From: Rene Rivera (grafik.list_at_[hidden])
Date: 2003-12-06 10:38:47
John Maddock wrote:
> I've just run the install routine on Linux again, and I still think there
> are a couple of things that could be improved:
> Header file location:
> I appreciate why you place these in a "versioned" directory, but the result
> is that the headers are not in the users include path, further it's very
> hard to see how a configure script could discover their location
> auto-magically. Likewise anyone building their project from a makefile
> would have to edit all their makefiles etc every time the boost version
> number changes.
This is the same issue Victor raised. And the conclusion is that yes it's the
goal to have either symlinks, or equivalent, from an unversioned name to the
versioned names. The only thing preventing this from occurring for the 1.31
release is the time to write the BBv1 support for doing that. I'll look at it
today to see if I can find an easy/fast way of doing it.
> 1) I don't see how an autoconf script can discover the right library version
There's a whole discussion about Open Source Development and the lack of
Configuration Management in it that I could insert here, but a reference is
..Perhaps write a small loop that iterates testing for
"/usr/local/include/boost-$(V)" to get you a list of all versions installed.
Then pick the latest??
> 2) I don't see how an autoconf script can discover the right compiler
I don't know much about how autoconf does things so I can't answer that :-)
But at minimum I would think it could easily construct the toolset part, after
all it detects the compiler you use. As for the rest.. if you are writing
makefiles it's not that difficult to write you Makefile.in to have some logic
for constructing the suffix in the same way we do in a top level / global
location. And then you have "-l$(BOOST_REGEX) -l$(BOOST_THREAD)", etc.
> Again a symlink would help, certainly for (1), not sure yet about (2).
> Is there a rule for symlinks in boost.build already? I've found one for
> hardlinks, but not surprisingly for symlinks?
No rule for symlinks, at least not in BBv1... There is one in BBv2 ;-)
-- -- Grafik - Don't Assume Anything -- Redshift Software, Inc. - http://redshift-software.com -- rrivera_at_[hidden] - grafik_at_[hidden] - 102708583_at_icq
Boost-Build list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk