Re: [MLB] Tag inferrence code needs improving

[Nick Hudson]

Robert Boehne wrote:
> 
> Nick Hudson wrote:
> >
> > I've been playing more with the MLB in NetBSD pkgsrc and I'm seeing a
> > few problems with tag inferrences. The problem as I see it boils down to
> > the fact that
> >
> > cc = gcc = /usr/bin/cc = /usr/bin/gcc = ...
> > c++ = g++ = /usr/bin/c++ = /usr/bin/g++ = ...
> >
> > What is the easiest way of getting libtool to infer the correct tag? The
> > introduction of CC_ALIASES or somesuch set at libtool configure/build
> > time seems sensible. Comments?
> 
> I'm not sure why this is a problem, if you don't have a "g++" tag
> configured, but already have "c++", then libtool will just ignore the
> "c++" tag and rebuild an identical one for "g++".  I'll admit this isn't
> beautiful, but I think it is better than a single line of code that
> would require maintenance or could cause a bug at some point.

libtool doesn't build a tag for each "alias" though. Or should I say I
haven't seen it do this. If it did then I agree it wouldn't be a problem
or beautiful.

As a stop gap I'll see if I can invoke this behaviour.

I've found a partial solution to the problems I was seeing. Different
configure scripts would pick a different c++ compiler which would then
mean libtool couldn't infer which tag to use. I've got round it in
making sure that CXX=c++ and so (most) configures will use c++ as the
compiler/linker.

Nick
-- 
aka address@hidden, address@hidden