Re: [Boost-testing] error: "8.3/runtime" is not a known subfeature value of <toolset>cw

From: David Abrahams (dave_at_[hidden])
Date: 2006-12-15 20:07:33

Aleksey Gurtovoy <agurtovoy_at_[hidden]> writes:

>> >> Because instead of having a single file that's says that versions
>> >> you have configured and how,
>> > 
>> > Sorry, I don't see why I should be required to write a file that says
>> > 
>> >     using msvc : 7.1 ;
>> > 
>> > instead of simply being able to request the same in the command line.
>> 
>> I would even suggest that user-config.jam use full paths to
>> compilers.
>
> If our auto-configuration code is _that_ unreliable, then we shoudn't
> advertise this functionality. But seriously, surely we can do better.
> If Python's distutils can do it, so can we.

It's not a matter of being unreliable.  The question is whether it's
conservative or liberal in finding tools.  For the newbie user, we
really want to try hard to find something.  For the regression report,
we really want to make sure the reported version number matches the
actual toolset found.

There's also the matter of historical precedent.  In the past we
didn't have such sophisticated configuration capabilities (e.g. the
SHELL rule), so falling back to using what was in the PATH was a
necessary last resort.  Now we can even verify that the version number
on the toolset we find in PATH matches the version number specified by
the user -- the Python toolset does that, for example.  

So, yes, we *can* do better, and we *should*.  Do we have to do that
much better before we release 1.34?  Probably not, IMO.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com