Re: to conditionally test, or not to conditionally test?

[Ralf Corsepius]

On Fri, 2005-05-27 at 09:44 +0200, Stepan Kasal wrote:
> Hi,
> 
> On Fri, May 27, 2005 at 08:17:58AM +0200, Ralf Corsepius wrote:
> > On Thu, 2005-05-26 at 09:50 +0200, Stepan Kasal wrote:
> > > 
> > > The advice ``perform all tests unconditionally'' is a workaround
> > > to fix current limitations of the implementation.
> ...
> > 
> > It actually should be "perform all tests for compilers unconditionally",
> > i.e. try using AC_PROG_<language> unconditionally, because autoconf
> > can't handle "using AC_PROG_<language> conditionally".
> 
> I wouldn't say so.
>
> The problem is that if macro BAR contains AC_REQUIRE([FOO]), then FOO
> is expanded inside the `if' and may escape execution.

True, ...
> It's hard to tell whether a macro calls AC_REQUIRE.  (It can call it
> indirectly.)
True, ... but ...

... in practice in most cases AC_PROG_<language> will be close to the
beginning of a configure script before "feature checks". So unless you
are doing something nasty, this in most cases works around the
AC_REQUIRE issue. If this wasn't the case, autoconf would completely
unusable.

However, autoconf apparently is usable, because the AC_REQUIRE issue
rarely shows effect ...

> A real fix will be to use shell functions to reimplement AC_REUIRE,
> in autoconf-3.
... otherwise autoconf-3 would already be existing ...

Ralf