The problem is that:
/tmp % ace --trace AC_PROG_CC
autoconf: warning: --macrodir is obsolete, use --autoconf-dir
configure.in:2:AC_PROG_CC:foo bar supercompiler
configure.in:2:AC_PROG_CC:
someone is requiring AC_PROG_CC *inside* the AC_PROG_CC itself. But
since AC_PROG_CC (the first one) is in elaboration, AC_REQUIRE or
something emits it anyway, and requirements are expanded *before*.
Therefore, AC_PROG_CC is expected twice. And it is always, this is
independent from the presence of a list of compilers.
AC_DEFUN([AC_PROG_CC],
[AC_BEFORE([$0], [AC_PROG_CPP])dnl
# Coucou
...
/tmp % cat configure.in
AC_INIT
AC_PROG_CC
/tmp % ace
/tmp % grep Coucou configure
# Coucou
# Coucou
The chain is:
AC_DEFUN([AC_PROG_CC],
[AC_BEFORE([$0], [AC_PROG_CPP])dnl
AC_LANG_PUSH(C)
....
AC_EXPAND_ONCE([_AC_EXEEXT])
AC_LANG_POP
])# AC_PROG_CC
and
define([_AC_EXEEXT],
[_AC_CYGWIN
...
and
define([_AC_CYGWIN],
[AC_CACHE_CHECK(for Cygwin environment, ac_cv_cygwin,
[AC_COMPILE_IFELSE([AC_LANG_PROGRAM([],
and
AC_DEFUN([AC_COMPILE_IFELSE],
[AC_LANG_COMPILER_REQUIRE()dnl
So we face, again, the problem that AC_PROG_CC depends upon
AC_COMPILE_IFELSE which depends upon AC_PROG_CC. Likewise for all the
languages, of course (we have to write a test for this one).
So I am back to something I raised some time ago: why the heck do we
have to compile to recognize Mingw etc.? Can't we just uname? While
anti-Autoconf (and btw, frankly, compiling __MINGW32__ makes no big
difference to me), it seems better in that, for instance, it makes it
possible to enable the support of these kind of machines without
requiring a C compiler.
