Re: GCC 4.5.0 release candidate available

Re: GCC 4.5.0 release candidate available

[Jerome Quinn]

I just downloaded the 20100406 snapshot and tried to build, however,
could not get cloog-ppl to build.  I know it's not mandatory, but
there's an issue here.

I pulled down both ppl and cloog-ppl from the same location as the
snapshot, so ppl-0.10.2 and cloog-ppl-0.15.9.  On a Centos 5.4 x86_64
system, I built ppl and installed it in a custom location.  This is
gcc 4.1.2-46.

./configure --prefix=/scratch/usr

Then I configured cloog-ppl:

./configure --with-ppl=/scratch/usr
make
...
make[1]: Entering directory `/scratch/usr/jlquinn/cloog-
ppl-0.15.9'
/bin/sh ./libtool --tag=CC --mode=link gcc -Wall -fomit-frame-pointer
-g -O2  -L/scratch/usr/lib  -o cloog  cloog.o libcloog.la -lgmp
-lppl_c -lppl -lgmpxx
libtool: link: gcc -Wall -fomit-frame-pointer -g -O2 -o .libs/cloog
cloog.o  -L/scratch/usr/lib ./.libs/libcloog.so
/scratch/usr/lib/libppl_c.so /scratch/usr/lib/libppl.so -lgmp -lgmpxx
-Wl,-rpath -Wl,/usr/local/lib -Wl,-rpath -Wl,/scratch/usr/lib
./.libs/libcloog.so: undefined reference to `mp_get_memory_functions'
collect2: ld returned 1 exit status
make[1]: *** [cloog] Error 1

Am I missing something here?  The gcc install docs imply there's
nothing special needed to build these two items.  If there is, we
should probably include that.

Thanks,
Jerry

Re: GCC 4.5.0 release candidate available

[Marc Glisse]

On Tue, 6 Apr 2010, Jerome Quinn wrote:

> undefined reference to `mp_get_memory_functions'

This means your version of GMP is too old. configure should probably catch 
this.

-- 
Marc Glisse

Re: GCC 4.5.0 release candidate available

[Jerome Quinn]

On Tue, Apr 6, 2010 at 3:29 PM, Marc Glisse
<marc.glisse+gcc@normalesup.org> wrote:
> On Tue, 6 Apr 2010, Jerome Quinn wrote:
>
>> undefined reference to `mp_get_memory_functions'
>
> This means your version of GMP is too old. configure should probably catch
> this.

What is the minimum gmp for cloog-ppl?  Is it the same as for gcc?  My
system gmp is too old for gcc so I've got gmp 4.3.2 placed in the gcc
source tree per the instructions.  It would be nice to have cloog use
the gmp that will be built as part of the gcc bootstrap.

Is there a clean way of making that happen?  If not, the gcc docs
should probably mention that this is an issue and how to work around
it - i.e. build gmp and install rather than making it part of the gcc
build.

thanks,
Jerry

Re: GCC 4.5.0 release candidate available

[Jonathan Wakely]

On 6 April 2010 21:10, Jerome Quinn wrote:
>
> What is the minimum gmp for cloog-ppl?  Is it the same as for gcc?  My
> system gmp is too old for gcc so I've got gmp 4.3.2 placed in the gcc
> source tree per the instructions.  It would be nice to have cloog use
> the gmp that will be built as part of the gcc bootstrap.
>
> Is there a clean way of making that happen?

Can't you just put the cloog sources in-tree as well?

Re: GCC 4.5.0 release candidate available

[Jonathan Wakely]

On 6 April 2010 21:10, Jerome Quinn wrote:
>
> What is the minimum gmp for cloog-ppl?  Is it the same as for gcc?  My
> system gmp is too old for gcc so I've got gmp 4.3.2 placed in the gcc
> source tree per the instructions.  It would be nice to have cloog use
> the gmp that will be built as part of the gcc bootstrap.
>
> Is there a clean way of making that happen?

Can't you just put the cloog sources in-tree as well?

