Status of devel/boost

[Max Brazhnikov]

On Fri, 20 Feb 2009 17:15:53 +0300, Alexander Churanov wrote:
> 2009/2/20 Li-Wen Hsu <lwhsu at freebsd.org>
>
> >  That is great and I made another patch based on your own, which fixed
> > some minor problems
> > (includes building on amd64):
> >
> > http://people.freebsd.org/~lwhsu/boost-1.37.diff
>
> I've examined your patch. It really adds useful things besides indentation
>
> :-) . I've also noted that my proposed patch is incomplete. Actually I've
>
> fixed Math and Test libraries, but since 'diff' was used to compare
> directories, new files were not included into the patch.
>
> > Pav had an exp-run and here is the list of the ports which failed:
> >
> >
> > http://people.freebsd.org/~lwhsu/boost-1.37.txt
>
> This also useful. How much time is takes for you, guys, to rebuild all
> ports that depend on boost? I've set up dedicated i386 and amd64 build
> boxes, but it still takes more than a day (don't know actual time for now)
> to rebuild all depending ports.
> I'm pretty happy with my current setup, however if somebody with faster
> hardware would take job of rebuilding all depending ports and telling me
> which failed - then that would make the process much faster. Since I would
> only work on fixing failed ports and ensuring they build successfully.
>
> > All OpenOffice.org ports are failed, openoffice.org-2.4.2_1.log is
> > just one of them.
>
> Bad news. OpenOffice is huge. However, I think it's possible to examine and
> fix this.
>
> > So are you working on a new patch currently?  Maybe you can based on
> > my 1.37 patch (appply that first then update it to 1.38), this may help
> > thing goes
> > more smoothly.
>
> Yes, sure. I've missed a flaw fixed by you in python/detail/config.hpp
>
> > That's really a lot of work, maybe we also need to notify maintainers
> > and see if upstream has newer version which supports newer version boost.
> >  For the currently
> > status, if we want have 1.37 in the tree, the fastest way is repocopy
> > devel/boost to devel/boost134
> > and changes all the LIB_DEPENDS of the ports which needs 1.34 to build.
> >  Maybe we just set
> > a time to make this happen?
>
> You are reading my thoughts. However, this would present a choice: either
> you have recent boost or you have applications that depend on 1.34 and
> boost-1.34, since they are unlikely to coexist. Having multiple versions of
> the same port installed on the system at the same time is ideal, but I do
> not beleive it's possible at present.
Multiple version of boost can coexist. But boost port should be fixed at first 
to install versioned includes/libs (which is supported by boost afaik) and 
this has to be done if multiple boost ports will be included to the portstree.
btw, it would be nice also to resolve conflict between boost and boost-python 
ports. Is there reason to have default boost without python support?

> My plan is to examine how much work is necessary to fix all ports, this
> looks like a better way for now.
Fixing ports to build vs versioned but supported boost is less hassled, that 
patching them to build vs latest boost.

> Alexander Churanov

Max