Re: try-ark problem: reveal arkbase
Unknown <[email protected]>
| Newsgroups | gmane.comp.sysutils.ark.devel |
|---|---|
| Message-ID | <[email protected]> |
Harlan wrote:
> Regardless, I gather we all agree that it is a Good Thing
> to be able to mix/match production/testing/development
> teams.
Yep.
> Next, I was thinking that (assuming it doesn't work this
> way already) while a policy team (sidai) is a logical
> place to put all of the packages that *MIGHT* be
> built/deployed/revealed, I think the policy team should
> only hold the *specs* on things and not do any "work".
> All of the compiling, revealing, and deploying should be
> done by glasli1-style teams.
Yes, that's the way it works. A typical scenario is: Say I
set about installing some random tool, let's say "MagicPoint
version 1.07" (I haven't got this one working yet, BTW >
:-)...
* I start with <prototype team="sidai" name="GNU">
(i.e. I hope "./configure && make && make install" works)
* If (sadly) I have to do Weird Stuff to get it to work:
- If it's something simple, like it has a non-standard
*.tar.gz filename, I can probably fix it with a <param>
setting.
- Let's say it's something stranger, like the tar file is
hidden inside another tar file (don't laugh -- these things
happen; installed lsof lately???)...
In that case, I write a special <assemble-source> method
and pop it in sidai/package/magicpoint.xml, and
change/add a prototype of
<prototype team="sidai" name="magicpoint" />
* I put that stuff in sidai on the simple rationale of "If
Matt wants to install this package, he'll have to do the
same thing". More importantly, he'll look over what I
wrote and point-out/fix any bugs.
* If whatever weirdness I need to do to install MagicPoint
is really specific to my site, I don't put in in Sidai;
rather, I just pop the override straight into the site
package/magicpoint--1.07.xml file.
All of the above is Everyday Routine ARK Business.
One thing the current ARK stuff does *not* address is sites
that aren't just like-minded (e.g. both use Sidai), but want
to be very, very similar -- a common toolset, modulo the odd
version difference. (Matt and I have teams like this.)
We do not currently have an efficient way to express this
situation.
Ideas? (Help, Jonathan :-?)
Will