This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Fix PR30957
- From: Hans-Peter Nilsson <hp at bitrange dot com>
- To: Revital1 Eres <ERES at il dot ibm dot com>
- Cc: gcc-patches at gcc dot gnu dot org, Janis Johnson <janis187 at us dot ibm dot com>, Zdenek Dvorak <rakdver at atrey dot karlin dot mff dot cuni dot cz>
- Date: Mon, 7 May 2007 07:39:01 -0400 (EDT)
- Subject: Re: [PATCH] Fix PR30957
- References: <OF1026C0EE.12F8E0E3-ONC22572D4.00156B18-C22572D4.0019230C@il.ibm.com>
On Mon, 7 May 2007, Revital1 Eres wrote:
> gcc-patches-owner@gcc.gnu.org wrote on 06/05/2007 18:56:51:
Heh, mailer gotcha?
> > On Sun, 22 Apr 2007, Revital1 Eres wrote:
> >
> > >(See attached file: pr30957-1.c.txt)
> >
> > I can't quote it by simple means in this mailer, as it was
> > attached rather than in-text, but did you really mean for the
> > test-case to xfail for powerpc-spe and vax (and why the vax)
> > rather than skip the test for them?
> >
> > Or rather: is the better choice not to skip the test for these
> > targets? I'm asking because the test fails for cris-elf as
> > well, and I suspect you don't know the test-machinery to express
> > the skipping. Perhaps best a check_effective_target_* thingy
> > (grep in testsuite/lib/*.exp) to express what the target must
> > have in order for the test-case to be valid?
> >
>
> I took the powerpc-spe and vax targets from ieee.exp script as
> described in - http://gcc.gnu.org/ml/gcc-patches/2007-04/msg01772.html.
> The intention was to disable the test on machines with no hardware
> support for IEEE arithmetic thus skipping those targets indeed seems a
> better solution.
Many more toolchains/chips than powerpc-spe and vax have no
hardware support for floating point arithmetic at all, and CRIS
is one of them. Several toolchains have multilibbed support for
hardware floating point present/absent, so a regular target
specifier is wrong as well.
> I am not sure how to skip just one testcase though;
> does adding .x file is the right way?
No, a check_effective_target_* perhaps called
check_effective_target_ieee_hw_fp. You could check an RTL dump
for calls in a function adding two floating point numbers.
> I am also not sure I understand
> why this test fail on cris-elf as well.
No hardware FP.
brgds, H-P