This is the mail archive of the
libstdc++@sourceware.cygnus.com
mailing list for the libstdc++ project.
Re: WCHAR_MIN; WCHAR_MAX
- To: Benjamin Kosnik <bkoz at cygnus dot com>
- Subject: Re: WCHAR_MIN; WCHAR_MAX
- From: Ulrich Drepper <drepper at cygnus dot com>
- Date: 07 Oct 1999 13:17:01 -0700
- Cc: Tom Tromey <tromey at cygnus dot com>, libstdc++ at sourceware dot cygnus dot com
- References: <Pine.LNX.4.04.9910071300280.24179-100000@decepticon.cygnus.com>
- Reply-To: drepper at cygnus dot com (Ulrich Drepper)
Benjamin Kosnik <bkoz@cygnus.com> writes:
> I can use AC_DEFINE and have things defined in config.h, and then have a
> header (say bits/std_cmath.h) that include config.h and has a mess of
> #if/#elif whatever stuff in it that determines what really happens.
Well, you said you needed it for mknumeric_limits. If you need the
information in the installed header then this is not acceptable, of
course, and we should shoot for the fixincludes solution.
> What I'm talking about is having autoconf generate a config.h, and
> the after all the tests have been run, and all the necessary macros
> defined, having autoconf use a header template (bits/std_cmath.h.in) that
> is used to generate a (bits/std_cmath.h) that only has the correct info in
> it, no extra defines. You can do something like this with Makefiles, using
> automake and AM_CONDITIONAL (see the libstdc++/libio/Makefile.* for an
> example).
You can do this with configure as well. You can use AC_OUTPUT to generate
arbitrary files based on all the sed information stored in config.status.
> I'm talking about this functionality for header and source files,
> and run at configure time not make time. I guess you are saying that
> this is what fixincludes is for. . . oh, joy.
Finxincludes is still the right answer, yes.
--
---------------. drepper at gnu.org ,-. 1325 Chesapeake Terrace
Ulrich Drepper \ ,-------------------' \ Sunnyvale, CA 94089 USA
Cygnus Solutions `--' drepper at cygnus.com `------------------------