This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
RE: [PATCHv4, MIPS] Add support for O32 FPXX and program header based ABI information
- From: Joseph Myers <joseph at codesourcery dot com>
- To: Matthew Fortune <Matthew dot Fortune at imgtec dot com>
- Cc: "'libc-alpha at sourceware dot org'" <libc-alpha at sourceware dot org>, "Moore, Catherine (Catherine_Moore at mentor dot com)" <Catherine_Moore at mentor dot com>, 'Andrew Pinski' <pinskia at gmail dot com>, "Rich Felker (dalias at libc dot org)" <dalias at libc dot org>, Rich Fuhler <Rich dot Fuhler at imgtec dot com>, "'macro at codesourcery dot com'" <macro at codesourcery dot com>
- Date: Fri, 2 Jan 2015 18:46:10 +0000
- Subject: RE: [PATCHv4, MIPS] Add support for O32 FPXX and program header based ABI information
- Authentication-results: sourceware.org; auth=none
- References: <6D39441BF12EF246A7ABCE6654B0235320F8DFA4 at LEMAIL01 dot le dot imgtec dot org> <alpine dot DEB dot 2 dot 10 dot 1412311718240 dot 26199 at digraph dot polyomino dot org dot uk> <6D39441BF12EF246A7ABCE6654B0235320F93196 at LEMAIL01 dot le dot imgtec dot org> <alpine dot DEB dot 2 dot 10 dot 1412312231120 dot 26199 at digraph dot polyomino dot org dot uk> <6D39441BF12EF246A7ABCE6654B0235320F93944 at LEMAIL01 dot le dot imgtec dot org>
On Fri, 2 Jan 2015, Matthew Fortune wrote:
> Joseph Myers <joseph@codesourcery.com> writes:
> > On Wed, 31 Dec 2014, Matthew Fortune wrote:
> >
> > > FAIL: conform/ISO/math.h/linknamespace
> >
> > I'd expect math.h linknamespace failures only for soft-float (and mips1
> > builds, because those use the dbl-64 sqrt implementation which uses
> > fegetround).
>
> = conform/ISO/math.h/linknamespace.out =
> [initial] __acos -> [libm.a(w_acos.o)] feraiseexcept
Should have been fixed by commit 0747f818110fa70323aed53fd7b50587adcbf64f.
> > > FAIL: conform/POSIX/fcntl.h/conform
> >
> > What are the conform/ failures? I've never run the conform/ tests for
> > MIPS as they're still native-only, but they probably indicate issues in
>
> Out of interest how do you run the testsuite if not natively? I got lots
> of failures from qemu user-mode testing especially with nptl tests when I
> last tried. N32 'stat*' is also very broken in qemu currently so N32 cross
> testing appears impossible until that is resolved.
test-wrapper="/some/where/scripts/cross-test-ssh.sh --timeoutfactor 50
some-board", with the build / source trees on a filesystem shared over NFS
(using physical hardware, with slow CoreFPGA processors, but this should
work with system-mode qemu as well to ssh to a virtual board).
> = N64 conform/POSIX/signal.h/conform.out =
>
> Testing for type of member sa_flags... FAIL
> Member "sa_flags" does not have the correct type. Compiler message:
Should now be fixed by commit 27dae0113e527aa2a94350d1f624c6e7c81139ce.
> = N64 conform/POSIX/termios.h/conform =
>
> Checking the namespace of "termios.h"... FAIL
> ------------------------------------------------------------------------
> Namespace violation: "TIOCSER_TEMT"
Should now be fixed by commit 253a59ccb571dc67b6dc303192d8d99378a35c75.
> = O32 conform/POSIX/fcntl.h/conform =
>
> Checking the namespace of "fcntl.h"... FAIL
> ------------------------------------------------------------------------
> Namespace violation: "pad"
> ------------------------------------------------------------------------
>
> = O32 conform/POSIX/mqueue.h/conform =
>
> Checking the namespace of "mqueue.h"... FAIL
> ------------------------------------------------------------------------
> Namespace violation: "pad"
Should now be fixed by commit 172019c0ecbf4714047e4ec233eab37628731d2b.
> (N64/O32/N32 unexpected pass)
> XPASS: conform/ISO11/complex.h/conform
> XPASS: conform/ISO11/stdalign.h/conform
> XPASS: conform/ISO11/stdnoreturn.h/conform
Expected whenever testing with GCC 4.7 or later - the XFAILs are there for
people using GCC 4.6 and can be removed when 4.6 is no longer supported.
> (O32/N32 unexpected pass)
> XPASS: conform/POSIX2008/fcntl.h/linknamespace
> XPASS: conform/POSIX2008/mqueue.h/linknamespace
> XPASS: conform/XOPEN2K/fcntl.h/linknamespace
> XPASS: conform/XOPEN2K/mqueue.h/linknamespace
> XPASS: conform/XOPEN2K/utmpx.h/conform
> XPASS: conform/XOPEN2K8/fcntl.h/linknamespace
> XPASS: conform/XOPEN2K8/mqueue.h/linknamespace
Except for utmpx.h, these XFAILs are removed by my patch
<https://sourceware.org/ml/libc-alpha/2014-12/msg00985.html> which is
pending review (they are present because the tests fail on x86_64, which
was the baseline for setting up linknamespace XFAILs).
> > > FAIL: inet/test-ifaddrs
> > > FAIL: inet/test_ifindex
> >
> > What are the failures here? It's odd for these to be architecture-
> > specific.
>
> Both of these fail to find any network interfaces even though there are
> 'lo' and 'mgmt0' listed by ifconfig. I'm running a modified 3.10.20
> kernel so it shouldn't be too old though I don't know exactly how it
> is modified! If there are no known issues here then I will try and
> find out what exactly is failing.
Maybe this is a page size issue? If so, you could retest the MIPS part of
the patch pointed to in my last comment in bug 16191, and commit it if it
works to fix the observed failures, and then file the separate bugs /
repost / ping for the IA64 / MicroBlaze parts of the patch?
--
Joseph S. Myers
joseph@codesourcery.com