This is the mail archive of the
mailing list for the GCC project.
Re: [PATCH] Fix bootstrap fail on spu vs. ira.c/genpreds
Jakub Jelinek wrote:
> On Mon, Apr 27, 2009 at 12:36:52PM -0400, Vladimir Makarov wrote:
>>> Your solution is clean, but if anyone ever wants to call
>>> regclass_for_constraint again in the future we might run into the same problem
>>> again, might we not? I don't really mind which patch goes in, it's your call.
>> There is a sense in your words too. So on the second thought, I'd go
>> for your patch. Unfortunately for this case, I am not a maintainer of
>> this part of compiler. So my ok is not an approval for committing the
>> patch. I think Ian could do that. I am CCing the email to him.
> Wouldn't it be better to just
> #define regclass_for_constraint(x) NO_REGS
> then in the generated header in that case?
I was just being consistent with the way the other functions in tm-preds.h
get macro wrappers. I didn't find any direct (unwrapped) calls of a handful
of the functions that I grepped for in the rest of the source, so I think it's
a convention to only access those functions through the wrappers, broken by
(as far as I could tell) just this single instance in ira.c.