Re: [PPL-devel] [CLooG-dev] 'make check; errors in ppl-0.10.2 and 'gcc'
Hi,
I think that this is more relevant to be posted to the PPL mailing list rather than the CLooG development list.
On Wed, Aug 25, 2010 at 12:38, sergstesh sergstesh@gmail.com wrote:
Hello,
while building ppl-0.10.2 I've noticed that 'make check' produces a number of errors (and ppl-0.11' 'make check' is clean IIRC).
OTOH, 'gcc' uses ppl-0.10.2.
So, are the 'make check' errors in ppl-0.10.2 significant and do they translate into 'gcc' bugs ?
I do not think that they would translate into gcc bugs.
If necessary, I can post screen output of 'make check' in question.
Please send the error messages that you are seeing, together with more information about the machine and OS that you are using.
Thanks, Sebastian
On 8/25/10, Sebastian Pop sebpop@gmail.com wrote:
Hi,
I think that this is more relevant to be posted to the PPL mailing list rather than the CLooG development list.
On Wed, Aug 25, 2010 at 12:38, sergstesh sergstesh@gmail.com wrote:
Hello,
while building ppl-0.10.2 I've noticed that 'make check' produces a number of errors (and ppl-0.11' 'make check' is clean IIRC).
OTOH, 'gcc' uses ppl-0.10.2.
So, are the 'make check' errors in ppl-0.10.2 significant and do they translate into 'gcc' bugs ?
I do not think that they would translate into gcc bugs.
If necessary, I can post screen output of 'make check' in question.
Please send the error messages that you are seeing, together with more information about the machine and OS that you are using.
Thanks, Sebastian
Please find 'make check' screen output attached.
Error messages are apparently at the end, but maybe some other info is necessary, so I decided to attach the whole file.
Thanks, Sergei.
On 08/25/10 21:21, Sergei Steshenko wrote:
On 8/25/10, Sebastian Popsebpop@gmail.com wrote:
I think that this is more relevant to be posted to the PPL mailing list rather than the CLooG development list.
On Wed, Aug 25, 2010 at 12:38, sergsteshsergstesh@gmail.com wrote:
Hello,
while building ppl-0.10.2 I've noticed that 'make check' produces a number of errors (and ppl-0.11' 'make check' is clean IIRC).
OTOH, 'gcc' uses ppl-0.10.2.
So, are the 'make check' errors in ppl-0.10.2 significant and do they translate into 'gcc' bugs ?
I do not think that they would translate into gcc bugs.
If necessary, I can post screen output of 'make check' in question.
Please send the error messages that you are seeing, together with more information about the machine and OS that you are using.
Please find 'make check' screen output attached.
Error messages are apparently at the end, but maybe some other info is necessary, so I decided to attach the whole file.
Dear Sergei,
as Sebastian anticipated, we need more information about the machine and OS you are using. We also need to know the versions of GCC, GMP and GLPK. The problem shown by the log file you attached concerns ppl_lpsol, an utility using the MIP_Problem class via the C interface of the PPL (the same interface used by GCC). In contrast, the same log file shows that testing via the C++ interface in tests/MIP_Problem/ succeeds. Looking forward to receive more information,
Roberto
On 8/26/10, Roberto Bagnara bagnara@cs.unipr.it wrote:
On 08/25/10 21:21, Sergei Steshenko wrote:
On 8/25/10, Sebastian Popsebpop@gmail.com wrote:
I think that this is more relevant to be posted to the PPL mailing list rather than the CLooG development list.
On Wed, Aug 25, 2010 at 12:38, sergsteshsergstesh@gmail.com wrote:
Hello,
while building ppl-0.10.2 I've noticed that 'make check' produces a number of errors (and ppl-0.11' 'make check' is clean IIRC).
OTOH, 'gcc' uses ppl-0.10.2.
So, are the 'make check' errors in ppl-0.10.2 significant and do they translate into 'gcc' bugs ?
I do not think that they would translate into gcc bugs.
If necessary, I can post screen output of 'make check' in question.
Please send the error messages that you are seeing, together with more information about the machine and OS that you are using.
Please find 'make check' screen output attached.
Error messages are apparently at the end, but maybe some other info is necessary, so I decided to attach the whole file.
