From: David Abrahams (dave_at_[hidden])
Date: 2005-06-14 11:14:52
Vladimir Prus <ghost_at_[hidden]> writes:
> On Saturday 11 June 2005 23:36, David Abrahams wrote:
>> Well, you can, but then any Jamfile that uses that feature, even
>> conditionally (e.g. <toolset>msvc:<asynch-exceptions>on) becomes
>> totally nonportable to any environment where the toolset isn't
>> configured. The validation system barfs when it comes across a
>> feature it (quite naturally) doesn't know about.
> To begin with, why do you want it to be a feature, if it's so
So I can enable it only when it's required by some library I'm using
(i.e. put it in usage-requirements).
> Do you want to build with different values of it?
> If not, using a compiler options can be more reasonable.
Of course; I've been around the block a few times on this one.
> I tend to believe that the validation can be just disabled. It's a
> maintenance nightmare to handle all cases where we should relax
> validation, like conditional properties.
I'm a little afraid of losing checking for things like spelling errors
and people mistakenly using Bbv1 feature names that you've changed in
v2. Maybe we just need to do validation much less often and in only a
few specific places, like the flags rule. That way, no validation
becomes the default and we start thinking about places where we want
to put it in.
-- Dave Abrahams Boost Consulting www.boost-consulting.com
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