http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57087
--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> 2013-04-27 12:47:38 UTC --- (In reply to comment #1) > > A response from Tim Prince suggested using the exact versions cited in > >the INSTALL HTML files. This referred to: gcc/contrib/download_prerequisites > >However, there is no mention of this in the INSTALL HTML files. That's not a bug, using the script is not required, it's just a convenience. You could have downloaded the correct versions by hand, you chose to use different versions. > >5. Repeat with cited versions > > > > To eliminate any side-effects from previous failed runs, the > > directories were cleared, and files re-extracted from ../gcc-4.7.2.tar.gz. Re-extracting the tarball is not necessary, the benefit of building in a separate obj dir is the src dir is untouched by the build process. > > Redid configure and make. This ran much further than previous attempts. > >So these versions of gmp, mpc, and mpfr seem to be better. That's why the installation docs say to use them. > >6. Summary of Bugs > > > >6.1 Incomplete information provided in the INSTALL HTML files > > > > There is no mention of contrib/download_prerequisites. There seem to be > >many other things in contrib/ which are also not explained. Not a bug. > > These should be included in the INSTALL files, and the point about > >"safer to use the exact versions" could be stressed or qualified with known > >problems. This specially applies to use of "(or later)". > > > > Commands like download_prerequisites need care not to use commands like > >wget, which is not available on my platform (however ftp seems to operate the > >same). It is available, you just have to install it. The download_prerequisites script is not necessary for building GCC, so it's not a bug that you had to download the files another way. > >6.2 An error reported by make > > > > Building failed as in section 4 above using: > > > >gmp-5.1.0 > >mpfr-3.1.1 > >mpc-1.0.1 > > > > Possibly this only concerns libmpfr, but it is not clear whether or > >not the other two have proceeded correctly. > > > > It is also not clear whether this is a problem specific to my platform. Possibly related to PR 50461 although the error is different.
