Hi, I confirm that there is a lot of work on upgrade to gcc-5.4.0 (a lot of packages to recompile) but FWIS could optimize a lot of package and there are packages like for example mongodb-3.4 that need a >=gcc-5.x.
Gcc-5.4.0 is now stable on gentoo and could be a good idea plan an upgrade. I test compilation of ebuild sys-devel/base-gcc and sys-devel/gcc for both amd64 and arm and installation is been complete correctly. I think that there is only a fix... avoid set of gcc-5 at the end of base-gcc and move it at the end of the installation of sys-devel/gcc. My cent G. On May 4, 2017 22:15, "Mitch Harder" <mitch.har...@sabayonlinux.org> wrote: > I'm seeing some discussion about updating our Sabayon gcc ebuilds to > version 5.4.0 > > You guys probably already knew this, but just in case this slipped > through the cracks, I want to make sure everyone is fore-warned that > this will be a really big chunk of work once we start using GCC-5.4.0 > on our packages. > > At a minimum, we need to rebuild every package that uses C++ since the > ABI changes. > > We may decide to rebuild everything just to insure consistency. > > Reading some discussions on the Gentoo M/L, mixing GCC 4.9.4 and 5.4.0 > is discouraged by several developers, but there do seem to be a few > who think it can be managed. > > The Gentoo news item for this upgrade recommends using a > revdep-rebuild command to identify the affected C++ packages: > > revdep-rebuild --library 'libstdc++.so.6' -- --exclude gcc > > When I ran this on my systems, I quickly discovered that this method > will not discern packages that have already been rebuilt with > GCC-5.4.0. It just gives me the same list every time. > > So, if we encounter build issues (and we almost certainly will since > some packages will attempt to build out of order), we'll have to come > up with our own system for keeping up with what's been rebuilt, and > what packages remain. > > I apologize if this has already been discussed, but I hadn't run > across any discussion on GCC 5.x, and wanted to make sure. > > Thanks, > Mitch... > >
