Re: Help in debugging the DirectFB backend - issue reprodicible with X11 but not a crasher

[Attilio Fiandrotti <fiandro tiscali it>]

Attilio Fiandrotti wrote:
> Attilio Fiandrotti wrote:
>
> > Carl Worth wrote:
> >
> >
> > > On Thu, 27 Jul 2006 01:06:43 +0200, Attilio Fiandrotti wrote:
> > >
> > >
> > >
> > > > I think cairodfb or gdkdfb must previously have corrupted memory 
> > > > somewhere, but i can't detect when nor where: can anyone reproduce this 
> > > > or give me an hint about how to catch it?
> > >
> > >
> > > Have you tried running the program under valgrind? Something like:
> >
> >
> > yes, and below is what vg tells me.
> > From a previous test using gdb (see log under vg's log) i noticed the 
> > stack was messed up, and going back from gtk_target_table_free() i found 
> > the gtk_text_layout_set_buffer() problem i reported about.
> > Note i often (but not always) get other strange crashes, like when i 
> > resize a window and the gtk app was linked using efence (i think efence 
> > unveils silent memory corruptions)
>
>
> <snip>
>
> rplying to myself, i report an experiment i did with GTK+ from CVS using 
> X11 backend and running GIMP as a test application.
> As you can see, below tk_text_buffer_free_target_lists() passes "8" as a 
> second parameter to gtk_target_table_free(), but in the stack i fing 
> "141806480".
> Moreover, the i counter in gtk_target_table_free() is decremented (from 
> 8 to 0) instead of being incremented from 0 to 8.
> This looks strange to me: is there any explanation for this ? could this 
> be a crasher bug with DFB and a silent bug with X11 ? can anyone 
> reproduce this ?

During a furter test i was able to crash the gimp, using the X backend, 
exactly the same way it did with GTKDFB.
The test was run on a debian stable using

-GTK+ from CVS
-Glib 2.12.0
-cairo from cvs
-pango-1.13.3

instead of standard deian stable corresponding libraries

Attilio