Re: [Boost-build] [boost] [bbv2][autolink] Re: Naming of32/64bitlibs

From: Rene Rivera (grafikrobot_at_[hidden])
Date: 2006-10-22 14:09:43

Vladimir Prus wrote:
> On Sunday 22 October 2006 19:43, Rene Rivera wrote:

>> * V1 does not have any special logic for figuring out if a toolset has
>> or doesn't a debug runtime. And hence sets the "g" unconditionally when
>> <runtime-build>debug is present.
> 
> Hmm, my experiments don't confirm this. I go to libs/thread/build and run:
> 
>    bjam -sBUILD="debug <runtime-build>debug"   
> 
> the libraries created are:
> 
>      libboost_thread-gcc-mt-d-1_35.a
>      libboost_thread-gcc-mt-d-1_35.so
>      libboost_thread-gcc-mt-d-1_35.so.1.35.0
> 
> And there's no 'g' in the library name. So it looks like some code figured out 
> that <runtime-build>debug does not affect gcc?

After investigating for an hour what I can figure out is that a feature 
is not included in the build when that feature is not handled by the 
toolset. That is if the toolset doesn't specify a "flags <toolset> X 
<runtime-build>debug : -x ;" the runtime-build feature will not show up 
to the name tagging rule, or any place else AFAICT. Hence for many V1 
toolsets the "g" doesn't show.

>> * The Boost V1 Jamfile builds all the variants resulting from
>> specifying: "debug release [debug-python] <runtime-link>static/dynamic
>> <threading>single/multi" on Windows (regardless of toolset). And
>> otherwise from: "debug release [debug-python] <threading>single/multi".
> 
> So, on Windows only, we also add variation on <runtime-link> property. Right?

Yes. But I don't think that's the correct thing to do, i.e. it was a kludge.

-- 
-- Grafik - Don't Assume Anything
-- Redshift Software, Inc. - http://redshift-software.com
-- rrivera/acm.org - grafik/redshift-software.com
-- 102708583/icq - grafikrobot/aim - grafikrobot/yahoo