Re: [PULL 15/26] configure, meson: move pthread_setname_np checks to Meson

[Peter Maydell <peter.maydell@linaro.org>]

On Mon, 15 Nov 2021 at 17:16, Daniel P. Berrangé <berrange@redhat.com> wrote:
>
> On Thu, Oct 14, 2021 at 06:29:27PM +0200, Paolo Bonzini wrote:
> > This makes the pthreads check dead in configure, so remove it
> > as well.
>
> This change appears broken
>
> On v6.1.0
>
> $ grep SETNAME_NP build/config-host.h
> #define CONFIG_PTHREAD_SETNAME_NP_W_TID 1
>
> While on git master
>
> $ grep SETNAME_NP build/config-host.h
> #define CONFIG_PTHREAD_SETNAME_NP_WO_TID
> #define CONFIG_PTHREAD_SETNAME_NP_W_TID
>
>
> it shoudn't be possible to have both these configs
> satisfied as they're 2 completely different impls.
>
> In the meson-log.txt we can see both tests passing, but with
> warnings
>
>   /home/berrange/src/virt/qemu/build/meson-private/tmpt191k3q1/testfile.c:9:5: warning: implicit declaration of function 'pthread_setname_np'; did you mean 'pthread_setcanceltype'? [-Wimplicit-function-declaration]
>
> So it isn't actually validating the function parameter
> signature, and then link is succeeding because the linker
> doesn't care about the incompatible function signatures.
>
> Peter meanwhile reports that he gets no pthread_Setname_np
> at all, and his meson-build.log shows -Werror set. This
> causes the missing fnuction signature to cause tests to
> fail in both cases.

More specifically, these tests get run with
"-Werror=implicit-function-declaration" (and this is not
because I've passed any kind of -Werror option in --extra-cflags;
I have passed --enable-werror but that is supposed to be the
default anyway...)

-- PMM