Re: [Gtkpod-devel] libgpod segfault

[Christophe F. <te...@gn...>]

Le lundi 28 novembre 2005 =C3=A0 23:33 +0900, Jorg Schuler a =C3=A9crit :

> I don't think libgpod should just read the ArtworkDB without claiming t=
o
> know what there should be. I believe the g_warning is placed well becau=
se it
> lets us know about changes, but we should not return the NULL and disca=
rd
> this mhni.

What do you suggest to use as a type ? IPOD_COVER_UNKNOWN, and let
libgpod users view/remove artowkr having this type, but not create new
ones ?

>=20
> The nasty thing here is that the calling function seems to use the NULL
> pointer in some bad way and segfaults... probably inside parse_mhod_3()
> (sorry, my gdb stopped working a long time ago, and I never could figur=
e out
> what is wrong).
>=20

Well, this needs to be fixed as well, or at least we need to understand
what happens instead of just papering over the bug by making sure it
can't be NULL. valgrind should be able to provide good backtraces too if
gdb doesn't work for you.

Christophe