Re: [jamboost] Boost.Build v2 destabilization

From: David Abrahams (dave_at_[hidden])
Date: 2003-02-18 11:59:00

Vladimir Prus <ghost_at_[hidden]> writes:

> David Abrahams wrote:
>> I'm about to check in some changes which pass all of our regression
>> tests, but which I fear may destabilize the build system for some of
>> you.
>>
>> The change is that the "composing-generator" distinction is now
>> obsolete. 
>
> Looking at your changes, I see that you've done something
> different. 

I wanted to make the smallest possible change that could work.

> You've allowed composing generators to be used
> everywhere. The distinction is still there: composing generators and
> ordinary generators differ in semantics.

So the difference you're talking about is just how many sources the
generator can consume, right? I wonder if we shouldn't use patterns
and have a single kind of generator instead? IOW, we'd specify the
RSP generator as 

OBJ * STATIC_LIB * -> RSP

While most 'C' compilers would be:

C -> OBJ

>> It was interfering with response files under NT and didn't
>> seem to be doing anything useful. 
>
> Agh... you created EXE from RSP and RSP generator won't be used
> because it's composing. 

Right.

> I believe the change you've made is reasonable.

That's a relief!

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