Re: bjam patch to provide necessary info for process_jam_log

From: David Abrahams (dave_at_[hidden])
Date: 2004-10-10 19:35:24

Aleksey Gurtovoy <alexy_at_[hidden]> writes:

> Aleksey Gurtovoy writes:
>> David Abrahams writes:
>>> "Aleksey Gurtovoy" <alexy_at_[hidden]> writes:
>>>
>>>> Folks,
>>>>
>>>> To elaborate -- currently bjam doesn't output failed/skipped targets that 
>>>> don't have associated actions. In cases when this happens, it makes the 
>>>> task of using bjam log to reengineer the dependency chain (the task of 
>>>> "process_jam_log" executable) impossible. A particular instance of this
>>>> situation can be observed in the current Boost.Python tests.
>>>
>>> I don't understand why a core bjam change was neccessary. Seems to
>>> me you could just add an empty action for the rule in question.
>>>
>>> No?
>>
>> And who is going to track this down next time somebody adds a similar
>> rule to their bjam file?
>
> To make it clear: I'm not insisting that the problem should be fixed
> this particular way. What I'm strongly opposite to is rejecting a
> possibly correct, at large, fix on the basis that a local workaround
> for one particular instance of the problem is possible. We've spent a
> considerable enough amount of time getting to the roots of the issue
> to feel that it should be fixed once and forever.
>
> I'd accept the fact that nobody has enough expertise in the core to be
> able to say whether the proposed change would break anything or to
> explain what was the motivation behind the original logic in the
> first place. I really hope that it's not the case, though.

FWIW, that area of the core is probably the least understood.

-- 
Dave Abrahams
Boost Consulting
http://www.boost-consulting.com