On 2024-09-11 01:58, Bruno Haible wrote:
1) Gnulib makes use of C99 __VA_ARGS__ in a couple of places.
I vaguely recall using other C99 features that can't easily be simulated
or finessed by Gnulib. These include designated initializers, compound
literals, and the subset of VLAs that are s
I agree in general, but I still find some value in supporting Debian
Woody for Libidn and will keep it as a local gnulib patch.
/Simon
signature.asc
Description: PGP signature
Hi Simon,
> There is no -Werror in use and the c99 module doesn't help
So it has no way of supporting ISO C 99.
> I think gnulib doesn't really support gcc 2.95
Indeed. gcc 2.95.3 is more than 23 years old [1].
The only sensible thing you can do with this compiler is to build
a newer GCC, such
Bruno Haible writes:
> This is an instance of "declaration after statement", which is a feature of
> C99,
> which Gnulib routinely assumes nowadays. We document this since 2017-04-24
> [1].
>
> I hope you aren't using -Werror, right?
>
> Does the compilation error go away if you add the 'c99' m
Hi Simon,
> I noticed a build failure of libidn on an old architecture that doesn't
> permit non-top variable declarations. How about the patch below? The
> rest builds fine and passes self-tests on the platform, so presumably
> this was a recently introduced regression in malloca.c related to
>
