1008 segfaults in genattr

[Robert Lipe]

> >   What's the simplest program you can provide that shows the problem?
> 
> >From memory:

[ hello, world ] 

Hmm.  That's pretty simple, all right. :-)

So this is definitely at least a different form of the problem than 
what we saw on OpenServer.

> This dumps core in (stop me if you've heard this before) 
> __do_global_dtors_aux, the 2nd time around.

Does the program I attached before do anything "funny" when built
with the native compiler?      If not, we're almost certainly barking
up the wrong tree.

Actually, what happens if you make the last line of "hello, world" into
"_exit(0);"?   Be sure to note the leading underbar.   This would bypass
the whole destructor/.fini thing.

> Jeff, I doubt that my platform includes sco5.h, even though it's now owned 
> by SCO.  I was proposing that I test a similar modification to my own 
> configuration header file.

Though Unixware and OpenServer have many similarities (the native 
compilers and libraries both come from The Organizatino Formerly Known
As ATT/USL/USG/Univel/Novell) they are, from gcc's view, distinct 
targets.   If you're picking up sco5.h on a Unixware target, I'd expect
it to spontaneously combust.

Though this isn't the "right" way to do this, edit config/i386/sysv4.h
(a.k.a. your tm.h file) to include an "#undef HAVE_ATEXIT" somewhere
after it's included "../svr4.h".   You'll have to rebuild (at least)
crt* and probably libgcc.a before testing to see if this cures what
ails you.


If this doesn't make the difference, then there's probably a problem 
with the way .fini is being built.

RJL