WCHAR_MIN; WCHAR_MAX
Benjamin Kosnik
bkoz@cygnus.com
Thu Oct 7 12:23:00 GMT 1999
> For now there should be autoconf tests. At the same time someone
> might want to add something to fixincludes to help saving the
> situation. On Solaris 7 machines (at least) the definitions are
> present (in wchar.h) so it iis obvious what fixincludes has to add.
Yeah. You've inadvertently stumbled apon something I really wanted added
as a feature to autoconf, but that Tom and Ben think can be handled with
sed or fixincludes. I have no idea how to do it though, and can't seem to
get the details out of anybody. So, excuse the rant.
I want something like AM_CONDITIONAL but for autoconf, so that I can
generate headers on the fly at configure time, depending on things that
autoconf has already tested for (like for the exampe below,
_GLIBCPP_DONT_USE_WCHAR removes the wchar bits for mknumeric_limits)
I suppose if I could figure out how to do it with sed or fixincludes I
could put it into yet another configure script and run it as part of
AC_OUTPUT_COMMANDS at the end of the configure process.
Does the java project do this kind of massive configuration-time hacking?
> If some platform make wchar_t no integer type it is broken beyond
> belief and we should not even think about supporting it. I don't know
> who ever could be so stupid (though I never saw was M$ is doing).
umm. Try HPUX 10.20 for size someday when you are feeling really patient.
On these platforms I usually just modify bits/c++config.h to not allow
wchar_t:
from:
#define _GLIBCPP_USE_WCHAR_T 1
to:
//#define _GLIBCPP_USE_WCHAR_T 1
> Alternatively require using using the versions of the OSes which have
> the necessary definitions or fixincludes cleans up the mess. I would
> prefer this since it's much cleaner.
Dude, I'll do most of the work, but I just need a clean example that I can
clone.
-Benjamin
More information about the Libstdc++
mailing list