Re: fatal() -> internal_error() jumbo patch

[jtc@redback.com (J.T. Conklin)]

>>>>> "Andrew" == Andrew Cagney <ac131313@cygnus.com> writes:
Andrew> Hello, I'm looking for comments.  Mainly on the new function
Andrew> internal_error().

It would have been a little easier had you split out that part of the
patch.  It took a bit of time to find amidst the boilerplate changes.

I think the idea is a good one.  Although, I must admit I've never
seen fatal() getting tripped in quite some years.

When internal_error() recurses three times, you write() to filedesc 3.
Did you mean 2?  Also that case calls then abort(), which has already
failed twice by that point.  Perhaps it should call _exit() or some
such, otherwise it may never exit.

Andrew> The new function asks the user before dumping core (quiting
Andrew> GDB) this gives the user a fighting chance of still being able
Andrew> to salvage something from a debug session.

IMO the question about dumping core and continuing should be separate.
I can envision answering yes to dump core and no to quit.  That way I
can dump the frotzed state of GDB (for later debugging), yet still try
to recover useful info from my debug session.

        --jtc

-- 
J.T. Conklin
RedBack Networks