Re: Why do no targets depend on the Jamfiles?

From: David Abrahams (dave_at_[hidden])
Date: 2005-04-06 11:13:14

Vladimir Prus <ghost_at_[hidden]> writes:

> On Tuesday 05 April 2005 18:26, David Abrahams wrote:
>
>> > I believe it was me. And I still consider it to be a major problem but I
>> > was alone and I did not manage to convice anybody about it
>> > seriousness.
>>
>> I've been convinced of it for years.
>
> Oh, well.
>
>> > Well, I agree that it is not a good idea to rebuild everything on
>> > every Jamfile change. IMHO the built files should depend on the
>> > commands used to build them...
>>
>> Right. We should write the contents of the commands out to some files
>> as part of the build process and then make inequality a precondition
>> of outdating new targets. 
>
> Should not be very hard.

Probably not too hard; just another thing to do.

>> Or we could wait for the Python integration 
>> to be finished and let Scons' mechanisms take care of it...
>
> And this might be hard. Not only we need to finish Python port, but also 
> integrate SCons.

I think (well, I hope) you're overestimating the difficulty of that
part.

>> ...which we can't do until BBv1 is retired. How's that project coming
>> along, Volodya?
>
> I have run all Boost tests (including Python) with V1 and V2 and compared the 
> results. Except for some noise (like random numbers used by some tests), they 
> are equal.
>
> What remains is:
>
> - tweak process_jam_log so that it handles V2 style grists (affects
> parsing of "skipped" messages)
>
> - make Python work on Windows. I'd appreciate help with implemeting
> or testing this.

I'll be very happy to try. I will be very busy with travel, though,
between tomorrow and the end of the month. What are the issues,
currently?

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