This is the mail archive of the
libffi-discuss@sourceware.org
mailing list for the libffi project.
Re: libffi rc3 on openbsd sparc/sparc64/mips64
- From: Richard Frith-Macdonald <richard at tiptree dot demon dot co dot uk>
- To: Sebastian Reitenbach <sebastia at l00-bugdead-prods dot de>
- Cc: libffi-discuss at sourceware dot org
- Date: Fri, 11 Feb 2011 14:35:12 +0000
- Subject: Re: libffi rc3 on openbsd sparc/sparc64/mips64
- References: <201102111526.06161.sebastia@l00-bugdead-prods.de>
On 11 Feb 2011, at 14:26, Sebastian Reitenbach wrote:
> Hi,
>
> attached the regression logs for libffi-rc3 on OpenBSD sparc/sparc64/mips64.
>
> patches to the testsuite for gcc-2.95 I created long time ago, that did not
> maded it into libffi:
> http://cygwin.ru/ml/libffi-discuss/2008/msg00161.html
> therefore the regression tests for sparc are bit meaningless.
>
> patches from the libffi port to actually compile libffi on OpenBSD mips64
> patch-src_mips_ffitarget_h
> patch-src_mips_ffi_c
>
> patches to enable libffi compilation on hppa, mips64(el) and powerpc
> patch-configure
>
> patches to the testsuite that are in the OpenBSD ports tree:
> patch-testsuite_libffi_call_float2_c
> patch-testsuite_libffi_call_return_uc_c
>
> there is also a patch for openbsd arm in the ports tree, but it did not
> applied to the new version, and I don't have an arm box around.
>
> libffi is used by my gnustep ports. gnustep-base based ports seem to work fine
> at least on mips64 tested so far, starting any GNUstep GUI application, this
> fails, since they seem to make use of stuff that doesn't pass the regression
> tests. On sparc64 some of the applications start up, but tend to crash early
> when trying to use it.
Would this be when the code passes structure arguments where the structures contain floats/doubles?
I seem to remember that passing those structures used to fail on 64bit intel until the last release of libffi, but I thought a regression test to catch it had been added then.