From: Vladimir Prus (ghost_at_[hidden])
Date: 2002-10-16 01:48:06
David Abrahams wrote:
>>Should <toolset> feature be symmetric. IOW, if "gcc" is default toolset
>>for me, should targets be built to
>>I think first alternative is preferred --- the shorter paths in the most
>>common case, the better. Other opinions?
> I think I agree. I worry about people rebuilding into the same
> directory by changing their default toolset, but that can be fixed by
> implementing Ullrich Koethe's idea of leaving behind one or more .jam
> files in the target directory which describe the build. If the toolset
> doesn't match up, we'll just force it to rebuild.
That Ullrich's idea seemed good to me before and still seems good now.
We'd need to implement it one day.
>>(BTW, it appears that 'symmetric features' are not yet supported,
>>although it should be extra easy to implementent)
> Thinking a little more, it seems to me that symmetric-ness might be
> determined based on target in some cases. For example, if a library
> requires shared=true, should we generate a shared/ subvariant
You're right. It's something to be remembered for future. (I thought it
might be good to enter this into tracker, but hell, sourceforge does not
even allow to search for all "active" issues, be it "task" or
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