (molumen) wrote :

Originator: NO

Cannot reproduce in 0.45devel build 04 jan 2006 on WIN XP SP2.

Revision history for this message
Bryce Harrington (bryce) wrote :

Originator: NO

Thanks gustav_b! Since this bug was able to be reproduced on Linux, I'm
removing the Win32-specific categorization. Since it appears to afflict
multiple platforms, this makes it more important to get fixed, too.

Revision history for this message
gustav_b (gustav-b) wrote :

Originator: NO

Could this be caused by the GTK bug mentioned in
https://sourceforge.net/tracker/?func=detail&atid=604306&aid=1580903&group_id=93438
?
(The GTK bug in question is
http://bugzilla.gnome.org/show_bug.cgi?id=374378). Looking at the warnings
they both fail on the same assertion.

I don't have time to build cvs head GTK on my system right now, so if
anyone using a GTK build no older than 2006-12-30 could try to reproduce
this using the steps in my last comment it would be of much help.

Revision history for this message
Johan Engelen (johanengelen) wrote :

Originator: NO

m_miamoto: bleeding edge versions of inkscape can be found at
http://inkscape.modevia.com/

Revision history for this message
Buliabyak-users (buliabyak-users) wrote :

Originator: NO

The " assertion failed: GTK_WIDGET_HAS_FOCUS" crash was fixed in gtk
2.10.7 and later, as explained in rel notes:

http://wiki.inkscape.org/wiki/index.php/ReleaseNotes045#Spinbuttons_may_crash_if_you_have_GTK.2B_older_than_2.10.7

Actually they didn't really fix it - they just removed the assert. The
warning is still fired. The deeper reason is gtk's unability to cope with
interruptible display which may call main loop iterations within a
value-changed callback.

But anyway, it does not crash so i'm closing this.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.