From: Vladimir Prus (ghost_at_[hidden])
Date: 2004-01-30 09:59:24
Vladimir Prus wrote:
> I don't think that will help. The issue is that if we decide that
> <foo>xxx <foo>yyy
> <foo>yyy <foo>xxx
> are different for all 'foo', then we have to create much more property-set
> instances that we do now. If we decide they are difference only of 'foo' is
> 'order-sensitive' feature, we have to create more property sets, but not as
> much. The question is if there's real performance effect.
Okay, first measurements. With CVS state "bjam -n" takes 5 seconds on some
project. If I don't sort properties, it takes 11 seconds and crashes later
with "Duplicate name of actual target" error.
Looks like this is not good alternative. I'll look into making specific
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