From: Vladimir Prus (ghost_at_[hidden])
Date: 2003-05-22 01:56:58
Andre Hentz wrote:
> On Mon, 19 May 2003 Ali Azarbayejani wrote:
> > The complaint is that CYGWIN is not a value of feature "os" or
> > "msvc" is not a value of feature "toolset. Of course, these won't be
> > values when I am on Linux. Let me know if you can see a logical
> > problem...otherwise I'll do a use case later.
> I want to second that. I have some statements in my Jamfile that
> look like this:
> lib xerces : : <name>xerces-c_2 <toolset>msvc ;
> lib xerces : : <name>xerces-c <toolset>gcc ;
> Unfortunately, that requires 'using msvc' to be issued, which cannot be
> done on FreeBSD (msvc.jam fails to initialize since the compiler is not
I want to third this ;-) I'm using the same physical drive for testing V2 on
Linux and NT (drive is Samba-mounted to NT) and quite often I configure msvc
from NT, and get errors from Linux.
However, I don't think the solution is to drop validation of values. Say you
lib xerces : : <name>xerces-c_2 <toolset>mvsc ;
Without validation, you might spend a lot of time before noting that "mvsc"
should be "msvc".... Would it be better if msvc can be dummy-initialized on
FreeBSD --- so that "msvc" is declared as value of <toolset>, but no warning
about missing cl is issued?
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