GNOME Bugzilla – Bug 169006
dia crashes when opening some older file
Last modified: 2005-06-10 16:54:27 UTC
Distribution/Version: Debian Sid opening the file http://developer.skolelinux.no/arkitektur/network-arch_de.dia with Dia 0.94 makes it crash! This is stderr output: ** (dia:11047): WARNING **: No attribute show_background ((nil)) or no data((nil)) in this attribute ** (dia:11047): WARNING **: No attribute text ((nil)) or no data((nil)) in this attribute ** (dia:11047): WARNING **: No attribute flip_horizontal ((nil)) or no data((nil)) in this attribute ** (dia:11047): WARNING **: No attribute flip_vertical ((nil)) or no data((nil)) in this attribute (dia:11047): GLib-GObject-CRITICAL **: g_object_ref: assertion `G_IS_OBJECT (object)' failed segmentation fault
Ooo, CORE DUMP!!! Confirming and bumping severity to major. (I'm using a Gnome+Python build of Dia 0.94 from source on Mandrake) Opens without any errors in Dia 0.92.2 I opened the file using Dia 0.92.2 and resaved it and now it no longer crashes 0.94 so I think we have a workaround. Do you have any idea which older version of Dia was used to create your file? I did a little experimentation and it seems that Dia will spew warnings if flip_horizontal or flip_vertical attributes are not present. If these attributes are not present then Dia should not apply any transformation and not spew any warnings.
Alan: the 'no attribute' warnings are usually not the reason for crashes. Here the not shown "text" attribute was causing the g_object_ref CRITICAL warning. And setting the target to 1.0 looks rather odd to me. 2005-06-10 Hans Breuer <hans@breuer.org> * lib/properties.h : define PROP_STD_SHOW_BACKGROUND_OPTIONAL ... * objects/custom/custom_object.c : ... and use it to avoid complains about missing attribute "show_background". It is safe to leave this uninitialized. Same for flip_* and "text", the latter with comment cause it usually is *not* safe to do it for the general case. Fixes remaining issues with bug #169006 which where caused by an incompatible change in the shape file.
Thanks for the explanation. > And setting the target to 1.0 looks rather odd to me. I prefer to set some target rather than none and I didn't think a 1.0 release would happen with crash bugs like this in it but if you prefer I'll try to restrain myself from setting the Target Milestone field.