This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: HPUX testsuite tweakage
> From: Zack Weinberg <zack@codesourcery.com>
>
> "Kaveh R. Ghazi" <ghazi@caip.rutgers.edu> writes:
>
> > Nice, this may help fix some spurious named section failures on irix6.
> > In addition to parse/attr-ctor1.C, I think gcc.dg/compat/sdata would
> > benefit too. Can you incorporate this update in there also? See:
> > http://gcc.gnu.org/ml/gcc-testresults/2004-04/msg00412.html
>
> Hoom, er, that's compat.exp which I don't know how to modify. cc:ing
> Janis for advice.
I tried sticking dg-require-named-sections in sdata-1_main.c but it
didn't like it. It said:
WARNING: compat.exp does not support dg-require-named-sections
Help would be appreciated, though this one in no way blocks your
current implementation.
> > However as you can see in the above test report, I run multiple passes
> > and I only get these failures with the o32 ABI. The other ABIs run
> > the named section tests just fine.
> >
> > So I'm wondering whether the cache variable
> > named_sections_available_saved is saved across multi-pass runs. I
> > wouldn't want to cache the result in that case.
>
> I have no idea. Would be good to find out.
> zw
Bad news, the testsuite harness caches the result across passes. So
the cached result of dg-require-named-sections depends on whether I
run the -32 pass first or not, with obvious bad results either way for
such testsuite entries.
Would you please insert an explanatory comment and avoid caching the
result? There's only a couple of such tests anyway so it's not a
critical optimization in the testsuite.
Thanks,
--Kaveh
--
Kaveh R. Ghazi ghazi@caip.rutgers.edu