Re: [Gtkpod-devel] itdb_init_ipod glitch

[P.G. R. <p.g...@ph...>]

Ah update to last...

Turns out it seems I was mistaken on the exact point of the crash. Rather t=
han=20
being the g_free(podpath) in itdb_create_directories, it is in fact the=20
freeing of the itdb database after the writes in itdb_init_ipod:

<snip>
writeok =3D itdb_shuffle_write(itdb, error);
	if(! writeok)
	{
	        itdb_free (itdb);
		return FALSE;
	}
=09
	/*itdb_free (itdb)*/;
	return TRUE;
<snip>

Literally commenting it out addresses the problem as far as muvon is=20
concerned. I will look more into this tomorrow and take up Christophe's=20
suggestion for a c test case as it maybe that the java is garbage collectin=
g=20
the itdb before the g_free (too clever for own good perhaps!).

Cheers for having a look Christophe.

Regards

PGR

On Monday 29 May 2006 22:31, Christophe Fergeau wrote:
> Le lundi 29 mai 2006 =E0 22:08 +0100, P.G. Richardson a =E9crit :
> > Jorg,
> >
> > Thanks for that...
> >
> > Just updated from cvs, hosed the shuffle and re-inited....
> >
> > ... and a small problem, which I have worked around but not yet resolve=
d.
> > In itdb_create_directories, I did g_free(podpath) as you did with
> > errordir originally then removed it before submitting the patch. With it
> > in place I get the attached backtrace.
>
> I glanced through the code, and I didn't find anything wrong with this
> g_free(podpath) (ie imo it's right and needed). It would be really nice
> if you had a C test case that you could run through valgrind, it would
> probably give really useful information.
>
> As a side note, the attached patch makes the error-handling in
> create_directories less invasive, thus the function is more readable imo
> (and it fixes a few leaks at the same time). The indentation with that
> patch is probably crap though :p I also removed the G_GNUC_INTERNAL
> since a simple static seems enough. I can make a patch fixing only the
> leaks and the G_GNUC_INTERNAL thing if you prefer. And I haven't even
> tried compiling that code...
>
> Christophe

=2D-=20
P.G. Richardson
Email:	p.g...@ph...