Re: 1008 segfaults in genattr

[Robert Lipe <robertl at dgii dot com>]

>   > That is part of the ELF spec, and I'm pretty sure that Unixware
>   > does exactly what you're asking of it.
> And since all svr4 systems are supposed to be using ELF, in theory we
> should be able to count on .fini support.

That, that does indeed sound like my assessment - at least in theory. 

> However, some targets (dgux, sh-elf, others?) seem to want to disable .fini
> support from svr4.h, and thus would need ATEXIT.  There may also be some targets
> that include svr4.h, but don't use crt*, so they'd need ATEXIT too.

Do they actively *want* to disable this, or is there some habit or
intertia involved?

> What a *()&@#$ mess.

And I thought I just didn't understand... :-)

> Considering that most targets seem to be mis-behaving in only minor ways
> with HAVE_ATEXIT & .fini support enabled, I think the safest way to go is
> to create a new target file for unixware which sucks in svr4.h, then
> undefs HAVE_ATEXIT.

acs, what is the exact triple generated by 'config.guess' for your
target?

Do we really want to try to make it *just* unixware here, or all
x86 SVR4 mutants?  This would include (at least) Solaris/x86, MP-RAS,
UHC, Microport, and probably a few others.    My inclination is to 
suspect that we should probably just #undef it for all x86 SVR4's,
so it could just #undef it in i386/sysv4.h if you're not comfortable
doing it for all the SVR4s.

Yes, doing it this way you do risk  [ breaking | fixing ] other ports.

Of course, I will defer to the elders on this if challenged.


RJL