Re: [Boost-testing] Metacom VC6.5 tests seem to be running MSVC7.1

From: David Abrahams (dave_at_[hidden])
Date: 2006-11-13 09:34:33

Aleksey Gurtovoy <agurtovoy_at_[hidden]> writes:

> Rene Rivera writes:
>> Aleksey Gurtovoy wrote:
>> > Rene Rivera writes:
>> >> AlisdairM wrote:
>> >>> There are suddenly a lot of unexpected passes for VC6.5, and looking at
>> >>> the output it seems that vcvars from Visual Studio 2003 is being used
>> >>> to set the paths.
>> >> Weird.
>> > 
>> > FWIW, we build bjam with Visual Studio 2003 (on every cycle), which
>> > involves running the corresponding "vcvars32.bat" from the same
>> > "regression.py" session that then proceeds with the regression tests
>> > themselves. Could this be the culprit?
>> 
>> Yes it could. I'm not sure what the effect, or non-effect, of running 
>> multiple vcvars scripts in the same env are.
>
> Doesn't Boost.Build itself have to run the corresponding 'vcvars'
> before invoking the compiler?

It does do that.

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