Re: GCC 4.5.0 release candidate available

[Jerome Quinn]

On Tue, Apr 6, 2010 at 4:27 PM, Jonathan Wakely <jwakely.gcc@gmail.com> wrote:
>> source tree per the instructions.  It would be nice to have cloog use
>> the gmp that will be built as part of the gcc bootstrap.
>>
>> Is there a clean way of making that happen?
>
> Can't you just put the cloog sources in-tree as well?

The install docs don't mention that possibility.  The last time I
tried to build with cloog, I think it wasn't supported in-tree.  I'll
try again...

Re: GCC 4.5.0 release candidate available

[Jonathan Wakely]

On 6 April 2010 22:02, Jerome Quinn wrote:
> On Tue, Apr 6, 2010 at 4:27 PM, Jonathan Wakely <jwakely.gcc@gmail.com> wrote:
>>> source tree per the instructions.  It would be nice to have cloog use
>>> the gmp that will be built as part of the gcc bootstrap.
>>>
>>> Is there a clean way of making that happen?
>>
>> Can't you just put the cloog sources in-tree as well?
>
> The install docs don't mention that possibility.  The last time I
> tried to build with cloog, I think it wasn't supported in-tree.  I'll
> try again...

I think it works as long as the directory is called "cloog" not "cloog-ppl"

I don't see any other way to get cloog to use a gmp you haven't built yet.

Re: GCC 4.5.0 release candidate available

[Jerome Quinn]

On Tue, Apr 6, 2010 at 5:11 PM, Jonathan Wakely <jwakely.gcc@gmail.com> wrote:
> On 6 April 2010 22:02, Jerome Quinn wrote:

>>>> Is there a clean way of making that happen?
>>>
>>> Can't you just put the cloog sources in-tree as well?
>>
>> The install docs don't mention that possibility.  The last time I
>> tried to build with cloog, I think it wasn't supported in-tree.  I'll
>> try again...
>
> I think it works as long as the directory is called "cloog" not "cloog-ppl"
>
> I don't see any other way to get cloog to use a gmp you haven't built yet.

If I try this the build still fails.  It's not picking up the ppl that is
in the gcc source tree.  The ppl dir is named ppl and the cloog dir is
named cloog.

It appears to be ignoring the ppl in the gcc source tree...


checking for sys/resource.h... yes
polylibgmp
Package gmp-include :
/scratch/usr/jlquinn/gcc-4.5.0-RC-20100406/host-x86_64-unknown-linux-gnu/gmp
Package gmp-library :
/scratch/usr/jlquinn/gcc-4.5.0-RC-20100406/host-x86_64-unknown-linux-gnu/gmp/.libs
checking whether gmp works... checking gmp.h usability... yes
checking gmp.h presence... yes
checking for gmp.h... yes
checking for __gmpz_init in -lgmp... yes
yes
checking for location of PolyLib... installed in standard location
checking for PolyhedronTSort in -lpolylibgmp... no
Can't find PolyLib.
checking for Parma Polyhedral Library (PPL)... installed at system prefix
checking ppl_c.h usability... no
checking ppl_c.h presence... no
checking for ppl_c.h... no
configure: error: Can't find PPL headers.
make[2]: *** [configure-stage1-cloog] Error 1
make[2]: Leaving directory `/scratch/usr/jlquinn/gcc-4.5.0-RC-20100406'
make[1]: *** [stage1-bubble] Error 2
make[1]: Leaving directory `/scratch/usr/jlquinn/gcc-4.5.0-RC-20100406'
make: *** [all] Error 2

GCC 4.5.0 release candidate available

[Richard Guenther]

A GCC 4.5.0 release candidate is available at:

ftp://gcc.gnu.org/pub/gcc/snapshots/4.5.0-RC-20100406/

Please test the tarballs and report any problems to Bugzilla.
CC me on the bugs if you believe they are regressions from
previous releases severe enough to block the 4.5.0 release.

Thanks,
Richard.

