headers
Will Partain | 1 Jul 2001 23:27
Picon
Picon
Favicon

updates 2001.07.01

Folks, I've just done my usual check-in-a-pile-of-bits to
the SourceForge CVS.

Main message: probably *stay away* from the ARK/arkbase/ark
changes unless you're Matt or Jonathan :-)  They work to at
least a basic level, but I haven't tested them enough beyond
that.

The changes are aiming toward much-improved "provenance"
handling (keeping track of what info came from what
prototypes), which should ultimately have major implications
for How Cool this ARK stuff is.  I will leave that for a
separate message sometime...

I killed the <pre-build> package method (we weren't using
it), and renamed <mkdir-deploy-build> to <mkdir-deploy>.
You shouldn't care.

The other thing on my hit list is to generalize
<constraint>s somewhat; Matt is doing some cool things, too,
even if he is too shy to tell you about them.

Anybody got anything they need help with?

Will
Permalink | Reply | 0 comments |
headers
Will Partain | 3 Jul 2001 21:42
Picon
Picon
Favicon

updates 2001.07.03

I checked in a few things; the CVS tree should be just about
as usable as it normally is.  I fixed one *horrible* bug...

	* utils.py (evalArkString): Use the "host context" we have in our
	hand; otherwise, ark.thing.ArkThing.__getattr__ will invent one --
	incorrectly.

and one Bug That Could Not Survive ...

	* xmlfile.py (ArkXmlFile.read): We must limit the exceptions we
	catch; otherwise, it catches things like parse errors (which we
	really want to hear about!).

The rest was nothing too interesting.  Enjoy.

Will
Permalink | Reply | 0 comments |
headers
Will Partain | 3 Jul 2001 22:18
Picon
Picon
Favicon

ark-events: coolness to come?

Here's something that I think will be Very Cool if/when it
all comes together...

Previously, when the ARK engine ran a "recordable" method,
it would 'touch' a file in its "state" directory.  So, for
example:

   % ls -l /d/ark-state/package/xemacs--21.1.14/compile--sparc-solaris2.6 
   -rw-r--r--   1 partain  sliadmin       0 Feb 23 21:48 /d/ark-state/package/xemacs--21.1.14/compile--sparc-solaris2.6

The system then knows that it doesn't need to recompile
xemacs again.  (Many details omitted)

OK, the next thing is to record, as an <ark-event>, details
of what went into doing the <compile> method for xemacs (or
whatever).  Pls find below a sample record for compiling
wget--1.7 (NB: definitely not the final format!).

We record the code that was run, the cvs-revision numbers
for all the files that told us anything, the values of
<param>s (not shown), the <constraints> that had to be
satisfied for this method to run, and so on.

And the coolness? -- we can process the <ark-events> and see
if the things that went into running the method continue to
be true/unchanged.  If not, we have the choice of re-running
the method.  This has *major* potential for helping to keep
the overall site "fresh".

Let's take xemacs, for example.  You might easily have a
(Continue reading)

Permalink | Reply | 0 comments |
headers
Jonathan Hogg | 3 Jul 2001 23:38

Stuff I've been sitting on


Hi all (especially Will),

I've just committed a couple of changes that I've been sitting on for a 
bit. Nothing major, but I was hoping that the adventurous in the crowd 
(Will) would try them out. I tinkered with the evaluation template stuff 
a bit to try and make it less magic. It seems to work OK for me, but I 
may have missed something.

There's a random change to your new event stuff Will as it was breaking 
on my machine. It couldn't cope with CVS files that came about as the 
result of a merge. It doesn't now to be honest, but it at least battles 
on instead of just expiring.

:-j

--

-- 
jonathan hogg, one good idea ltd, 131 queen margaret dr., glasgow g20 8pd
http://www.onegoodidea.com/ tel:+44-(0)7976-614338 fax:+44-(0)7970-537451
Permalink | Reply | 0 comments |
headers
Harlan Stenn | 7 Jul 2001 09:44
X-Face
Favicon

CONS?

Anybody looking at CONS as a possible framework to replace Make?

From what little I have read, I gather it can easily "fetch" things in a CVS
repository, which may greatly ease the use of CVS-based packages for ARK.

For example, I track the CVS versions of NTP and PIKT (and ark), and if we
had the ability to "use" this in ARK instead of requiring a tarball, it
would make some things lots easier for me.

Oh, I'm working hard on getting the openpackages.org framework to
bootstrap.  I hope to have the first pass finished within the next week.

H
Permalink | Reply | 0 comments |
headers
Harlan Stenn | 8 Jul 2001 10:43
X-Face
Favicon

openpackages.org

We're probably just a few days away from having the bootstrap (non-*BSD)
version of openpackages coughing along.  We got Darwin bootstrapped tonight
and will be doing one of the Linux distros tomorrow.  I should have HP-UX
working by Monday night.

I believe this will be a great help in bootstrapping and running ARK.

If anybody can help get this off the ground, it would be nice.

The site is <http://www.openpackages.org> and the CVS version of the code is
the way to go (at the moment).

Once I get it up and running I want to see source packages written for ark
and its dependencies (python may be the only one now).

I'm also really curious to look at CONS as an alternative to make as I hear
CONS has support for syncing with remote CVS repositories, which could be a
really good thing for users/developers of ARK (and PIKT and NTP).

H
Permalink | Reply | 0 comments |
headers
berto | 8 Jul 2001 15:12

Re: openpackages.org


   From: Harlan Stenn <Harlan.Stenn <at> pfcs.com>
   Sender: ark-dev-admin <at> lists.sourceforge.net

   I'm also really curious to look at CONS as an alternative to make as I hear
   CONS has support for syncing with remote CVS repositories, which could be a
   really good thing for users/developers of ARK (and PIKT and NTP).

Harlan:

I haven't looked at CONS.  But I think that it would have to be a really,
really significant improvement over standard make for us to begin requiring
its use for ARK (or PIKT or NTP).  It's still another package for the user
to fetch, build, and install.  Are those steps as easy as for current (GNU)
make?

PIKT is now available for ten different OSes.  Does CONS run on all OSes,
or all that matter?

Also, I can see where syncing with the ARK CVS repository might be a good
thing, but I'm not so sure with PIKT (or NTP?).  Naturally, syncing with a
CVS repository is optional, right?

Berto

-------------------------------------------------------------------------------
Robert Osterlund, Unix Systems Manager                berto <at> gsbrgo.uchicago.edu
Grad School of Business, U of Chicago                       phone: 773/702-8898
1101 E. 58th Street, #309, Chicago, IL 60637, USA             fax: 773/702-0233
(Continue reading)

Permalink | Reply | 501 comments |
headers
Will Partain | 8 Jul 2001 15:18
Picon
Picon
Favicon

Re: CONS?

Harlan asks:

> Anybody looking at CONS as a possible framework to replace Make?

No.  Another tool that I've heard some (Java) people raving
about is Ant.  Any hands-on review of either would be very
welcome.  Of course, because ARK is a "both/and" kind of
thing (rather than "either/or"), we could very plausibly
have package stuff that used all three :-)

> For example, I track the CVS versions of NTP and PIKT (and
> ark), and if we had the ability to "use" this in ARK
> instead of requiring a tarball, ...

ARK doesn't "require" a tarball; it's just that the bits of
code that happen to be lying around today happen to require
them.  The fundamental thing in ARK land is "same as that
guy over there, except this little bit different".

Below is an (untested) <assemble-source> method that's nothing
to do with tarballs.

Will

<assemble-source>
  <constraint><dependency type="essential" name="." on-method="get-source" /></constraint>
  <param name="CVS"> <at> host:CVS <at> </param>
  <param name="PKG_NAME">!pkg_name</param>
  <param name="PKG_VERSION">!pkg_version</param>
(Continue reading)

Permalink | Reply | 0 comments |
headers
Will Partain | 8 Jul 2001 15:21
Picon
Picon
Favicon

Re: openpackages.org

Harlan writes:

> The site is <http://www.openpackages.org> and the CVS version of the code is
> the way to go (at the moment).
> 
> Once I get it up and running ...

When you get to that point, could you give us a
micro-tutorial, along the lines of "I typed X, then Y, then
Z, and the following four wonderful things happened?"

It would be a lovely thing indeed to build on a
purpose-built packaging tool.

Will
Permalink | Reply | 0 comments |
headers
Daniel Hagerty | 8 Jul 2001 21:06

Re: openpackages.org

 > I haven't looked at CONS.  But I think that it would have to be a really,
 > really significant improvement over standard make for us to begin requiring
 > its use for ARK (or PIKT or NTP).  It's still another package for the user
 > to fetch, build, and install.  Are those steps as easy as for current (GNU)
 > make?

    My company has started looking at cons; we're pushing the limits
of what one can do (sanely, maintainably) with make.  Many other
moderate scale projects seem to run into make horrors quickly.  For
some interesting insight into why, see
http://www.pcug.org.au/~millerp/rmch/recu-make-cons-harm.html .

    I'm not really the cons point man for our company, but my
understanding is that cons is just perl code, so should be fairly
portable to anything that runs perl (the web page claims that NT is
known to work well).
Permalink | Reply | 501 comments |

Gmane