|
Boost-Build : |
From: Vladimir Prus (ghost_at_[hidden])
Date: 2006-11-13 05:04:52
On Monday 13 November 2006 12:54, Johan Nilsson wrote:
> Roland Schwarz wrote:
> > Johan Nilsson wrote:
> >> Isn't this more of a compiler-specific issue than O/S specific?
> >
> > Yes, of course, I was a little sloppy.
> >
> >> I'd prefer only platform B generating an error. If I explicitly
> >> request a shared runtime library I'd like to either get it, or fail.
> >
> > Perhaps what irritates me is "error" and "fail", which might mean:
> > stop
> > everything, which in case of building a larger set of independent
> > targets is too limiting.
>
> I think I agree with you here, unless "bjam -q" is explicitly requested. In
> that case I'd like everything to stop.
>
> > What I would prefer instead is to put <build>no into the build
> > in this case. E.g when building the install target of boost this
> > would not stop everything but simply skip building this single
> > variant.
>
> I'm not so sure. In the general case I'd prefer error messages being output
> and a summary error status returned from bjam. Still not saying everything
> should stop though.
>
> I realize that when by default building a large set of variants, such as
> when invoking "bjam install" from boost root, the <build>no feature would
> be very nice to use.
>
> OTOH, what should happen if someone requested e.g.
> bjam --toolset=msvc --threading=single --runtime-link=shared from boost
> root? For this case it should definitely not silently generate nothing,
> IMHO.
Please check out the discussion Rene an I had in the "impossible features"
thread. In the case of all errors, you'll still have the error message, and
an error summary right at the end of output.
- Volodya
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