Re: Documenting AC_LANG_CASE

[Akim Demaille]

>>>>> "Alexandre" == Alexandre Oliva <address@hidden> writes:

Alexandre> On Nov 29, 2000, Akim Demaille <address@hidden> wrote:
>> Now there are: AC_LANG_SOURCE etc.  That's the only reason why they
>> needed AC_LANG_CASE.

Alexandre> I see.  With AC_LANG_SOURCE, it's easy enough to construct
Alexandre> AC_LANG_CASE out of nested ifs.  

:) :) :)

Alexandre> But then, why not provide it in the first place?

My point is that AC_LANG_CASE is extremely low level, and I'm
convinced Autoconf should not expose too low levels, but provide the
medium layer which people need.  AC_LANG_PROGRAM, and family are this
medium layer, the new generation of AC_CHECK_FUNCS etc. are the high
level interface etc.

More seriously, what did you mean with ``With AC_LANG_SOURCE, it's
easy enough to construct AC_LANG_CASE out of nested ifs''?  It seems
to me that the motivation for introducing AC_LANG_CASE here was not
right.

Again, I'm not strictly against exposing AC_LANG_CASE, but I'd like to
be proved it's needed, and it's really needed (i.e., some code needs
it to achieve its point).  Otherwise it's exposing too much Autoconf,
which means even more burden in the future.