From: Mark Elston (m.elston_at_[hidden])
Date: 2004-06-16 13:20:34
Ah, I begin to see. I haven't seen the references to <variant>debug
or <toolset>gcc in the docs (are they there?).
With that mechanism in place I see why things are the way they are.
There is nothing really *magic* happening. There are just some
shortcuts (saving keystrokes). If we were to write scripts to use
bjam and wanted to "fully" document the scripts we could modify
the command-line I gave below like this:
bjam <variant>debug <variant>release <toolset>msvc <toolset>gcc MyTarget
This is very explicit and, knowing what is going on here, I can
freely use the shortened form without any lingering questions.
Except, perhaps, for compiling a "complete" listing of all available
features and any associated "shortcuts".
I'll have to go back and reread the section on features now.
BTW, is it possible to combine similar features? For example, in
the above command line could we shorten it to something like:
bjam <variant>"debug release" <toolset>"msvc gcc" MyTarget
I know that was available in v1 for the toolset since it was specified
with a -sTOOLS="..." option where multiple tool types could be specified
inside the quotes.
Thanks again for the explanation.
> -----Original Message-----
> From: Vladimir Prus [mailto:ghost_at_[hidden]]On Behalf Of Vladimir
> Sent: Tuesday, June 15, 2004 10:57 PM
> To: jamboost_at_[hidden]
> Subject: Re: [jamboost] Re: Some fundamental misunderstandings (v2)
> Mark Elston wrote:
> > I *had* read this but didn't get the significance. I'm not sure
> > I still make the connection.
> > Are you saying that the choice of compiler is now just another
> > feature? And these features can be combined with all other
> > features? So now I can do this:
> > bjam debug gcc MyTarget msvc release
> > to build debug and release versions of MyTarget using both gcc
> > and msvc compilers (not that I would do it this way...). Is this
> > correct?
> Yes, exactly. Above, "debug" is shortcut for <variant>debug and "gcc" is
> shortcut for <toolset>gcc, where both <variant> and <toolset> are
> > If so, I think I prefer the old way since it was actually somewhat
> > clearer about what was happening.
> That's the case where opinions differ, I'm afraid. It's hard to change
> behaviour now, since V2 exists for quite some time and V2 users are
> accustomed to new behaviour.
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