This is the mail archive of the
libstdc++@sourceware.cygnus.com
mailing list for the libstdc++ project.
Re: making conforming headers.
- To: Steven King <sxking at uswest dot net>
- Subject: Re: making conforming headers.
- From: Benjamin Kosnik <bkoz at cygnus dot com>
- Date: Thu, 4 May 2000 17:30:57 -0700 (PDT)
- cc: "Martin v. Loewis" <martin at loewis dot home dot cs dot tu-berlin dot de>, libstdc++ at sourceware dot cygnus dot com
> Its the cname/name.h problem that I was looking at. The cname headers
> are, as most of them note, non-conforming and the name.h headers dont
> exist.
with any luck nathan will revisit this stuff soon. As far as I'm
concerned, the "mkcshadow" code needs to be dusted off and made
working, and added to the configure scripts by default. If you're looking
for something fun to do, this might be a good place to start....
The perennial problem that these approaches have (based on my experiences
with std_cmath.h) is that autoconf does a good job of being able to test
of things, and communicate clearly via AC_OUTPUT what functions and
headers and structs are around, but some kind of intermediate step is then
necessary to correctly prune and sculpt the correct headers given this
information. This seems to be the place where most of the v3 stuff falls
down: for std_cmath.h we gave up, and just littered the code with
macros... anyway.
Supposedly gtk-- has this same problem. It would be cool if somebody could
come up with an elegant solution (even if it is just a sed/perl/bash
script.)
Having this kind of functionality will simplify the build process in a
bunch of ways (the biggest one: we wont' need c++config.h to be a
superset of autoconf's config.h...)
(getting off soapbox)
benjamin