Re: pkginclude and pkglib?

[Gerard Jungman <jungman at lanl dot gov>]

Dave Morrison wrote:
> 
> For what it's worth, I'd be happy to help implement this change.  How do
> you see it working during the build of gsl itself?  You would need a
> "gsl" directory somewhere in the picture so that

Yes, annoying isn't it. I have done that sort of thing
in the past. We can create a gsl/ directory at the top,
stick a Makefile.am in it with a rule to create
the links, and put "gsl" first on the SUBDIRS list in
the top Makefile.am.


> One possible downside to this
> method is that it doesn't automatically guarantee that all headers
> mentioned in the various 'include_HEADERS' will have symlinks in
> $top_builddir/gsl.

I think we just have to live with this. We have been very
scrupulous about making sure all exported headers have the "gsl_"
prefix. Although that looked a little silly when we went over
to using pkginclude, it happens to be very useful now; they
are easy to find in the source tree (w/o having to grep
the Makefile.am's themselves... BLECHHHH).


> I'm sure there are more sophisticated ways of
> getting this done.  Do you have any thoughts on the subject?

Unfortunately, I don't know any better way. I guess the general
problem is that automake does not (as far as I know) support the notion
of striped builds, where the various subdirs are passed over
in different phases or stripes. If it did, then it would
be possible to formalize this sort of thing in a "code generation"
pass, which would occur before any compilation pass, and
could automatically "do the right thing" when you use
pkginclude_HEADERS (is Tom Tromey listening...?).

This general failing of automake has been an annoyance in the past.
Oh well.

Anyway, I would like to wait for Brian to cast his
vote before doing anything on this. He might have
an idea for something better.


-- 
G. Jungman