On Mon, Mar 03, 2003 at 08:48:12AM -0700, Marc Aurele La France wrote:
>On Sun, 2 Mar 2003, David Dawes wrote:
>
>> On Sat, Mar 01, 2003 at 08:27:49PM -0700, Marc Aurele La France wrote:
>> >On Sat, 1 Mar 2003, Mark Vojkovich wrote:
>
>> >>   setjmp is a *macro* (for __sigsetjmp) defined in /usr/include/setjmp.h.
>> >> This is libc 2.2. so it doesn't set HAS_GLIBC_SIGSETJMP.
>> >> SYMCFUNCALIAS chokes on this.  This is gcc 2.95.3.
>
>> >>    I think the HAS_GLIBC_SIGSETJMP set logic is wrong.
>
>> >You've got glibc 2.2.1, I'll guess.  The #if's should be looking for glibc
>> >< 2.2.2, not 2.2, although a host.def override is available (see
>> >xfree86.cf).
>
>> OK, so using the version macros in <features.h> isn't good enough here,
>> and it has to be done with the imake LinuxCLib*Version parameters instead.
>
>I don't think that's necessary.  It is simpler to #define HAS_GLIBC_SIGSETJMP
>for all of glibc 2.2.*, which is, in part, what I'll be committing
>shortly, after I iron out my libc5 problem.

So that would mean using __sigsetjmp(env, 0) on all glibc 2.x.  I guess
that has to work for compatibility reasons.  It's definitely cleaner
and lower-impact than trying to keep track of the two cases separately.
As we've seen, the host.def override was incompletely implemented in
4.3.0.

David
-- 
David Dawes
Release Engineer/Architect                      The XFree86 Project
www.XFree86.org/~dawes
_______________________________________________
Devel mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/devel

Reply via email to