On Sun, May 16, 2021 at 4:37 PM Bernd Edlinger <bernd.edlin...@hotmail.de> wrote: > > On 5/14/21 10:20 PM, Andrew Pinski via Gcc wrote: > > On Fri, May 14, 2021 at 3:27 AM Richard Biener via Gcc <gcc@gcc.gnu.org> > > wrote: > >> > >> On May 14, 2021 10:53:21 AM GMT+02:00, "Martin Liška" <mli...@suse.cz> > >> wrote: > >>> On 5/12/21 10:51 AM, Richard Biener via Gcc wrote: > >>>> On Tue, May 11, 2021 at 6:34 PM Serge Belyshev via Gcc > >>> <gcc@gcc.gnu.org> wrote: > >>>>> > >>>>> > >>>>>> $ egrep "mpfr\.h" log/cfg/cfg.gcc-11.1.0.log > >>>>>> checking for the correct version of mpfr.h... buggy but acceptable > >>>>>> > >>>>>> It appears "gcc-11.1.0/contrib/download_prerequisites" > >>>>>> specifies "mpfr-3.1.4.tar.bz2" whereas top level 'configure' > >>>>>> has "MPFR_VERSION_NUM(3,1,6)". > >>>>>> > >>>>>> Is there a reason mpfr 3.1.6 isn't being used? > >>>>> > >>>>> No good reason: that script was not updated with new versions. > >>> GCC-11 is > >>>>> also known to work fine with the most recent mpfr version 4.1.0. > >>>> > >>>> Yes, the update of the minimum version and the buggy check was > >>>> done by Janne w/o adjusting the download script. > >>>> > >>>> Richard. > >>>> > >>> > >>> Should I adjust download_prerequisites accordingly? > >> > >> Yes. > > > > It looks like the adjustment was made but the file was not added uploaded: > > [apinski@xeond2 gcc]$ ./contrib/download_prerequisites > > 2021-05-14 13:18:41 > > URL:http://gcc.gnu.org/pub/gcc/infrastructure/gmp-6.1.0.tar.bz2 > > [2383840/2383840] -> "gmp-6.1.0.tar.bz2" [1] > > http://gcc.gnu.org/pub/gcc/infrastructure/mpfr-4.1.0.tar.bz2: > > 2021-05-14 13:18:42 ERROR 404: Not Found. > > error: Cannot download mpfr-4.1.0.tar.bz2 from > > http://gcc.gnu.org/pub/gcc/infrastructure/ > > > > Yes, indeed, and please do not forget to update also the hashes > in contrib/prerequisites.md5 and contrib/prerequisites.sha512 . > > However, I wonder why we change only the default mpfr version to latest > available version but keep gmp and mpc versions from 2015?
I've reverted the change and instead upped to 3.1.6 as I intended to suggest to avoid printing the offending warning on configure time. This can be picked by branches that are similarly affected. Any discussion on upping versions further needs to be accompanied by actual testing on target/host combos where it's usually used (non-Linux). Richard. > > Thanks > Bernd. > > > > > > Thanks, > > Andrew Pinski > > > > > >> > >> Richard. > >> > >>> Martin > >> > >
