Re: Future of cons?

[Steven Knight]

> Hi all, I am relly glad this thread got so far and I hope we soon start
> talking about what is to be done.
> 
> Before I get to the serious stuff I'd like to comment a mail that pissed me
> off.
> 
> > -----Original Message-----
> > From: Doug Alcorn [mailto:address@hidden
> > Rajesh Vaidheeswarran <address@hidden> writes:
> > 
> > > Having said this (and before my net connectivity dies)... I would like
> > > address the more specific issues at hand once I return to the US on the
> > > 22nd of June.
> > 
> > Sounds good to me.  There's no hurry to do anything that I know of.
> 
> As much as I have respect for Rajesh private life and travels, I find it
> very frustrating to read that "there is no hurry". Is this some kind of bad
> joke or don't give a **** for what the users write. I have been complaining
> for months so the web page is updated, so no new comer get in the same mess
> as I did. This week I saw a "can't compile hello_world.c" mail. Wake up
> Doug, the web site is a mess, it waists new comers and old users time
> because YOU think it's OK to wait another month and another month to take
> away a broken version. If the web page IS the place where things are to be
> found than it HAS to be right. Find someone else to run the web site or find
> a better system.

I agree this is a serious problem, but cut Doug some slack, will you?
The issue is that dsmit.com is a commercial site, and only Rajesh can
update it.  If

> Perl front end to Scons: Yes it _is_ possible (have a look at
> Inline::Python). If we decide that it is the main stream, I can run some
> tests. This would allow us to write scripts in perl and python both using a
> python core.

That would be helpful.  I just took a quick glance, and it looks like
Inline::Python should provide enough functionality to get at the SCons
build engine from Perl.  I'm sure there'll be a fair amount of time lost
to figuring out quirks, but at least the technology is there.

Do you know if Inline::Python requires either a specific version of Perl
or specific version of Python?

> Using SCons as the core: I think we must have a serious look at that. I mean
> that some people should try to present what is good and not. What the
> limitations. Is that what we want?

I'm biased, but I'd suggest we follow this path.  The SCons build engine
is architected for it from the ground up, it has a very slick job
architecture for parallel builds (courtesy Anthony Roach), it supports
easy specification of new Builder and Scanner objects, and 0.08 will
have a framework for simple and flexible specification of Platform and
Tool specifics in environments.

> Core language: if the front end problem is fixed, what core language is to
> be used? Perl, python, C, C++ other. 

My vote should be obvious...

> Linux: I read a mail that states that we should try to get the linux kernel
> build through Xcons. Ok the idea is fun but it's no more than that. When
> we've learn can walk we can try to run after the linux build system.

I agree, but it does make sense to pay a little attention to the linux
kernel build requirements along the way so that you can make sure the
architecture doesn't preclude doing it some day.

        --SK