From: Jurko GospodnetiÄ (jurko.gospodnetic_at_[hidden])
Date: 2008-09-01 16:19:30
>> Hmmm... but isn't this target actually the wrong one to use here?
>> Isn't the intention here to just install the headers belonging to this
>> one library and not all libraries every time? Not sure how that
>> 'ideal' would be implemented though as a library can depend on many
>> headers from other libraries as well.
> I think that this would require that each library specify
> a) which headers belong to it.
> b) what other libraries it depends on.
> A library can specify its headers and the dependencies can be
> found automatically.
> The former is probably safer, since header scanning is imperfect.
Bah... I do not see any of that happening until something external
starts forcing a feature like that. And until then I guess the current
solution is as good as it gets.
Header scanning really can be imperfect while forcing library authors
to list their dependencies would need a lot of pushing power. And I do
not think library authors could be made to 'like' the extra work, at
least without some great enough benefit they can recognize.
Possibly some kind of header scanning combined with a way to
explicitly specify any missed headers but all that seems like too much
effort for not enough gain.
I'll tackle the other part of the post in a separate reply.
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