Re: [PATCH] tests/tcg: Suppress compiler false-positive warning on sha1.c

["Alex Bennée" <alex.bennee@linaro.org>]

Peter Maydell <peter.maydell@linaro.org> writes:

> On Tue, 4 Mar 2025 at 14:29, Alex Bennée <alex.bennee@linaro.org> wrote:
>>
>> Peter Maydell <peter.maydell@linaro.org> writes:
>>
>> > On Tue, 4 Mar 2025 at 13:44, Alex Bennée <alex.bennee@linaro.org> wrote:
>> >>
>> >> Peter Maydell <peter.maydell@linaro.org> writes:
>> >> > I think it's new-ish (gcc 11?). On the other hand
>> >> > -Wno-unknown-warning-option is quite old, and would suppress
>> >> > this error. If we do
>> >> >  CFLAGS+=-Wno-unknown-warning-option -Wno-stringop-overread
>> >> >
>> >> > does that work?
>> >>
>> >> Yes, I did:
>> >>
>> >> modified   tests/tcg/hexagon/Makefile.target
>> >> @@ -18,7 +18,7 @@
>> >>  # Hexagon doesn't support gdb, so skip the EXTRA_RUNS
>> >>  EXTRA_RUNS =
>> >>
>> >> -CFLAGS += -Wno-incompatible-pointer-types -Wno-undefined-internal
>> >> +CFLAGS += -Wno-incompatible-pointer-types -Wno-undefined-internal -Wno-unknown-warning-option
>> >>  CFLAGS += -fno-unroll-loops -fno-stack-protector
>> >
>> > I think we should do this where we add -Wno-stringop-overread,
>> > not just for the hexagon tests -- or are the tcg tests
>> > guaranteed to be run with a fixed compiler from a container
>> > regardless of the local dev environment?
>>
>> I can move it, but hexagon is unusual in being clang based. However the
>> oldest compilers we use are 10.2 in the qemu/debian-legacy-test-cross
>> container.
>
> My question was more "do we only ever build this test with
> a fixed set of compilers that we control, or are we instead
> maybe sometimes using the user's clang/cc/gcc" ?

Its certainly possible - at least on a Debian system or with the
appropriate --cross-cc-$ARCH flags. I expect 99% of devs either have
Debian or Docker/Podman to fall back on.

> If the latter, we definitely need to associate the
> "don't warn about unknown warnings" with the place we are
> adding the option that's not known across all our supported
> compilers. If the former, then putting it in the hexagon
> specific file seems OK I guess.
>
> -- PMM

-- 
Alex Bennée
Virtualisation Tech Lead @ Linaro