-- 
Richard Guenther <rguenther@suse.de>
Novell / SUSE Labs
SUSE LINUX Products GmbH - Nuernberg - AG Nuernberg - HRB 16746 - GF: Markus Rex

Re: GCC 4.5.0 release candidate available

[Jack Howarth]

On Tue, Apr 06, 2010 at 03:45:27PM +0200, Richard Guenther wrote:
> 
> A GCC 4.5.0 release candidate is available at:
> 
> ftp://gcc.gnu.org/pub/gcc/snapshots/4.5.0-RC-20100406/
> 
> Please test the tarballs and report any problems to Bugzilla.
> CC me on the bugs if you believe they are regressions from
> previous releases severe enough to block the 4.5.0 release.
> 
> Thanks,
> Richard.
> 
> -- 
> Richard Guenther <rguenther@suse.de>
> Novell / SUSE Labs
> SUSE LINUX Products GmbH - Nuernberg - AG Nuernberg - HRB 16746 - GF: Markus Rex
> 

Richard,
   Any chance we can get r157822 reapplied on gcc 4.5 branch 
along with...

http://gcc.gnu.org/ml/gcc-patches/2010-04/msg00129.html

to fix emutls? As can be seen from the testsuite run on
x86_64-apple-darwin10...

http://gcc.gnu.org/ml/gcc-testresults/2010-04/msg00444.html

when all the "dg-require-effective-target tls_native" tests
are converted to "dg-require-effective-target tls", we only
see failures in...

FAIL: gcc.dg/tls/opt-4.c scan-assembler tcc1@
FAIL: gcc.dg/tls/opt-4.c scan-assembler tcc2@
FAIL: gcc.dg/tls/opt-4.c scan-assembler-not tcc1[^@]
FAIL: gcc.dg/tls/opt-4.c scan-assembler-not tcc2[^@]
FAIL: gcc.dg/tls/section-1.c conflict with user-defined section (test for errors, line 12)

at -m32 and...

FAIL: gcc.dg/tls/section-1.c conflict with user-defined section (test for errors, line 12)

at -m64. Iain thinks the gcc.dg/tls/opt-4.c failures are expected
but the gcc.dg/tls/section-1.c failure might be showing up a
previously unknown bug in emutls.
            Jack

Re: GCC 4.5.0 release candidate available

[Richard Guenther]

On Tue, 6 Apr 2010, Jack Howarth wrote:

> On Tue, Apr 06, 2010 at 03:45:27PM +0200, Richard Guenther wrote:
> > 
> > A GCC 4.5.0 release candidate is available at:
> > 
> > ftp://gcc.gnu.org/pub/gcc/snapshots/4.5.0-RC-20100406/
> > 
> > Please test the tarballs and report any problems to Bugzilla.
> > CC me on the bugs if you believe they are regressions from
> > previous releases severe enough to block the 4.5.0 release.
> > 
> > Thanks,
> > Richard.
> > 
> > -- 
> > Richard Guenther <rguenther@suse.de>
> > Novell / SUSE Labs
> > SUSE LINUX Products GmbH - Nuernberg - AG Nuernberg - HRB 16746 - GF: Markus Rex
> > 
> 
> Richard,
>    Any chance we can get r157822 reapplied on gcc 4.5 branch 
> along with...
> 
> http://gcc.gnu.org/ml/gcc-patches/2010-04/msg00129.html
> 
> to fix emutls? As can be seen from the testsuite run on
> x86_64-apple-darwin10...
> 
> http://gcc.gnu.org/ml/gcc-testresults/2010-04/msg00444.html
> 
> when all the "dg-require-effective-target tls_native" tests
> are converted to "dg-require-effective-target tls", we only
> see failures in...
> 
> FAIL: gcc.dg/tls/opt-4.c scan-assembler tcc1@
> FAIL: gcc.dg/tls/opt-4.c scan-assembler tcc2@
> FAIL: gcc.dg/tls/opt-4.c scan-assembler-not tcc1[^@]
> FAIL: gcc.dg/tls/opt-4.c scan-assembler-not tcc2[^@]
> FAIL: gcc.dg/tls/section-1.c conflict with user-defined section (test for errors, line 12)
> 
> at -m32 and...
> 
> FAIL: gcc.dg/tls/section-1.c conflict with user-defined section (test for errors, line 12)
> 
> at -m64. Iain thinks the gcc.dg/tls/opt-4.c failures are expected
> but the gcc.dg/tls/section-1.c failure might be showing up a
> previously unknown bug in emutls.

Not before the 4.5.0 release.

Richard.