configure.in is unable to find ppl over 0.10 Reproducible: Always Steps to Reproduce: 1. 2.emerge -a1v cloog-ppl 3. apply patch cloog-ppl-0.15.10-ppl_not_found.patch
Created attachment 332852 [details, diff] patch for ppl versions over 0.10
I don't know if cloog-ppl will even work with ppl-1.0. We have to do some testing first.
i think we're going to drop cloog-ppl entirely now that gcc-4.8+ supports the main cloog package
*** Bug 529202 has been marked as a duplicate of this bug. ***
With GCC 4.9 now ~arch, has the time come to mask it for removal?
(In reply to Tony Vroon from comment #5) > With GCC 4.9 now ~arch, has the time come to mask it for removal? The problem with this is cross-toolchains older than 4.9 (like gcc-4.7.4, which I am stuck on for MIPS due to PR61538, which upstream is ignoring) will dep on cloog-ppl if USE graphite. The proposed patch appears to work. At least, it compiles. There is this warning, however: * This package has a configure.in file which has long been deprecated. Please * update it to use configure.ac instead as newer versions of autotools will die * when it finds this file. See https://bugs.gentoo.org/426262 for details. Which the patch touches, so that will have to be dealt with for the long term.
(In reply to Tony Vroon from comment #5) i think the critical part is that gcc-4.8 is stable pretty much everywhere. so we can unwind graphite support from <=gcc-4.7.
(In reply to Joshua Kinard from comment #6) > (In reply to Tony Vroon from comment #5) > > With GCC 4.9 now ~arch, has the time come to mask it for removal? > > The problem with this is cross-toolchains older than 4.9 (like gcc-4.7.4, > which I am stuck on for MIPS due to PR61538, which upstream is ignoring) > will dep on cloog-ppl if USE graphite. > Do you need graphite? Removing it from <=4.7 also cleans up the toolchain.eclass a bit.
(In reply to Anthony Basile from comment #8) > (In reply to Joshua Kinard from comment #6) > > (In reply to Tony Vroon from comment #5) > > > With GCC 4.9 now ~arch, has the time come to mask it for removal? > > > > The problem with this is cross-toolchains older than 4.9 (like gcc-4.7.4, > > which I am stuck on for MIPS due to PR61538, which upstream is ignoring) > > will dep on cloog-ppl if USE graphite. > > > > Do you need graphite? Removing it from <=4.7 also cleans up the > toolchain.eclass a bit. Doubtful. Never really tried it, and not really inclined. Probably better to save that and LTO testing for 4.9, when I can get that to run on these SGI platforms.
We would remove USE=graphite from <=4.7 like we did for 4.5 previously.
With the revelation that PR61538 was unknowingly addressed upstream, this means I'm not stuck on 4.7 with these MIPS systems anymore. I'll be kickstarting a stage build using 4.9.2 shortly just to make sure there aren't any more problems. Whatever you guys want to do to 4.7 now, go ahead!
Created attachment 396912 [details, diff] Patch against toolchain.eclass This patch drops pre 4.8 support for graphite and nicely cleans up the whole cloog-ppl mess.
Comment on attachment 396912 [details, diff] Patch against toolchain.eclass >+ # graphite was added in 4.4 but we only support it in 4.8+ due to external >+ # library issues. >+ confgcc+=( $(use_with graphite cloog) ) >+ use graphite && confgcc+=( --disable-isl-version-check ) I believe this will call `use_with` on versions where graphite isn't in IUSE which is undefined behaviour wrt. PMS. It also makes an automagic dependency. It needs to be explicitly disabled for 4.4-4.7. We might as well support 5.0 while we're at it. 5.0 drops the dependency on cloog and just requires isl-0.14 (actually 0.12.2 but 0.14 is recommended). Something like (untested): if tc_version_is_at_least 5.0 ; then confgcc+=( $(use_with graphite isl) ) use graphite && confgcc+=( --disable-isl-version-check ) elif tc_version_is_at_least 4.8 ; then confgcc+=( $(use_with graphite cloog) ) confgcc+=( $(use_with graphite isl) ) use graphite && confgcc+=( --disable-isl-version-check ) elif tc_version_is_at_least 4.4 ; then confgcc+=( --without-cloog ) confgcc+=( --without-ppl ) fi (Plus the required RDEPEND bits.) Bleh. At least the include paths are gone.
*** Bug 541898 has been marked as a duplicate of this bug. ***
i've pushed this now that merges all of our versions: http://sources.gentoo.org/eclass/toolchain.eclass?r1=1.655&r2=1.656
Commit message: Block cloog-ppl as we are punting it http://sources.gentoo.org/dev-libs/ppl/ppl-1.1.ebuild?r1=1.3&r2=1.4
Commit message: Block cloog-ppl as we are punting it http://sources.gentoo.org/dev-libs/cloog/cloog-0.18.0.ebuild?r1=1.13&r2=1.14 http://sources.gentoo.org/dev-libs/cloog/cloog-0.18.1.ebuild?r1=1.13&r2=1.14 http://sources.gentoo.org/dev-libs/cloog/cloog-0.18.3.ebuild?r1=1.2&r2=1.3 http://sources.gentoo.org/dev-libs/cloog/cloog-9999.ebuild?r1=1.4&r2=1.5
cloog-ppl itself is now dropped as well: http://gitweb.gentoo.org/repo/gentoo.git/commit/?id=b813ba983b1784bf746fdc62284067a5f9394b36
