multithreaded gdbserver again. Should I use gdb5.3, or a snapshot?

multithreaded gdbserver again. Should I use gdb5.3, or a snapshot?

[Dan Kegel]

Hi all,
I had the multithreaded gdbserver from gdb-5.3 working, but
never tried really debugging anything with it.  I recently
brought it back up, and tried it out a bit.  I ran into the usual
set of newbie problems, namely:

* I didn't have solib-absolute-prefix set properly;
   this was made clear with the commands
     break main
     cont
     info shared
   which showed the wrong path.  I gave the command
     set solib-absolute-prefix /opt/cegl-1.5/hardhat/devkit/sh/sh4_le/target
   and that helped gdb find the shared libraries

* The libthread_db.so on the target was stripped, which seems to have
   caused SIG32 messages.  I made sure both gdb and gdbserver had
   the same unstripped libthread_db.so.  I think that got rid of
   the SIG32's, but...

Oddly, gdb doesn't seem to automatically load the shared libraries,
even though auto-solib-add is on.  I have to give the command
     shared
after hitting the breakpoint at main for the libraries to load,
else I get that SIG32 message.

Maybe I don't understand how shared libraries get loaded into
gdb.  I didn't think I should have to give the 'shared' command
normally.  Any idea what's up there?

And even then, I seem to get the SIGTRAP message referred to in
http://sources.redhat.com/ml/gdb/2003-03/msg00227.html
when I step a bit after hitting a breakpoint in a thread.
I tried applying the PREPARE_TO_PROCEED patch Daniel mentioned in
http://sources.redhat.com/ml/gdb/2003-03/msg00228.html,
and the original patch
http://sources.redhat.com/ml/gdb-patches/2002-08/msg01056.html
but neither of them seemed to apply cleanly against gdb-5.3.
Each had at least one reject.  I fixed up the rejects,
but the resulting debugger didn't work too well.

I'm sure my difficulties are because I'm sleepy, but perhaps
it's worth asking: should I stick with gdb5.3, or is a more
recent snapshot a better bet?  I'm running gdb on x86 linux,
and I'm running gdbserver on ppc405, ppc750, and sh4 linux.

Thanks,
Dan

-- 
Dan Kegel
http://www.kegel.com
http://counter.li.org/cgi-bin/runscript/display-person.cgi?user=78045

Re: multithreaded gdbserver again. Should I use gdb5.3, or a snapshot?

[Daniel Jacobowitz]

On Mon, Apr 07, 2003 at 11:15:51PM -0700, Dan Kegel wrote:
> Hi all,
> I had the multithreaded gdbserver from gdb-5.3 working, but
> never tried really debugging anything with it.  I recently
> brought it back up, and tried it out a bit.  I ran into the usual
> set of newbie problems, namely:
> 
> * I didn't have solib-absolute-prefix set properly;
>   this was made clear with the commands
>     break main
>     cont
>     info shared
>   which showed the wrong path.  I gave the command
>     set solib-absolute-prefix /opt/cegl-1.5/hardhat/devkit/sh/sh4_le/target
>   and that helped gdb find the shared libraries
> 
> * The libthread_db.so on the target was stripped, which seems to have
>   caused SIG32 messages.  I made sure both gdb and gdbserver had
>   the same unstripped libthread_db.so.  I think that got rid of
>   the SIG32's, but...
> 
> Oddly, gdb doesn't seem to automatically load the shared libraries,
> even though auto-solib-add is on.  I have to give the command
>     shared
> after hitting the breakpoint at main for the libraries to load,
> else I get that SIG32 message.
> 
> Maybe I don't understand how shared libraries get loaded into
> gdb.  I didn't think I should have to give the 'shared' command
> normally.  Any idea what's up there?

That suggests that something is wrong with your dynamic linker on the
target; is that stripped too?

It doesn't matter if the copies on the target are stripped; also, it
shouldn't matter if libthread_db.so is stripped; but libpthread.so and
ld.so on the host should be unstripped.

> And even then, I seem to get the SIGTRAP message referred to in
> http://sources.redhat.com/ml/gdb/2003-03/msg00227.html
> when I step a bit after hitting a breakpoint in a thread.
> I tried applying the PREPARE_TO_PROCEED patch Daniel mentioned in
> http://sources.redhat.com/ml/gdb/2003-03/msg00228.html,
> and the original patch
> http://sources.redhat.com/ml/gdb-patches/2002-08/msg01056.html
> but neither of them seemed to apply cleanly against gdb-5.3.
> Each had at least one reject.  I fixed up the rejects,
> but the resulting debugger didn't work too well.
> 
> I'm sure my difficulties are because I'm sleepy, but perhaps
> it's worth asking: should I stick with gdb5.3, or is a more
> recent snapshot a better bet?  I'm running gdb on x86 linux,
> and I'm running gdbserver on ppc405, ppc750, and sh4 linux.

A more recent snapsot won't help you.  I've been too busy with other
projects to pursue the change to the remote protocol needed to fix the
SIGTRAPs.  Sorry.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer