Re: [Tack-devel] Question about other ACK distributions...

[David G. <dg...@co...>]

On Friday 27 January 2006 16:05, Peter Fr=F6hlich wrote:
[...]
> So I gather that you want to (a) keep support for old/strange platforms
> and (b) improve the thing enough to make it useful for "modern"
> development platforms? That would be in line with what I would like to
> see as well, and in that case I'll start delving into the sources a bit
> more now. :-)

I want to keep support for *targeting* all the old platforms, but I don't s=
ee=20
much need right now for being able to cross-compile the ACK itself for one =
of=20
those platforms. The ACK's main value, as I see it, is allowing Big Machine=
 X=20
to compile for Small Machine Y.

> BTW, is there a short "hackers guide" or other overview document that
> describes the structure of the whole beast? I got a little dizzy
> navigating around the tree yesterday.

Unfortunately no, and I know exactly how you feel. The build tree is very,=
=20
very big, and very, very weird. There are automatically generated headers a=
nd=20
strange scripts everywhere. A lot of the code is incredibly archaic K&R C=20
that does things like prototype their own private copies of malloc inside a=
=20
function so they don't have to bother including stdlib.h. Fun! This was all=
=20
fine back in the day, but the world has moved on and gcc isn't very happy=20
about it.

It's worth browsing the documentation, which is a bit spotty but what there=
 is=20
is pretty good. It should give you an idea as to how the system's put=20
together, but you may need to do a certain amount of reading between the=20
lines. It's mostly all online at http://tack.sourceforge.net.

The build system itself is based on several layers of recursive makefiles a=
nd=20
is, to put it mildly, incomprehensible and undocumented. (It's one of those=
=20
things whose design exists only in the head of the person who wrote it. I'v=
e=20
done code like that. Sigh.) I've pretty much given up trying to touch it an=
d=20
am ripping it out completely and replacing it with something more modern.

> Also, at least on OS X, it's=20
> kinda frustrating to build right now: The scripts can't make their /tmp
> dirs for some reason, and even once you make all of them by hand some
> things apparently didn't get copied right (see below). Any hints where
> those things should have been created?

I did try to build 5.6 on SourceForge's OSX machine, and failed, but I forg=
et=20
exactly why. My development machine is Linux, so that's what I do the bulk =
of=20
the testing on.

I'm going to wait until I have the build system up and running, which is go=
ing=20
to be a while, I'm afraid. Then I can do another Linux release so people ca=
n=20
play with it, and then after *that* tackle the portability issues... sigh.=
=20
The ACK is quite capable of eating all my spare time if I let it, as are al=
l=20
my other projects (one of which is to rebuild my house, which I'm afraid ha=
s=20
priority right now).

=2D-=20
+- David Given --McQ-+=20
|  dg...@co...    | "While I write this letter, I have a pistol in one
| (dg...@ta...) | hand and a sword in the other." --- Sir Boyle Roche
+- www.cowlark.com --+=20