From: Vladimir Prus (ghost_at_[hidden])
Date: 2004-05-28 05:53:31
> > And how do we know
> > 1. What toolset to load?
> > 2. What version to configure?
> > I'm thinking we better treat <toolset> in a special way and allow
> > undeclared values of that feature.
> I'm +1 for this, too.
> It is nearly impossible to support "standard" configurations for some
> 20+ toolset with several subvariants.
> For example.
> I have 3 different gcc configurations already and have to disable some
> warning for gcc-3.4.0 (-Wno-deprecated) when evaluating this compiler.
> My co-workers still use gcc-3.3.2 and do not even have 3.4.0 installed.
> The same issue arises when testing other compilers ... (and commiting
> the Jamfiles, of course...)
The patch I've just comitted allows to specify any value of <toolset>,
including subfeatures, inside condition. On the other hand, other feautures
in condition are now validated. For example
now generates an error, since correct spelling is <link>shared. A similiar
error has occured recently in one of the Boost libraries.
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