dmalloc and __throw

[Michael Vance <briareos at lokigames dot com>]

I'm running into nasty nasty problems with the last weeks' worth of
daily CVS builds. All of them seg fault inside of a call to
_chunk_alloc() inside of an fopen() call. The code in question has not
changed, nor or is there any different behaviour in terms of global
constructors (the code is among the first executed). For what it's
worth, when I run the program linked against malloc, or with
MALLOC_CHECK_ set, it does *not* segfault, but it does do very very
strange things. First, the data inside of the object being constructed
goes from perfectly fine to hopelessly garbled as I immediately step
out of the constructor. Second, with a break set in dmalloc_error(), I
get breaks inside of __throw--some for 0-byte mallocs (fine), but
others for free( 0x0 ) attempts--bad.

While I would love to contribute test cases, I haven't been able to
build any that fail and the code in question is not open. Mostly, I
just need to know whether I'm on crack, my binutils are fubarred, I
need to look harder for problems dmalloc wouldn't catch, or if there
really may be nasty problems inside of egcs that I'm exposing with
this 500,000 line project.

m.

-- 
Programmer                         "I wrote a song about dental floss,
Loki Entertainment Software         but did anyone's teeth get cleaner?"
http://lokigames.com/~briareos/              - Frank Zappa, re: the PMRC