User:0xdc/substrate/20191201

New release time, it's been a while since the last set back in August! Haven't needed new stages but I did get a new modern laptop so I had to dive back in. Main updates are for the GCC 9.2 stabilisation and the 17.1 profile on amd64.

Notes

amd64

The December stages have been built with the 17.1 profiles with the split multilib.

The only minor note is having to set PORTDIR= in make.conf for the stage1 (I left it for the other builds tho). I hope portage will use /var/db/repos/gentoo as the default going forward.

The bindist builds haven't been looked at much. Something in the plasma stages has some conflicts, and I'm slowly getting away from using plasma for sway so I think those stages might just rot.

armv7a

This caused me a few issues.

stage1 update_seed

Updating the seed stage with `--rebuild-if-new-ver` caused gcc-7.3.0 from the upstream tarball get rebuilt, and for some reason it failed to compile because of a missing header. gcc-7.3.0 isn't needed as we can install gcc-9.2 directly. Well...

stage2 zlib gcc

When building the stage2, zlib was the first package that requires a C compiler, and gcc was throwing errors that compiler warnings were too strict. Turns out that gcc was straight-up failing to find libmpfr.so.

In the seed stage3, gcc was running against mpfr-3, and couldn't update to mpfr-4 due to dependencies on gcc-7.3 and mpc. But in the stage1chroot that conflict doesn't exist, and as such was installing mpfr-4 libs to the stage1, but compiling gcc against the seed library mpfr-3.

I pulled mpfr-3 into the chroot explicitly. I hope that the stage2 and/or stage3 removes it...