httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Greg Stein <gst...@lyra.org>
Subject Re: FINAL_* flags ?
Date Tue, 09 Jan 2001 22:46:14 GMT
On Tue, Jan 09, 2001 at 10:12:44PM +0000, Ben Laurie wrote:
> Greg Stein wrote:
> > 
> > On Tue, Jan 09, 2001 at 12:48:07PM -0500, Jim Jagielski wrote:
> > > Jim Jagielski wrote:
> > > >
> > > > Here's the concept: We create a family of env vars, called
> > > > FINAL_* (eg: FINAL_CFLAGS) that are passed to all configures.
> > > > Looks like we could then add this to the COMPILE line in
> > > > rules.mk if I'm reading that right.
> > >
> > > with a bit of minor addition to fastgen.sh to store it in
> > > the Makefiles
> > 
> > Hrm... we're getting a real explosion of flags, rather than dropping back
> > and looking at the problem from scratch. This will mean we have CFLAGS,
> > EXTRA_CFLAGS, and FINAL_CFLAGS. Then throw in OPTIM for good measure, and I
> > bet there are one or two others that I'm not thinking of.
> > 
> > IMO, we should just have:
> > 
> > input: CFLAGS
> > output: (autoconf-discovered-)CFLAGS USER_CFLAGS
> > 
> > We can pass USER_CFLAGS to the sub-configures as CFLAGS. Like any normal
> > autoconf script, they read CFLAGS on input and propagate that into the
> > output CFLAGS.
> > 
> > Note that (I believe) it is normal behavior for an autoconf script to read
> > CFLAGS on input. That's at least the expectation (and evidenced by Ben's
> > attempt to do "CFLAGS=-Werror ./configure"
> 
> Then the subconfigures will fail in eactly the same way!

No, no... I didn't explain completely, I guess. You and Jim both read it
this way.

"First" lines in configure.in:

  USER_CFLAGS="$CFLAGS"
  CFLAGS=""

Then at the very end, we can either munge them together, or subst both of
them and have the compile line use both values (I'd prefer the latter).

That should cure the problem.

Cheers,
-g

-- 
Greg Stein, http://www.lyra.org/

Mime
View raw message