Remote Debugging Of Multithreaded Application on PPC 82xx

Remote Debugging Of Multithreaded Application on PPC 82xx

[Frank.Christ]

I am trying to setup a remote debugging session, so that I can debug my
multithreaded
application running on PPC 82xx and use gdb on my Intel host (i686, SuSE 7.0).

I downloaded gdb version 5.3 and compiled gdb on the host (with target
powerpc-linux).
I also cross-compiled gdbserver (with option powerpc-linux).

I have /lib/libthread_db.so.1 on the 8xx
  have the target filesystem layout available on the host
  have solib-absolute-prefix set to point to the target libraries
(solib-absolute-prefix=/opt/elinos/cdk/ppc/82xx/libc6/powerpc-linux)

I can start the gdbserver on the target and gdb on the host and have a
connection by tcp/ip.
The application is multithreaded and uses the libpthread library. I can put
breakpoints in
every thread and break into my code. "info threads" gives all my threads and
everything
seems fine.

However:
I can step through the main thread, but I cannot step through all my other
threads. When
I try to step with next in a client thread I get the following message:

-------------------------------------
 (gdb) next

Program received signal SIGTRAP, Trace/breakpoint trap.
[Switching to Thread 1024]
0x0feee8a4 in nanosleep () from
/opt/elinos/cdk/ppc/82xx/libc6/powerpc-linux/lib/libc.so.6
------------------------------------

Any hints what my be wrong with my configuration?

Thanks in advance,
Frank Christ

Re: Remote Debugging Of Multithreaded Application on PPC 82xx

[Daniel Jacobowitz]

On Fri, Mar 14, 2003 at 03:03:25PM +0100, Frank.Christ@RSBICK.rohde-schwarz.com wrote:
> I am trying to setup a remote debugging session, so that I can debug my
> multithreaded
> application running on PPC 82xx and use gdb on my Intel host (i686, SuSE 7.0).
> 
> I downloaded gdb version 5.3 and compiled gdb on the host (with target
> powerpc-linux).
> I also cross-compiled gdbserver (with option powerpc-linux).
> 
> I have /lib/libthread_db.so.1 on the 8xx
>   have the target filesystem layout available on the host
>   have solib-absolute-prefix set to point to the target libraries
> (solib-absolute-prefix=/opt/elinos/cdk/ppc/82xx/libc6/powerpc-linux)
> 
> I can start the gdbserver on the target and gdb on the host and have a
> connection by tcp/ip.
> The application is multithreaded and uses the libpthread library. I can put
> breakpoints in
> every thread and break into my code. "info threads" gives all my threads and
> everything
> seems fine.
> 
> However:
> I can step through the main thread, but I cannot step through all my other
> threads. When
> I try to step with next in a client thread I get the following message:
> 
> -------------------------------------
>  (gdb) next
> 
> Program received signal SIGTRAP, Trace/breakpoint trap.
> [Switching to Thread 1024]
> 0x0feee8a4 in nanosleep () from
> /opt/elinos/cdk/ppc/82xx/libc6/powerpc-linux/lib/libc.so.6
> ------------------------------------
> 
> Any hints what my be wrong with my configuration?

You may want to try the PREPARE_TO_PROCEED patch which I posted to
gdb-patches yesterday; you can find the list archives on
http://sources.redhat.com/ml/gdb-patches/2003-03/.

There are also some bugs related to stepping alternate threads that
require a change to the remote protocol, but I didn't get a consensus
on how to solve them yet (and I ran out of time to pursue it; I'll try
again soon).

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer