On Tue, 2008-08-12 at 01:12 +0200, Dominique Dhumieres wrote:
> > but nothing complains and it seems to work fine.
>
> The tests gcc.target/i386/pr32000-2.c and return-3.c are run
> (and fail) on i686-apple-darwin9 which does not support decimal
> floating point. So I think the dg-require is not properly
> enforced.
I read the fine manual and was reminded that dg-require-effective-target
takes a single effective-target keyword as an argument and not an
effective-target expression. Tests that need an expression must instead
use dg-skip-if, as in
{ dg-skip-if "reason for skipping" { ! { ilp32 && dfp } } }
I'm testing a patch to error out for using the wrong number of arguments
with dg-require-effective-target. I could instead modify it to accept
an expression but don't see a good reason for doing that Both of these
directives were added before target expressions were supported, with
dg-skip-if supporting a list of targets and dg-require-effective-target
supporting a single effective-target keyword.
Janis
