From: Felipe Magno de Almeida (felipe.m.almeida_at_[hidden])
Date: 2007-04-25 23:49:29
On 4/25/07, David Abrahams <dave_at_[hidden]> wrote:
> on Wed Apr 25 2007, Vladimir Prus <ghost-AT-cs.msu.su> wrote:
> > However, I still at loss what real examples call for this behaviour. Would
> > you mind providing some?
> I already did; the python examples call for this behavior. To
> abstract it a bit more:
I support David for the heuristic behavior and I'll try to explain why.
IMO, when somebody uses bjam feature1=value1 feature2=value2 ...
featureN=valueN. What the user really want is to build everything, but
forcing those features for their values. But, it doesnt have anything
to do, directly, with alternative selection. Alternatives are created
by the project's mantainer and usually is an implementation detail of
the library. And as such, shouldn't popup to the user's face as
ambiguous error. The user shouldn't even know, in most cases, that
there is an alternative selection.
Though, I'm not yet convinced about how the heuristic should really
work. Maybe counting how many sequenced matches would be better? (I
dont know if it was clear).
> Dave Abrahams
> Boost Consulting
-- Felipe Magno de Almeida Consultor - Expertise Solutions http://www.expertisesolutions.com.br
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