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

From: Vladimir Prus (ghost_at_[hidden])
Date: 2006-12-15 05:53:23

Aleksey Gurtovoy wrote:

> Vladimir Prus writes:
>> I'm pretty sure this is yet another fallout from "--toolset" option and
>> multiple toolsets.
>> 
>> May I suggest that the "--toolset" change be reverted, and the
>> command line be:
>> 
>>    borland-5.6.4 borland-5.8.2 msvc-6.5 msvc-7.0
>>    cw-8.3/runtime-link=static
> 
> Don't we want this working before we release?

Why would we? "toolset=" autoconfiguration is a way to help first time
users.

I think saying that if one wants to use more that one toolset he has to
configure them is acceptable.

> 
> 
>> If I understand correctly, change to use "toolset=" or "--toolset"
>> is not a pressing at all -- the only difference is that "toolset="
>> and "--toolset" performs toolset auto-configuration and:
>> 
>>         1. Most testers have perfectly working user-config.jam
> 
> Sort of, I'd say. We (metacomm) still have an outstanding CW issue
> which we are not sure what to attribute to (toolset config, toolset
> command line specification, something else?):
> http://thread.gmane.org/gmane.comp.lib.boost.testing/3515/focus=3698

I'm aware of this issue; I plan to look at it this weekend.

>>         2. Using autoconfiguration for regression tests might be a
>>         bad idea in general.
> 
> Why?

Because instead of having a single file that's says that versions you have 
configured and how, autoconfiguration adds yet another variable that 
depends on your environment. That's like using

        using msvc : all ;

in user-config.jam -- that's handy for end users, but for regression test
this is not very good -- nobody knows what's stored in your registry.

- Volodya