Re: [PATCH] configure.ac: add detection of symver gcc attribute

[Andreas Grünbacher]

Hello,

this looks mostly reasonable, but

Am Fr., 16. Feb. 2024 um 22:11 Uhr schrieb Giulio Benetti
<giulio.benetti@benettiengineering.com>:
> From: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
>
> On non-ELF platforms, such as microblaze, builds will fail when trying
> to add symver information because  __attribute__((symver ..)) is not
> supported even though __has_attribute(__symver__) returns true.
>
> Support for symver needs to be detected via a compile test since
> __has_attribute can report false positives [0].
>
> Add a configure compile check for __attribute__((symver ..)) to ensure
> it is supported and define a variable to advertise support.
>
> [0] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101766#c1
>
> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
> Signed-off-by: Giulio Benetti <giulio.benetti@benettiengineering.com>
> ---
>  configure.ac       | 15 +++++++++++++++
>  libattr/syscalls.c |  2 +-
>  2 files changed, 16 insertions(+), 1 deletion(-)
>
> diff --git a/configure.ac b/configure.ac
> index 7e362e9..98477b5 100644
> --- a/configure.ac
> +++ b/configure.ac
> @@ -57,6 +57,21 @@ AS_CASE([$host_os],
>         [linux*], [os_linux=yes])
>  AM_CONDITIONAL([OS_LINUX], [test "x$os_linux" = "xyes"])
>
> +AC_CACHE_CHECK(whether __attribute__((__symver__())) is supported,
> +       gcc_cv_symver_attribute,
> +       [cat > conftest.c <<EOF
> +void foo (void) {}
> +__typeof(foo) foo __attribute__ ((__symver__("foo@foo")));
> +EOF
> +       gcc_cv_symver_attribute=no
> +       if ${CC-cc} -Werror -S conftest.c -o conftest.s >/dev/null 2>&1; then 
> \
> +           gcc_cv_symver_attribute=yes
> +       fi
> +       rm -f conftest.[cs]
> +])
> +AS_IF([test $gcc_cv_symver_attribute = yes],
> +      [AC_DEFINE(HAVE_SYMVER_ATTRIBUTE, [], [GCC supports symver 
> attribute])])
> +
>  AC_CONFIG_COMMANDS([include/attr],
>         [dnl
>         rm -rf include/attr
> diff --git a/libattr/syscalls.c b/libattr/syscalls.c
> index 907560a..caa70bc 100644
> --- a/libattr/syscalls.c
> +++ b/libattr/syscalls.c
> @@ -32,7 +32,7 @@
>   * fall back to traditional .symver asm directive otherwise.
>   */
>  #ifdef __has_attribute
> -# if __has_attribute(__symver__)
> +# if defined(HAVE_SYMVER_ATTRIBUTE)

checking for __has_attribute doesn't make sense when
HAVE_SYMVER_ATTRIBUTE is defined now. Could you please fix the if
nesting?

>  #  define SYMVER(cn, vn) __typeof(cn) cn __attribute__((__symver__(vn)))
>  # elif __has_attribute(__no_reorder__)
>     /*
> --
> 2.34.1
>

Thanks,
Andreas