This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [RFA:] fix lack of #ifdef HAVE_C99_RUNTIME in builtins-20.c and elsewhere


> From: Geoff Keating <geoffk@apple.com>
> Date: Thu, 5 Jul 2007 17:39:08 -0700

> On 05/07/2007, at 5:19 PM, Hans-Peter Nilsson wrote:
> >>> +proc check_effective_target_c99_math { } {

> I agree with Kaveh here; in each of these cases what you want to do is  
> to #include builtins-config.h and add options lines as
> appropriate.

No, I *don't* want to do that *only* and I already mentioned
why: the tests *besides* builtins-20; to wit builtins-59.c,
builtins-61.c and builtins-62.c, are performed by grepping dump
files, and that can't be helped by an include and stuff #ifdef'd
out.  That only helps builtins-20.c.  Those tests need a
dejagnu-test way to be skipped, and that way is
check_effective_target_c99_math.

> It's an error in the testsuite to be testing C99 functions without  
> switching them on.

Of course, the right options and a "#include builtins-config.h"
may have to be added *in addition* to
check_effective_target_c99_math, but that function is still
needed.

> ... so, sorry, still not OK.

IMHO, it's an error to test the wrong way, so maybe the
test-cases should be changed to target-specific and restricted
to certain high-profile primary targets.

And some of the tests require specific patterns IIRC like sin,
cos and tan to pass, not just C99 math.  The tests should be
rewritten, but I've lost interest.  Let them fail.

brgds, H-P


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]