2017-10-06 9:43 GMT+02:00 Paul Menzel paulepanter@users.sourceforge.net:
Having the code base compatible with future toolchains is quite important and convenient in my opinion.
That's a great argument to switch out the toolchain 4 months before a release. Which means 3.5 months ago.
Especially for bisecting an issue, people don’t rebuild the toolchain in each step.
Which means that no matter what we do to crossgcc, there's no reproducibility. That's not a good argument for doing something or anything at all.
The good side is, that most warnings and errors of newer toolchains actually point out real problems – most of them don’t affect anything though –, so fixing those would be a good thing.
Fixing real issues? Yes, please. Swapping out the entire code optimizer phase, potentially introducing regressions through no code change at all in coreboot itself? Not at this time.
Patrick