Re: autoconf 2.61: AC_DEFINE variable with parenthesis

[Steven G. Johnson]

Paul Eggert wrote:

But there are at least two regressions here.  First, the one reported
by Andrey Simonenko in
<http://lists.gnu.org/archive/html/bug-autoconf/2006-12/msg00026.html>,
where AC_DEFINE([DEF(x)], [somevalue]) does not work in Autoconf 2.61
as it did in Autoconf 2.59.  Second, the one I introduced in
<http://lists.gnu.org/archive/html/bug-autoconf/2006-12/msg00031.html>,
which explicitly generates a warning when you invoke
AC_DEFINE([DEF(x)], [somevalue]).

Ah, I didn't notice that you had added this warning.







Attached is a patch, with ChangeLog entry below.

Steven

2006-12-28  Steven G. Johnson  <[EMAIL PROTECTED]>

        * lib/autoconf/general.m4 (AC_DEFINE_TRACE): Don't include
        preprocessor macro arguments in traced name.
        * doc/autoconf.texi (Defining symbols): Document longstanding
        support for AC_DEFINE-ing macros with arguments, and document
        behavior when the same variable has multiple AC_DEFINEs.
        * lib/autoconf/fortran.m4 (_AC_FC_WRAPPERS): Revert to the
        old implementation which AC_DEFINEs the FC_FUNC and FC_FUNC_
        macros directly, giving much shorter and simpler code.

Regarding Andrey Simonenko's complaint, it has to do with old-style #define templates in the config.h.in file, which are now deprecated according to the manual. In particular, he was relying on the old (no longer documented?) behavior that if you #define VAR VALUE in the .h.in file then it gets replaced, but it doesn't get replaced if you do just #define VAR [no value]. He wanted to do the same thing for a macro with arguments, i.e. #define VAR(args) [no value] and have it not be replaced. I'm not sure what is the rationale to support this deprecated, no-longer-documented behavior in the context of a "new" feature (macros with arguments). (It's not a regression since it was never documented.) However, the source of this "regression" (if you want to call it that) seems to be revision 1.83 of status.m4 (see your 2006-01-06 entry in the ChangeLog). In particular, prior to that change, the conftest.defines sed pattern skipped to the close-parenthesis (if there was an open-parenthesis after the macro name) before checking to see whether the #define had a value. It would be fairly easy to go back to this behavior if we wanted to, I guess, but I don't really see the point. Regarding the second problem, I have fixed it in the attached patch. In particular, it now always traces just the name of the macro, not including the argument list, matching the behavior of AH_TEMPLATE.

diff.out
Description: Binary data