On Fri, Nov 13, 2020 at 4:45 PM Segher Boessenkool
<segher(a)kernel.crashing.org> wrote:
On Fri, Nov 13, 2020 at 04:11:41PM -0800, Nick Desaulniers wrote:
> On Fri, Nov 13, 2020 at 3:49 PM Segher Boessenkool
> <segher(a)kernel.crashing.org> wrote:
> > On Fri, Nov 13, 2020 at 10:59:26AM -0800, Nick Desaulniers wrote:
> > > The `optimize` attribute is both non-portable across toolchains (hence
> > > this warning)
> >
> > Like *all* GCC extensions.
> >
> > > and a little quirky in GCC.
> >
> > How so? Don't spread FUD please, say what *is* wrong, then people can
> > decide for themselves whether they want it or not.
>
> Spread FUD? Ard literally sent TO YOU:
>
https://lore.kernel.org/lkml/CAMj1kXHxX+u5-cN0v3SLdqZTSiKsWsFOvc2SC5=-Sca...,
> and it was referenced again in
>
https://lore.kernel.org/lkml/20201028081123.GT2628@hirez.programming.kick....
>
> Was it FUD when Ard sent it to you?
He didn't say "this option is a little quirky". He simply quoted our
wiki entry for it, which says "use this only for debugging" (just like
the user documentation btw). The FAQ also goes on to explain the
attribute is very hard to use, it is not obvious at all what flags you
can and cannot set, it's a user interface disaster. It explains what is
bad with it, it doesn't just say "ooh I don't understand it, do not use
it". (It does say "no one really understands it, do not use it", there
is that ;-) )
Are we splitting hairs here? I want both toolchains to be successful;
though maybe next time I see something like this patch/0day report go
by, I'll just keep my mouth shut and we can deal with the runtime bugs
later?
--
Thanks,
~Nick Desaulniers