On Wed, Oct 16, 2019 at 10:21 PM Masahiro Yamada
<yamada.masahiro@xxxxxxxxxxxxx> wrote:
>
> I tested -std=gnu99 for ARM with pre-built Linaro toolchains.
>
> GCC 4.9.4 was NG,
> GCC 5.3.1 was OK.
Ok, so the gcc-5.1 cut-off from my gcc git tree conversion looks to be
the right one. I wasn't sure how official/complete the git conversion
was.
> If we increase the minimal GCC version, we might end up with dropping
> more architecture.
That I wouldn't worry about. If some architecture can't get a gcc
version from the last five years, I think we _should_ drop it.
Historically, the problem has more been distro gcc versions. An
unmaintained architecture that has a compiler that is ancient I don't
much care about, but if we lose testers that use ancient distros, that
loses real coverage.
That's true even if it's just one or two actual users that upgrade
kernels - we found a real bug not that long ago because rmk used some
ancient Debian install with a new kernel. That's the kind of odd use
we want to encourage, and that matters. Hexagon? Not so much.
Although I think rmk actually had a new compiler and cross-built the
new kernel, so that likely wasn't the issue in _that_ particular case,
but in other cases it might have been.
Linus
