Re: ac-lang-compiler-gnu.patch

[Pavel Roskin]

Hello, Akim!

> The aim is to reach the day where all the languages share the same
> skeleton.

Sounds good.

> Index: ChangeLog
> from  Akim Demaille  <address@hidden>
>       * aclang.m4 (_AC_LANG_COMPILER_GNU): New macro, which unifies...
>       (_AC_PROG_F77_GNU, _AC_PROG_CC_GNU, _AC_PROG_CXX_GNU): Remove.
>       Adjust dependencies.

Ok provided that
1) The testsuite passes (my favourite condition :-))
2) You read my comments and make adjustments when needed.

> -[if test "ac_gnu_compiler" = yes; then
> +[if test "$ac_compiler_gnu" = yes; then

Missing "$" must be a typo. If it is please apply it separately.

> -ac_gnu_compiler=$ac_cv_prog_gcc
> +ac_compiler_gnu=$ac_[]_AC_LANG_ABBREV[]_cv_compiler_gnu

Could you please avoid such renamings?
I can imagine many configure.in's using it e.g. to add -Wall to CFLAGS.

> +[_AC_COMPILE_IFELSE([AC_LANG_SOURCE([[#ifndef __GNUC__
> +  choke me
> +#endif

How about that semicolon - shouldn't we still worry about NeXT:

> -[# The semicolon is to pacify NeXT's syntax-checking cpp.
> -cat >conftest.$ac_ext <<_ACEOF
> -#ifdef __GNUC__
> -  yes;
> -#endif

> -if test $ac_cv_prog_gcc = yes; then
> +if test $ac_c_cv_compiler_gnu = yes; then

Please, leave those names alone! You can just define them like
_AC_LANG_ABBREV for every language.

I'm sure a lot of code depends on the traditional names.

> -#ifdef __GNUC__
> -  yes;
> -#endif

Note this semicolon again for C++ - there must be a reason.
Hopefully it's irrelevant for Fortran.

Regards,
Pavel Roskin