From: Vladimir Prus (ghost_at_[hidden])
Date: 2004-11-20 08:04:57
On Saturday 20 November 2004 15:48, Reece Dunn wrote:
> >>E.g. bjam cw-8.3 target-os=mac # os=NT
> > Why would you want to have host os as property? After all, you can't
> > build with two host os at the moment. In other words
> > bjam cw-8.3 host-os=NT host-os=LINUX
> > makes no sense, so Boost.Build can just detect host os itself and use
> > this information when needed. And that means "os" can mean "target os".
> What I mean is that bjam defines os to be the OS that the user is
> running on (i.e. the host os). This is set internally, so you don't set
> it on the command line (thst is not what I meant).
This need to be tweaked: you current os should be default value of the "os"
feature, but still allow user to specify a different value. What I'm saying
that we need only one feature, there's no need for both host-os and
> >>One of the motivations for using <gui> as a feature, or a better name
> >>might be <os-api>, is to allow support for different APIs, e.g.:
> >> bjam msvc-8.0 os-api=win32
> >> bjam msvc-8.0 os-api=win64
> >> bjam gcc-3.3 target-os=linux os-api=x11
> >> bjam gcc-3.3 target-os=linux os-api=motif
> > Maybe you're right, if you want to target Qt and gtk, then <os> would be
> > too coarse grained.
> We can have the following rules:
>  set <os> to the operating system being run (already implemented);
>  if <target-os> is not defined, define it to be equal to <os> (note:
> you may want to define <target-os> as NT when <os>=CYGWIN).
>  validate that <target-os> has an appropriate value (already
> provided via feature).
>  if <os-api> is not defined, define it to an appropriate default (e.g.:
>  validate that the <os-api> value is appropriate for the value of
> <target-os> (e.g. <target-os>LINUX:<os-api>cocoa is invalid).
Interesting, it means some combination of features should be declared invalid
and it should be checked by Boost.Build. OTOH, this is purely optional
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