Re: [PATCH v2 00/10] Rework READ_ONCE() to improve codegen

[Arvind Sankar]

On Thu, Jan 23, 2020 at 10:45:08AM -0800, Nick Desaulniers wrote:
> On Thu, Jan 23, 2020 at 9:32 AM David Laight <David.Laight@xxxxxxxxxx> wrote:
> >
> > From: Will Deacon
> > > Sent: 23 January 2020 17:17
> > >
> > > I think it depends how much we care about those older compilers. My series
> > > first moves it to "Good luck mate, you're on your own" and then follows up
> 
> I wish the actual warning was worded that way. :P
> 
> > > with a "Let me take that off you it's sharp".
> 
> > Oh - and I need to find a newer compiler :-(
> 
> What distro are you using? Does it have a package for a newer
> compiler?  I'm honestly curious about what policies if any the kernel
> has for supporting developer's toolchains from their distributions.
> (ie. Arnd usually has pretty good stats what distro's use which
> version of GCC and are still supported; Do we strive to not break
> them? Is asking kernel devs to compile their own toolchain too much to
> ask?  Is it still if they're using really old distro's/toolchains that
> we don't want to support?  Do we survey kernel devs about what they're
> using?).  Apologies if this is already documented somewhere, but if
> not I'd eventually like to brainstorm and write it down somewhere in
> the tree.  Documentation/process/changes.rst doesn't really answer the
> above questions, I think.
> 
> -- 
> Thanks,
> ~Nick Desaulniers

Reposting Arnd's link
https://www.spinics.net/lists/linux-kbuild/msg23648.html

Seems like there is nothing still on versions between 4.6 and 4.8. It
might be useful to put a list of distro's that are running the current
minimum supported toolchain versions in process/changes.rst?

I think the issue is not just kernel devs. Users may need to compile a
custom kernel or at least build a module.