Re: Can't debug Threaded application

[Russell Shaw <rjshaw at netspace dot net dot au>]

Daniel Jacobowitz wrote:
> On Tue, Aug 30, 2005 at 03:06:39AM +1000, Russell Shaw wrote:
>
> > Daniel Jacobowitz wrote:
> >
> > > On Mon, Aug 29, 2005 at 12:08:53PM -0400, jo@atsresults.com wrote:
> > >
> > >
> > > > Does no one have answer to this problem?
> > > >
> > > > Is GDB useless for threaded applications now?
> > >
> > > Of course not.  No one's responded because no one has advice that you
> > > can't find in the list archives.  This almost always means something is
> > > broken on your installation.  You may want to report it to Trustix,
> > > since that's your distribution.
> >
> > I thought it was just me that had thread problems. In a program,
> > when i step over:
> >
> >  result  = gnome_vfs_directory_list_load (&dir_list, directory,
> >                      GNOME_VFS_FILE_INFO_DEFAULT);
> >
> > I get:
> >
> > (gdb) next
> >
> >  Program received signal SIG33, Real-time event 33.
> >  [Switching to Thread -1230148688 (LWP 3083)]
> >  0xb7a975a9 in poll () from /lib/tls/libc.so.6
> >
> > and execution seems to abort because i can't reach any
> > breakpoints after this point. The program was compiled
> > with gcc-3.4.4 -g -O0. Gdb is version 6.3.
>
>
> I recommend submitting a complete test case, since this sort of thing
> does work for many people.

It's too hard for me to isolate to a simple test case, because i
think it's caused by an earlier part of the program that sets up
threading, but i know that this library function is pretty ugly
(i was debugging someone elses program).
All i can do is submit the whole project, or else leave it till
one day when i'm writing my own threaded code and find the same
bug that i could make a simple test case for.