RFC: Stack saving/tracing functionality.

[Poul-Henning Kamp]

In message <200506111546.j5BFkToq011515 at gate.bitblocks.com>, Bakul Shah writes:
>> ... but it would be neat if it could also save/print userland stacks
>> so that we could get tracebacks from abort()'ing userland programs.
>
>Along these lines; wouldn't it be neat if there was a sysctl
>to leave a segfaulted or aborted process around so that you
>can attach a debugger to it and find out what went wrong (and
>may be even correct it!)?  Debugging a live process (even if
>fatally injured) yields more clues as you can poke around at
>its I/O connections, its caller process etc.  A separate
>program can be used to create a coredump if you really wish
>to preseve the dead body for later autopsy.

Ideally a coredump should include anything sockstat and fstat can
tell you about the process.

/me longs for the MVS dumps level of details...

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.