[Intel-gfx] [PATCH 2/2] drm/i915: Ensure intel_engine_init_execlist() builds with Clang

Nick Desaulniers ndesaulniers at google.com
Mon Oct 22 21:10:25 UTC 2018


On Wed, Oct 17, 2018 at 12:25 AM Jani Nikula <jani.nikula at intel.com> wrote:
>
> On Tue, 16 Oct 2018, Nick Desaulniers <ndesaulniers at google.com> wrote:
> > On Tue, Oct 16, 2018 at 8:35 AM Stephen Boyd <swboyd at chromium.org> wrote:
> >>
> >> Quoting Jani Nikula (2018-10-16 05:29:38)
> >> > Clang build with UBSAN enabled leads to the following build error:
> >
> > I'm overjoyed that you're testing this configuration! If you find more
> > bugs, we're happy to help triage if you let us know about them here:
> > https://github.com/ClangBuiltLinux/linux/issues.
>
> Everyone, thanks for the patches/reports, reviews, and testing! These
> two have now been pushed to drm-intel-next-queued; due to timing they're
> expected to land upstream in the merge window for v4.21 i.e. not the
> imminent one. (So much for the quick turnaround! ;)
>
> Also glad to connect you all here. :)
>
> So I don't mind making small concessions to Clang build such as these in
> the interest of gaining better static analysis. Our (i915) own pre-merge
> build testing uses primarily gcc and sparse, with a bunch of extra flags
> (see the top of drivers/gpu/drm/i915/Makefile), but I'm not sure about
> adding Clang to the mix anytime soon. Without a tight and constant
> feedback loop new issues will pop up though. Maybe we'll spot inline
> function use in constant expressions in review, maybe we don't. I'm sure
> there are other Clang specific issues.
>
> Do you have any plans for setting up something like the 0day? Without
> the feedback loop I think you'll take one step back for every two steps
> you progress...

Yes, I was in contact with Intel's 0day folks (out of Shanghai, I
believe).  Unfortunately, the patch requiring asm-goto landed just as
they started looking into supporting clang which really really
derailed things.  See
http://lkml.iu.edu/hypermail/linux/kernel/1804.0/00720.html where no
one working on Clang was CC'ed about this.

In the meantime, we've been focusing on arm64 and kernel-ci with
Linaro.  More on that soon.

>
> BR,
> Jani.
>
> --
> Jani Nikula, Intel Open Source Graphics Center



-- 
Thanks,
~Nick Desaulniers


More information about the Intel-gfx mailing list