Boost logo

Boost :

From: Jeff Garland (jeff_at_[hidden])
Date: 2000-12-11 08:47:46


> At 07:49 AM 12/11/2000 -0500, John Maddock wrote:
>
> >... I don't like the idea of having all the cpp files in one directory -

I agree with this also. However all the .cpp files should be under
boost_top/src/libname.

> >there is no guarentee that all of them will build (not all libraries are
> >supported on all platforms). Nor do I think there should be a single
> >"boost" library build, although I use large parts of boost, I have never

This I disagree with. As I described in my earlier post, this is
inconsistent with current boost header policy. If you don't want a build
which can put all of boost in a single library then we should also move the
headers into their own library specific includes.

> >used all of it in one project, users should not pay for what they don't
> >use, and would prefer to see the boost libraries continue to build as
> >separate libraries. For me that means that putting source in
> ><boost>/libs/libname/src is fine, as long as it is properly documented.

Agreed. I would extend this with a configuration option which allows for
easy subsetting of the library to only the desired libs.

Beman Dawes wrote:
> In addition to his points, consider that boost will almost certainly grow
> far beyond the current size and scope. It will be unmanageable
> if we just
> toss all source files into a single source directory.

Absolutely. And if we make everyone build and link every single library one
at a time they will throw up their hands in frustration. I only link one
library to get I/O streams and STL from the standard library. Of course you
might think this is a bad thing :)

> I can't even count the number of times I've considered using a
> component of
> some library, only to reject it because I had to buy into the whole
> unbounded library just to get the one component. Boost has to figure out
> how to avoid that fate. We don't know the whole answer yet, but keeping
> separate library directories seems likely to be part of the solution.

Agreed. See the suggestion in my earlier post.

Jeff


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk