Re: [PATCH] meson: Avoid implicit declaration of functions

[Paolo Bonzini <pbonzini@redhat.com>]

On 5/29/23 12:24, Peter Maydell wrote:
> On Mon, 29 May 2023 at 08:58, Michal Privoznik <mprivozn@redhat.com> wrote:
>>
>> While detecting a presence of a function via 'cc.links()'
>> gives desired result (i.e. detects whether function is present),
>> it also produces a warning on systems where the function is not
>> present, e.g.:
>>
>>    qemu.git/build/meson-private/tmph74x3p38/testfile.c:2:34: \
>>    warning: implicit declaration of function 'malloc_trim' [-Wimplicit-function-declaration]
> 
> Produces a warning where ? On stdout/stderr ?
> 
> The linked bug report says "in the configure logs"
> which is kinda vague. Warnings in the logfiles are
> not a problem; warnings in the terminal output are...

It seems to be a lint:

  * QA Notice: Found the following implicit function declarations in configure logs:
  *   /var/tmp/portage/app-emulation/qemu-7.2.0-r3/work/qemu-7.2.0/tools-build/meson-logs/meson-log.txt:562 - malloc_trim
  *   /var/tmp/portage/app-emulation/qemu-7.2.0-r3/work/qemu-7.2.0/tools-build/meson-logs/meson-log.txt:591 - statx
  *   /var/tmp/portage/app-emulation/qemu-7.2.0-r3/work/qemu-7.2.0/tools-build/meson-logs/meson-log.txt:627 - statx
  *   /var/tmp/portage/app-emulation/qemu-7.2.0-r3/work/qemu-7.2.0/softmmu-build/meson-logs/meson-log.txt:718 - malloc_trim
  *   /var/tmp/portage/app-emulation/qemu-7.2.0-r3/work/qemu-7.2.0/softmmu-build/meson-logs/meson-log.txt:747 - statx
  *   /var/tmp/portage/app-emulation/qemu-7.2.0-r3/work/qemu-7.2.0/softmmu-build/meson-logs/meson-log.txt:783 - statx
  * Check that no features were accidentally disabled.
  * See https://wiki.gentoo.org/wiki/Modern_C_porting.

So, not an actual bug but more of a favor that we can do to the distros.

>> +    cc.has_function('malloc_trim', prefix: '#include <malloc.h>') and \
>>       cc.links('''#include <malloc.h>
>>                   int main(void) { malloc_trim(0); return 0; }''')
> 
> This seems super clunky -- surely this isn't the way Meson
> intends people to write tests ?

For this one can remove the cc.links test altogether, so this part of the
patch is okay.

The statx is less clear.

> 
> +statx_test = statx_prefix + '''
>    int main(void) {
>      struct statx statxbuf;
>      statx(0, "", 0, STATX_BASIC_STATS, &statxbuf);
>      return 0;
>    }'''
>  
> -has_statx = cc.links(statx_test)
> +has_statx = cc.has_function('statx', prefix: statx_prefix) and \
> +    cc.links(statx_test) 

Here we could replace cc.links with a cc.has_header_symbol('sys/stat.h',
'STATX_BASIC_STATS'), and likewise for has_statx_mnt_id below.

If it was just has_statx I would rather leave things as is.  However,
given that there is has_statx_mnt_id too, I'd apply a v2 that removes
usage of cc.links for these three tests.

Paolo