Dear Sergei,
as Sebastian anticipated, we need more information about the machine and OS you are using. We also need to know the versions of GCC, GMP and GLPK. The problem shown by the log file you attached concerns ppl_lpsol, an utility using the MIP_Problem class via the C interface of the PPL (the same interface used by GCC). In contrast, the same log file shows that testing via the C++ interface in tests/MIP_Problem/ succeeds. Looking forward to receive more information,
Roberto-- Prof. Roberto Bagnara Applied Formal Methods Laboratory Department of Mathematics, University of Parma, Italy http://www.cs.unipr.it/~bagnara/ mailto:bagnara@cs.unipr.it
This is the info:
" uname -a Linux amdam2 2.6.27.48-0.2-default #1 SMP 2010-07-29 20:06:52 +0200 i686 athlon i386 GNU/Linux "
- this is 32 bit SUSE-11.1 in simple English.
Regarding versions - I am building as much as possible from sources with a tool of mine, so the versions can be seen in the attached autogenerated 'config_wrapper.sh' script used to run 'confugure'.
I have also attached 'configure' screen output - please see it in 'configure.log' file.
Thanks, Sergei.
Another piece of info I forgot to mention.
In the end I need to build gcc-4.4.4. I am building it in two stages in addition to whatever stages used by 'gcc' build mechanism itself.
First I build gcc-4.4.4 without 'cloog', then using this clogg-less gcc-4.4.4 I build 'cloog' and its dependencies, and then having built 'cloog' I build the final gcc-4.4.4 with 'cloog' - still using the cloog-less gcc-4.4.4.
So now you know the gory details of the 'gcc' I'm using to build the stuff :).
Again, IIRC, ppl-0.11 builds just fine using this scheme, but ppl-0.10.2 produces 'make check' errors we are discussing in this thread.
Thanks, Sergei.
On 8/27/10, Sergei Steshenko sergstesh@gmail.com wrote:
On 8/26/10, Roberto Bagnara bagnara@cs.unipr.it wrote:
On 08/25/10 21:21, Sergei Steshenko wrote:
On 8/25/10, Sebastian Popsebpop@gmail.com wrote:
I think that this is more relevant to be posted to the PPL mailing list rather than the CLooG development list.
On Wed, Aug 25, 2010 at 12:38, sergsteshsergstesh@gmail.com wrote:
Hello,
while building ppl-0.10.2 I've noticed that 'make check' produces a number of errors (and ppl-0.11' 'make check' is clean IIRC).
OTOH, 'gcc' uses ppl-0.10.2.
So, are the 'make check' errors in ppl-0.10.2 significant and do they translate into 'gcc' bugs ?
I do not think that they would translate into gcc bugs.
If necessary, I can post screen output of 'make check' in question.
Please send the error messages that you are seeing, together with more information about the machine and OS that you are using.
Please find 'make check' screen output attached.
Error messages are apparently at the end, but maybe some other info is necessary, so I decided to attach the whole file.
Dear Sergei,
as Sebastian anticipated, we need more information about the machine and OS you are using. We also need to know the versions of GCC, GMP and GLPK. The problem shown by the log file you attached concerns ppl_lpsol, an utility using the MIP_Problem class via the C interface of the PPL (the same interface used by GCC). In contrast, the same log file shows that testing via the C++ interface in tests/MIP_Problem/ succeeds. Looking forward to receive more information,
Roberto-- Prof. Roberto Bagnara Applied Formal Methods Laboratory Department of Mathematics, University of Parma, Italy http://www.cs.unipr.it/~bagnara/ mailto:bagnara@cs.unipr.it
This is the info:
" uname -a Linux amdam2 2.6.27.48-0.2-default #1 SMP 2010-07-29 20:06:52 +0200 i686 athlon i386 GNU/Linux "
- this is 32 bit SUSE-11.1 in simple English.
Regarding versions - I am building as much as possible from sources with a tool of mine, so the versions can be seen in the attached autogenerated 'config_wrapper.sh' script used to run 'confugure'.
I have also attached 'configure' screen output - please see it in 'configure.log' file.
Thanks, Sergei.
On 08/26/10 23:15, Sergei Steshenko wrote:
Another piece of info I forgot to mention.
In the end I need to build gcc-4.4.4. I am building it in two stages in addition to whatever stages used by 'gcc' build mechanism itself.
First I build gcc-4.4.4 without 'cloog', then using this clogg-less gcc-4.4.4 I build 'cloog' and its dependencies, and then having built 'cloog' I build the final gcc-4.4.4 with 'cloog' - still using the cloog-less gcc-4.4.4.
So now you know the gory details of the 'gcc' I'm using to build the stuff :).
Again, IIRC, ppl-0.11 builds just fine using this scheme, but ppl-0.10.2 produces 'make check' errors we are discussing in this thread.
