From: Vladimir Prus (ghost_at_[hidden])
Date: 2008-04-23 03:19:24
On Wednesday 23 April 2008 10:27:19 Roland Schwarz wrote:
> Vladimir Prus wrote:
> > On Wednesday 23 April 2008 02:27:29 Roland Schwarz wrote:
> > So:
> > 1. We codify that the important properties line cygwin/mingw should
> > be fixed for a given version configured in user-config.jam, and that
> > adding -mno-cygwin by hand is not OK.
> Not sure what you mean by "properties line", I guess "flavor"? If yes, I
> mean that every
> gcc need to get one.
You've misunderstand what I say above. For your solution to work, we need
to require that 'flavour' or whatever is *fixed* for a given version
configured in user-config.jam. So, if you have "gcc-blahblah' it's always
gcc version 22.214.171.124 targeting cygwin on Z80 processor, whatever OS,
and it's not OK to change any of that using command line flags. Then, of
course, the question is how to document this adequately. See, gcc manual
does not mention any 'flavour' and it's not possible to say if any given
option will change that 'flavour' or not.
> > 2. -mno-cygwin actually changes the output of -dumpmachine?
> Yes. Output without -mno-cygwin is
> and with:
> >> My ideas are like so:
> >> 1) Make threadapi an optional feature
> >> 2) Change the <conditional> requirement to add a threadapi
> >> only if none specified yet and make it conditioned on
> >> flavor of compiler.
> > This seems like a good approach, except that we don't have this
> > 'conditional feature as default' thing you appear to suggest above.
> > Or I musunderstood, and it's actually doable now?
> I tried a little yesterday, and it almost works. The only thing I need
> is the change for flavor.
> Altough the code does not look obvious, and perhaps can be done better.
> I post to the
> list as soon I have something to show.
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