Re: [Boost-testing] Branch independent testing?

From: Beman Dawes (bdawes_at_[hidden])
Date: 2007-10-13 17:13:49

Rene Rivera wrote:
> Beman Dawes wrote:
>> Rene Rivera wrote:
>>> Beman Dawes wrote:
> 
>>   ...One think I'm kinda struggling with right now is having to build
>>> process_jam_log as it depends on Boost.Filesystem. Hence it makes 
>>> creating an independent set of test tools hard. At the end getting rid 
>>> of PJL is ideal, but we aren't there yet.
>> Having bjam generate the xml files directly would be a big step forward 
>> in reliability. But I'd prefer we get 1.35.0 out the door before 
>> investing any time in that.
> 
> Well, it's now going to be a real problem. PJL depends on Filesystem, 
> which depends on System. But there is no System in the release branch.

But the version of Filesystem in the release branch doesn't use System. 
System will be added to the release branch before or at the same time as 
Filesystem gets updated.

> Which either prevents PJL from using the "current" Filesystem, or we 
> make testers have both Trunk and Release around...

I been assuming that testers will need a full boost tree for anything 
they test. Anything else is complex, messy, and runs the risk of testing 
some files that are other than what the runner is supposed to be testing.

 > Or, someone could port PJL to Python?

Please, no! I really think you are worrying about a non-problem!

--Beman