Re: autoconf (was: Re: [Gnu-arch-users] [ANNOUNCE] tlator-0.1 initial release)

[wave++]

In local.unid.arch, you wrote:
>>   - Relocated builds:
>>   - Ability to avoid recursiviness:
>>   - Ability to use standard make syntax:
>
> These are possible with Make, in my experience.  Maybe easier with 
> makepp, but it's hard to tell.

Of course, you can do *whatever* you want using make.
If you want to reduce the concept to "generic dag solver", you don't
need anything more than make.

>>   - The ability to setup arbitrarily complex rules, through perl
>>     expressions.
>
> This looks like a bug.  Likewise building makedepend into the core.

Make isn't a simple DAG solver, anyway. You don't need to know about
files when solving a dag. But you are solving a DAG around your build
system.

How many passes can you remove in the process? Can you use dynamic rules
to do that without using a two-pass process? How complex is your build
system?

In fact, automake IS using Make!

> Oh well, whatever works for you.  I'll probably go look again at jam.  
> I'm not sure any has enough advantage to be worth the cost, as with 
> others.
>
> Returning to the original point, most people on the list seem to 
> appreciate that arch is an quantum leap on from CVS.  It's not clear 
> that any of the alternative build tools have made a similar size leap 
> forwards from Make.  They still try to solve essentially the same 
> problem: construct a DAG and solve it.

Basically yes. I perfectly agree with you.

Like arch basically "constructs changesets and exchanges them".
Nothing that tar + diff --recursive + mail can't do :)

-- 
'(wave++ "Yuri D'Elia" "http://www.yuv.info/";)