FW: (panic action)

[Andrew Bartlett]

David Lee wrote:
> 

> This thread seems very close to a thread about two weeks ago entitled
> "Automatic backtrace".  Its topic was automating a default "panic action"
> at configure time, taking into account the myriad quirks of OSes,
> compilers and debuggers.  The default would do its best to obtain a
> backtrace into a file in a simple and reliable way.  (Naturally, there
> would need to be override capability for the experts who want to dive
> deeper.)

It incresingly looking like somthing that would be very worthwhile.

> I think we reached the conclusion that it would be possible, and that a
> fair amount of table-driven work, at both configure-time and compile-time
> would be necessary.

That was certainly the conclusion as I understood it

> Any progress on this?  I cannot commit much time to it.  But if I knew
> that any such time would be likely to have a fruitful outcome and
> reception, then I'd see about warping time a little...

I recon that with the XFILE stuff we now have in HEAD we should be able
to hijack the raw file descriptors and fddup() (I think thats the
function) stderr to them, rather than to /dev/null as we do at the
moment.  

This will allow us to write a 'normal' panic action that doesn't need to
know about logfiles etc, but has smbd to the legwork getting it to
stable storage.  If sombody gets to this before me all the better!

Andrew Bartlett

-- 
Andrew Bartlett                                 abartlet at pcug.org.au
Manager, Authentication Subsystems, Samba Team  abartlet at samba.org
Student Network Administrator, Hawker College   abartlet at hawkerc.net
http://samba.org     http://build.samba.org     http://hawkerc.net