[patch][graphite] Check cloog revision when configure.

[Richard Guenther]

On Fri, 6 Nov 2009, Sebastian Pop wrote:

> On Fri, Nov 6, 2009 at 05:33, Tobias Grosser <grosser@fim.uni-passau.de> wrote:
> > On Fri, 2009-11-06 at 18:15 +0800, Li Feng wrote:
> >> Hi,
> >> On Fri, Nov 6, 2009 at 11:46 AM, Tobias Grosser
> >> <grosser@fim.uni-passau.de> wrote:
> >> > On Fri, 2009-11-06 at 11:05 +0800, Li Feng wrote:
> >> >> Hi,
> >> >>
> >> >> In polyhedral benchmark, aermod is miscompiled because of the
> >> >> older cloog revision number.
> >> >>
> >> >> Here is the bug and discussion:
> >> >> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41924
> >> >> http://groups.google.com/group/gcc-graphite/browse_thread/thread/ee4485e5fd1e4302
> >> >>
> >> >> To avoid this happen again from both user and develper,
> >> >> I would like to check cloog revision when configuring.
> >> >> Sebastian has proposed the same thing some time ago:
> >> >> http://gcc.gnu.org/ml/gcc-patches/2009-07/msg01839.html
> >> >>
> >> >> Here is the patch:
> >> >>
> >> >> diff --git a/configure.ac b/configure.ac
> >> >> index 48fa580..5fab23a 100644
> >> >> --- a/configure.ac
> >> >> +++ b/configure.ac
> >> >> @@ -1593,7 +1593,7 @@ if test "${ENABLE_CLOOG_CHECK}" = "yes"; then
> >> >> Â  Â CFLAGS="$CFLAGS $clooginc $gmpinc $pplinc"
> >> >> Â  Â AC_MSG_CHECKING([for correct version of CLooG])
> >> >> Â  Â AC_TRY_COMPILE([#include "cloog/cloog.h"],[
> >> >> - Â #if CLOOG_VERSION_MAJOR != 0 || CLOOG_VERSION_MINOR != 15
> >> >> + Â #if CLOOG_VERSION_MAJOR != 0 || CLOOG_VERSION_MINOR != 15 ||
> >> >> CLOOG_VERSION_REVISION < 4
> >> >
> >> > I personally believe moving "||" to the next line and identing the new
> >> > row would look better.
> >>
> >> Because of the #if directives, we should put them in one line. Otherwise,
> >> it will not checking this.
> >>
> >> >
> >> >> Â  Â choke me
> >> >> Â  Â #endif
> >> >> Â  Â ], [AC_MSG_RESULT([yes])], [AC_MSG_RESULT([no]); clooglibs= ; clooginc= ])
> >> >>
> >> >> Ok for trunk?
> >> >
> >> > You still need to add a ChangeLog entry and to update the "configure"
> >> > file using autotools.
> >> >
> >> Ok.
> >>
> >> > Otherwise I believe the patch is useful and required since a while, I
> >> > just did not do it. So thanks for your affords.
> >> >
> >> > If there are no concerns until Sunday it should be fine to commit,
> >> > however I would like to see the updated patch, before saying yes.
> >> >
> >> > Tobi
> >> >
> >> >
> >> here is the patch:
> >
> > It is fine with me. Please commit if bootstrapped (or configured) and
> > there are no concerns until Sunday.
> >
> 
> From the original discussion at
> http://gcc.gnu.org/ml/gcc-patches/2009-07/msg01839.html
> Richi was concerned by such a patch, so I think that we should get
> an approval from Richi and from a configure maintainer before
> committing this patch to trunk.

Well, as GCC doesn't bootstrap with the older versions the check is
ok with me.  But you still need a configure maintainer to approve
the patch.

Richard.