Re: netbsd-5 _lwp_park() SIGSEGV

[Andrew Doran <ad%netbsd.org@localhost>]

On Fri, Jan 16, 2009 at 01:57:50AM -0500, Matthew Mondor wrote:

> On Wed, 7 Jan 2009 20:46:02 +0000
> Andrew Doran <ad%netbsd.org@localhost> wrote:
> 
> > This doesn't appear to be the thread that has crashed. Have you looked
> > at the other threads in the core?
> 
> You were right; I could do further debugging today and noticed that the
> threads appeared to behave well despite their rather complex
> interactions, and that the problem is related to another issue in the
> main thread (LWP 1).
> 
> Btw, is it expected that gdb threading information is only available
> for core dumps?

Our current gdb does not have the necessary code to deal with threads in a
running program, and to my knowledge nobody wants to write that.

All the kernel infrastructure is in place to support threaded debugging but
the user<->kernel interface (ptrace) does not support multiple threads
reliably.

The informal plan is to kill ptrace() completely and provide Sun's threaded
/proc interface to the debugger. The decision was made on the basis that the
interface works and no additional gdb code needs to be written. I am not
aware of anyone working on it at the moment.

Thanks,
Andrew