Dear Sergei,
thanks for the additional information. I have completed the investigation and I can say that your observation (which I have been able to reproduce) is a problem in one of the demo programs of PPL 0.10.2 that in no way can affect GCC.
Said that, I warmly recommend to use PPL 0.11, as PPL 0.10.2 will not receive any further maintenance. We have already identified a few minor issues in PPL 0.11 and we plan to release PPL 0.11.1 soon. I urge Graphiters to help us test PPL 0.11 on as many platforms as possible in order to make sure PPL 0.11.1 will include all the necessary patches. All the best,
Roberto
Regarding "I warmly recommend to use PPL 0.11" - well, I can't, please see
http://groups.google.com/group/cloog-development/browse_thread/thread/db7eee... .
Thanks, Sergei.
On 8/27/10, Roberto Bagnara bagnara@cs.unipr.it wrote:
On 08/26/10 23:15, Sergei Steshenko wrote:
Another piece of info I forgot to mention.
In the end I need to build gcc-4.4.4. I am building it in two stages in addition to whatever stages used by 'gcc' build mechanism itself.
First I build gcc-4.4.4 without 'cloog', then using this clogg-less gcc-4.4.4 I build 'cloog' and its dependencies, and then having built 'cloog' I build the final gcc-4.4.4 with 'cloog' - still using the cloog-less gcc-4.4.4.
So now you know the gory details of the 'gcc' I'm using to build the stuff :).
Again, IIRC, ppl-0.11 builds just fine using this scheme, but ppl-0.10.2 produces 'make check' errors we are discussing in this thread.
Dear Sergei,
thanks for the additional information. I have completed the investigation and I can say that your observation (which I have been able to reproduce) is a problem in one of the demo programs of PPL 0.10.2 that in no way can affect GCC.
Said that, I warmly recommend to use PPL 0.11, as PPL 0.10.2 will not receive any further maintenance. We have already identified a few minor issues in PPL 0.11 and we plan to release PPL 0.11.1 soon. I urge Graphiters to help us test PPL 0.11 on as many platforms as possible in order to make sure PPL 0.11.1 will include all the necessary patches. All the best,
Roberto-- Prof. Roberto Bagnara Applied Formal Methods Laboratory Department of Mathematics, University of Parma, Italy http://www.cs.unipr.it/~bagnara/ mailto:bagnara@cs.unipr.it
On 08/27/10 19:09, Sergei Steshenko wrote:
Regarding "I warmly recommend to use PPL 0.11" - well, I can't, please see
http://groups.google.com/group/cloog-development/browse_thread/thread/db7eee... .
Hi Sergei,
the answer to your query is "it's only a matter of changing 'configure'" I am using GCC 4.5.1 and CLooG-PPL 0.15 will PPL 0.11 since three weeks without any problem. All the best,
Roberto
On Fri, Aug 27, 2010 at 12:09, Sergei Steshenko sergstesh@gmail.com wrote:
Regarding "I warmly recommend to use PPL 0.11" - well, I can't, please see
http://groups.google.com/group/cloog-development/browse_thread/thread/db7eee... .
So, is it possible to make 'cloog-ppl' compatible with ppl-11.0 ? Or interface of the latter changed, so it's not only a matter of changing 'configure' ?
I will not make cloog-ppl-0.15.x compile with ppl-0.11: cloog-ppl-0.15 is a branch of cloog.org that will not evolve anymore. The only changes that are going to be integrated in cloog-ppl-0.15 will be bug fixes. Please use cloog.org trunk if you want to configure and build with ppl-0.11. Also please note that gcc-4.6 will not work with cloog-ppl-0.15, and so you will have to use cloog.org. There will be an official release of cloog.org, with a version number 0.16.x, before gcc-4.6 is released.
Sebastian
On Fri, Aug 27, 2010 at 12:49:27PM -0500, Sebastian Pop wrote:
I will not make cloog-ppl-0.15.x compile with ppl-0.11: cloog-ppl-0.15 is a branch of cloog.org that will not evolve anymore. The only changes that are going to be integrated in cloog-ppl-0.15 will be bug fixes.
I'd be as happy as the next guy to see cloog-ppl die, but allowing it to be compiled with ppl-0.11 sounds like a bug fix to me.
Please use cloog.org trunk if you want to configure and build with ppl-0.11.
Does that mean that Andreas's patches made into gcc? I thought he had only tested the isl backend. In any case, you would need cloog-parma if you want to use PPL.
skimo
On Fri, Aug 27, 2010 at 13:05, Sven Verdoolaege skimo-cloog@kotnet.org wrote:
On Fri, Aug 27, 2010 at 12:49:27PM -0500, Sebastian Pop wrote:
I will not make cloog-ppl-0.15.x compile with ppl-0.11: cloog-ppl-0.15 is a branch of cloog.org that will not evolve anymore. The only changes that are going to be integrated in cloog-ppl-0.15 will be bug fixes.
