Re: current git segfaults, gdb backtraces included, pl;ease help

[Bastien Nocera <hadess@hadess.net>]

On Sun, 2009-02-22 at 17:16 +0200, Johan Hedberg wrote:
> Hi Jelle,
> 
> On Sun, Feb 22, 2009, Jelle de Jong wrote:
> > ==9308== Conditional jump or move depends on uninitialised value(s)
> > ==9308==    at 0x49AEEE0: (within /lib/i686/cmov/libc-2.7.so)
> > ==9308==    by 0x49A21F5: vsprintf (in /lib/i686/cmov/libc-2.7.so)
> > ==9308==    by 0x498A48A: sprintf (in /lib/i686/cmov/libc-2.7.so)
> > ==9308==    by 0x492B922: ba2str (bluetooth.c:88)
> > ==9308==    by 0x13262C: l2cap_get (btio.c:731)
> > ==9308==    by 0x13312F: get_valist (btio.c:983)
> > ==9308==    by 0x133447: bt_io_get (btio.c:1061)
> > ==9308==    by 0x48312A2: connect_event_cb (server.c:102)
> > ==9308==    by 0x130F37: server_cb (btio.c:180)
> > ==9308==    by 0x48A8F2C: (within /usr/lib/libglib-2.0.so.0.1800.4)
> > ==9308==    by 0x4872367: g_main_context_dispatch (in /usr/lib/libglib-2.0.so.0.1800.4)
> > ==9308==    by 0x4875A02: (within /usr/lib/libglib-2.0.so.0.1800.4)
> > ==9308==  Uninitialised value was created by a stack allocation
> > ==9308==    at 0x483121F: connect_event_cb (server.c:91)
> 
> This was actually a pretty stupid thing I should have noticed earlier.
> The parameter list to bt_io_get should end in BT_IO_OPT_INVALID but it
> was ending in NULL for this call. I also found two other places with
> problematic parameters to bt_io_get. All occurences have been fixed in
> git now.

Make BT_IO_OPT_INVALID be 0 in the enum, and you can use the GCC
sentinel feature. Otherwise, it would have to wait for:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28319

Cheers