Re: gcc 4.0 build failures on m68k, mips, and mipsel

Martin Pitt <martin(at)piware(dot)de> writes:
>>> However, Thiemo Seufer dealt with this and created a proper tas()
>>> function for PostgreSQL on mips, which also works with gcc 4. Of
>>> course this is much better, so I applied his patch in Debian.
>>> I attach the patch. Do you consider applying it upstream?
>>
>> Hm, does this patch actually compile as-is? I'd think you'd need to
>> have the MIPS definition of slock_t placed earlier than the assembly
>> code fragment.

> I attach the updated patch against 8.0.

OK, better, but I think not quite there yet. AFAICS the s_lock.c code
should simply go away, as there are no platforms for which it'd be used.
Also, the constraints for the inline assembler ought to include memory
and cc clobbers, according to the comments near the head of s_lock.h.
I've applied the attached modified patch to CVS HEAD, but can't test it
without a MIPS machine ... could you double-check it?

regards, tom lane