Re: [jamboost] Common Lib Subdir

From: Rene Rivera (grafik666_at_[hidden])
Date: 2003-08-15 18:49:45

[2003-08-15] Administrator wrote:

>Hi all,
>
>I was wondering what was the status of utilizing a common $(BOOST_ROOT)/lib
>subdir to place compiled libraries into was, and what, if anything, needs
to
>be done in order to help out.

Last we left the relevant conversation at...

[2003-08-05] Beman Dawes wrote:
>At 02:01 AM 8/5/2003, Rene Rivera wrote:
> >>* We need to test this build process on a daily basis. I don't know
> >>if that means adding to the current regression tests or adding a
> >>separate test framework.
> >
> >IMO adding a separate framework is easier than trying to integrate to the
> >current one. We would not want to break the regression testing, given 
>that
> >it's working nicely. And testing the build is considerably easier.
> >
> >Initially just making a build script, ala tools/regression/run_tests.sh,
> >that automates the fetch/build/process. And like the regression parse the
> >log for errors. etc.
> >
> >Is process_jam_log flexible enough to parse the output of a regular 
>build?
> >I could write a quick Perl script, otherwise.
>
>I'd rather wait a bit until the requirements become clearer. The point for 
>now is to just start thinking about the idea that not only must this 
>build/install procedure work, but we also need to test regularly to make 
>sure it stays working.

...So I'd say the best way you can help is by taking a look at what the
proposed build+install procedure does and giving feedback. In addition to
the current CVS, which you already have, check out the boost-root/Jamfile on
the build_for_distribution branch...

cvs co -r build_for_distribution boost/Jamfile

It tries to mimic configure type options, which you can get help on by
doing...

cd boost-root
bjam --help

One immediate feedback question I have for you is wether it's more important
to build+install, as autoconf/configure/make does, or just build, and let
the user and/or package builders install?

>For my own sake, I pulled down the latest CVS sources, and patched the
>relevant Jamfiles in order to have this functionality. If anyone is
>interested, I can post the patchfile to this list.

I looked at the patches you posted on the Boost.Users list, but feel free to
post them here for those others that may be interested.

The one comment I have about your patches is that the approach you took is
unlikely to win favor with the Boost library authors. This is because they
would have to maintain the stage rules themselves. Most authors tend to want
less things to worry about, rather than more ;-)

That said.. having something like an "install" rule for library authors to
specify what they want installed could be a workable and prefered
alternative.

-- grafik - Don't Assume Anything
-- rrivera (at) acm.org - grafik (at) redshift-software.com
-- 102708583 (at) icq