code in cvs is broken

[Derek Atkins]

linas@linas.org (Linas Vepstas) writes:

> > -derek, who is trying to track down an apparent memory corruption bug.
> 
> I guess libefence didn't help.

No, efence required way too much memory, at least when I compiled the
whole program.  I changed over to using dmalloc (although I might
still have used efence on my module because I was stupid).

As I mentioned in a more recent email, I've tracked it down to an
unbalanced gtk_main() (without gtk_main_quit()) that was destroying
some of my widgets from under me.

> Note that gdb allows you to 'watch' memory locations for read/write
> access.  But its *extremely* slow -- like hours for a few cpu-secsonds
> real time.   I dunno if its broken, or is designed for use with a hardware
> debugger behind it.

I tried that, but it wasn't helping.

> You might try a stunt of unmapping the page which you think is getting
> corrupted, and then seeing who faults on that page access.

Hrm, I hadn't thought of that; the problem is that I really wanted
it to fail when it was being freed.  Unfortunately there is no
way (that I know) to say:
	break in g_free when the argument is 0xXXXXX

-derek

-- 
       Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
       Member, MIT Student Information Processing Board  (SIPB)
       URL: http://web.mit.edu/warlord/    PP-ASEL-IA     N1NWH
       warlord@MIT.EDU                        PGP key available