I'd be as happy as the next guy to see cloog-ppl die, but allowing it to be compiled with ppl-0.11 sounds like a bug fix to me.
Ok, then let's release cloog-ppl-0.15.10 that will work with both ppl-0.10 and ppl-0.11.
Please use cloog.org trunk if you want to configure and build with ppl-0.11.
Does that mean that Andreas's patches made into gcc?
Yes, the major part of Andreas patches are in GCC trunk, the change to GCC's configure scripts is still missing, as we need some functions or macros to determine the version of cloog.org. With that it should be straight forward to include his configure changes to GCC trunk.
Sebastian
On Fri, Aug 27, 2010 at 13:17, Sebastian Pop sebpop@gmail.com wrote:
Ok, then let's release cloog-ppl-0.15.10 that will work with both ppl-0.10 and ppl-0.11.
Fixed like this: http://repo.or.cz/w/cloog-ppl.git/commit/8cd75413eb37f925de96a1ae77acf43eb31... I am uploading cloog-ppl-0.15.10.tar.gz to gcc/infrastructure/
Sebastian
On Fri, Aug 27, 2010 at 14:49, Sebastian Pop sebpop@gmail.com wrote:
On Fri, Aug 27, 2010 at 13:17, Sebastian Pop sebpop@gmail.com wrote:
Ok, then let's release cloog-ppl-0.15.10 that will work with both ppl-0.10 and ppl-0.11.
Fixed like this: http://repo.or.cz/w/cloog-ppl.git/commit/8cd75413eb37f925de96a1ae77acf43eb31... I am uploading cloog-ppl-0.15.10.tar.gz to gcc/infrastructure/
ftp://gcc.gnu.org/pub/gcc/infrastructure/cloog-ppl-0.15.10.tar.gz
Thanks.
My tool at the moment is building that cloog-less gcc4, so it will take a while until it reaches the new 'cloog' - first ppl-0.11 has to be built, and its 'make check' takes quite a while (time and disk space :) ).
I'll keep you posted.
Regards, Sergei.
On 8/28/10, Sebastian Pop sebpop@gmail.com wrote:
On Fri, Aug 27, 2010 at 14:49, Sebastian Pop sebpop@gmail.com wrote:
On Fri, Aug 27, 2010 at 13:17, Sebastian Pop sebpop@gmail.com wrote:
Ok, then let's release cloog-ppl-0.15.10 that will work with both ppl-0.10 and ppl-0.11.
Fixed like this: http://repo.or.cz/w/cloog-ppl.git/commit/8cd75413eb37f925de96a1ae77acf43eb31... I am uploading cloog-ppl-0.15.10.tar.gz to gcc/infrastructure/
ftp://gcc.gnu.org/pub/gcc/infrastructure/cloog-ppl-0.15.10.tar.gz
'make' of cloog-ppl-0.15.10 fails in the very beginning with the following message: " make: *** No rule to make target `m4/libtool.m4', needed by `Makefile.in'. Stop. ".
Meanwhile I'll revert to cloog-ppl-0.15.9 and ppl-0.10.2.
Regards, Sergei.
On 8/28/10, Sergei Steshenko sergstesh@gmail.com wrote:
Thanks.
My tool at the moment is building that cloog-less gcc4, so it will take a while until it reaches the new 'cloog' - first ppl-0.11 has to be built, and its 'make check' takes quite a while (time and disk space :) ).
I'll keep you posted.
Regards, Sergei.
On 8/28/10, Sebastian Pop sebpop@gmail.com wrote:
On Fri, Aug 27, 2010 at 14:49, Sebastian Pop sebpop@gmail.com wrote:
On Fri, Aug 27, 2010 at 13:17, Sebastian Pop sebpop@gmail.com wrote:
Ok, then let's release cloog-ppl-0.15.10 that will work with both ppl-0.10 and ppl-0.11.
Fixed like this: http://repo.or.cz/w/cloog-ppl.git/commit/8cd75413eb37f925de96a1ae77acf43eb31... I am uploading cloog-ppl-0.15.10.tar.gz to gcc/infrastructure/
ftp://gcc.gnu.org/pub/gcc/infrastructure/cloog-ppl-0.15.10.tar.gz
participants (4)
-
Roberto Bagnara -
Sebastian Pop -
Sergei Steshenko -
Sven Verdoolaege