Re: Please add new Android branches

On Mon, Sep 25, 2023 at 1:16 AM Guillaume Tucker
<[email protected]> wrote:
>
> On 20/09/2023 18:57, Nick Desaulniers wrote:
> > On Tue, Sep 19, 2023 at 5:12 PM Todd Kjos <[email protected]> wrote:
> >>
> >> On Tue, Sep 19, 2023 at 4:19 PM Nick Desaulniers
> >> <[email protected]> wrote:
> >>>
> >>> Half of the red exclamation marks on that link are are "0 warnings - 0
> >>> errors".  What's up with that?
> >>
> >> This more recent build has "0 warnings - 0 errors":
> >> https://linux.kernelci.org/build/android/branch/android15-6.1/kernel/ASB-2023-07-05_14-6.1-2620-gf2d0380464d3/
> >>
> >> Here is the error (not sure why it bubbled up as 0/0):
> >>
> >>   AR      drivers/base/power/built-in.a
> >> ../fs/namei.c:2172:13: error: use of bitwise '|' with boolean operands
> >> [-Werror,-Wbitwise-instead-of-logical]
> >>         } while (!(has_zero(a, &adata, &constants) | has_zero(b,
> >> &bdata, &constants)));
> >>
> >> ~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> >>                                                    ||
> >> ../fs/namei.c:2172:13: note: cast one or both operands to int to
> >> silence this warning
> >>   CC      drivers/base/firmware_loader/builtin/main.o
> >> 1 error generated.
> >> make[3]: *** [../scripts/Makefile.build:250: fs/namei.o] Error 1
> >> make[3]: *** Waiting for unfinished jobs....
> >> make[2]: *** [../scripts/Makefile.build:500: fs] Error 2
> >
> > Guillaume,
> > Any idea what's up with that? Seems like the existence of diagnostics
> > is detected correctly; the number of warnings and errors is not.
>
> Yes I think there's some heuristics in the legacy log parsing
> tool to count the number of compiler warnings and errors.  This
> probably is outdated with Clang now, so the build exit status is
> used to detect a failure but then the stats are off.

Right, I think clang-17 changed the structure of its output to use
pipes `|` similar to GCC.

I've been keeping my eye on something called SARIF, which is some kind
of structured interchange format for diagnostics produced by compilers
and consumed by IDEs. (Pretty sure it's just JSON).  That might be
interesting to pursue in KernelCI so that the diagnostics don't need
hand rolled parsers (that can break).

>
> It's better to avoid fixing this code as it's going to be retired
> in a few months' time to be replaced with a fresh implementation
> based on the new API.  I believe some tweaks could be done to fix

Oh? Is there a ticket I can sub to to track that?

> critical issues, but if the detail of warnings and errors can be
> found even though the stats are off then maybe we can save the
> effort of fixing it?
>
> Cheers,
> Guillaume
>


-- 
Thanks,
~Nick Desaulniers