Re: mc ongoing bug

[Ronald de Man <deman win tue nl>]

On Sat, Apr 03, 1999 at 04:05:30AM +0100, Mike McEwan wrote:
> Ronald de Man <deman@win.tue.nl> writes:
> 
> > I experienced this today for the first time. I decided to make a backtrace,
> > so recompiled gmc with options -g -O2. Now the bug was gone. My compiler
> > is pgcc-1.1.1 and by default I compile with -O6. What compiler and
> > compilation flags are you using?
> 
>   I'm also using pgcc-1.1.1, however I have been using the default
> CFLAGS provided in the gmc Makefile. I believe this is `-g -O2' (or
> perhaps just `-g' with no optimisation) by default - no? The same
> flags which do *not* give you the problem.

The default is '-g -O2', and yes, this is giving me no problems.

>  
> > I'm not saying that it must be a compiler bug. It might be a subtle bug
> > in gmc that's only triggered by some compilers (for example an uninitialized
> > variable, whose initial value will depend on the compiler used).
> 
>   Well, I guess we're both pgcc-1.1.1 - wonder what others with the
> problem are using?
> 
> > If you want this fixed, you should definitely try to provide the
> > developers with a stacktrace of the coredump.
> 
>   But I'm not getting a coredump??

Then either you're running gmc from a non-writable directory, or
you have a limit on the size of coredumps.

You don't need the coredump itself as long as you can reproduce the
problem (which you can). Just do:

gdb
file /usr/local/bin/gmc
attach <pid of gmc>
continue
<now make gmc crash by right clicking on an icon>
bt

Ronald