Re: [Boost-testing] Apple Mac OS testers?

From: Aleksey Gurtovoy (agurtovoy_at_[hidden])
Date: 2007-10-25 13:25:24

On Thu, 25 Oct 2007 10:18:46 -0500, Vladimir Prus <ghost_at_[hidden]> wrote:

> Aleksey Gurtovoy wrote:
>
>> On Wed, 24 Oct 2007 15:36:53 -0500, Rene Rivera <grafikrobot_at_[hidden]>
>> wrote:
>>
>>> Niko Demmel wrote:
>>>> http://www.boost.org/tools/regression/xsl_reports/runner/instructions.html
>>>
>>> Yea, we need to make new instructions.
>>
>> I'd much prefer Boost.Build to reduce the level of the initial
>> setup hassle, which has raised significantly by v1 to v2 switch.
>> The more intimidating and demanding the setup instructions are,
>> the less volunteers we get.
>
> The V2 specific instructions, found at:
>
>         https://zigzag.cs.msu.su:7813/boost.build/wiki/TestingOnBoost
>
> involve writing two lines in user-config.jam.

Firt of all, you *have* to have user-config.jam, which is a usability
regression from v1 (from regression.py user's standpoint anyway).

Second of all, and we had this discussion before [1], I do believe
that the whole toolset pre-declaration business is redudant and we can
do better. Repeating myself, if Python's distutils can do it, so can we.

> Everything else is regression.py voodoo.

regression.py does not require anything else, that's the point. We've
spent a considerable amount of thought to make its requirements
as simple as

     * Python 2.3 or higher
     * Some spare disk space (~5 Gb per each tested compiler)

Naturally, its upsetting for us to see these degrade without a good
reason.

[1] http://thread.gmane.org/gmane.comp.lib.boost.testing/3770/focus=3782

--
Aleksey Gurtovoy
MetaCommunications Engineering