Re: Pointers for packaging appreciated (icecream)
On Sat, Dec 18, 2004 at 12:18:39AM +0100, Magnus Therning wrote:
> Package: icecc
> Description: Distributed compiler, client and server (icecream)
> to the fastest free server and is as this dynamic. This advantage pays off
> 1. The original package doesn't have any manpages. Rather than writing
> my own right now I'd like to (temporarily) use the 'undocumented'
> man page. How do I do that?
Apparently you don't; quoting from dh_undocumented(1):
Do not run!
This program used to make symlinks to the
undocumented.7 man page for man pages not present in a
package. Debian policy now frowns on use of the
undocumented.7 man page, and so this program does
nothing, and should not be used.
If you *really* want to you could make a symlink from man/man1/icecc.1
to dh_undocumented.1.gz. Instead, I would encourage you to add it to
./debian/TODO and let lintian continue to warn you. And nag upstream.
> 2. How do I handle the cache directory it uses? Currently I only
> create it at install, but don't remove it afterwards. I don't want
> to remove it in case of an upgrade, only with an actual remove.
I'm not sure I understand entirely; but if you just include the
directory (/var/cache/icecc/?) in the package as a normal directory
(possibly via ./debian/dirs), then it will do what you want. If its
name is determined at install-time by the admin, then you can just use
the postrm script to query debconf and rmdir it.
> 3. The monitor is a KDE app. I don't use KDE myself (and I have no real
> interest in digging any deeper into it). Instead of heavily
> modifying makefiles to exclude it at build time I opted to put it in
> a separate package. I'd really like some feedback on that package
> (I don't even know if it works).
A separate package is probably good, especially so that the
Distributed Compiler doesn't depend on 675==$(aptidue search kde |wc -l)
> 4. I renamed the package from icecream to icecc since the package
> icecream already exists in Debian. I've written something about it
> in the README.Debian, is this a good way of handling it?
Sounds good to me.
> 5. I would love if someone who has access to some serious machines
> would install and try it out. I only have tried on three single-CPU
> machines so far, it'd be nice to get confirmation that it works
> properly also for larger setups.
If I am offered the job I want, I'll have access to a 24 node cluster
and I will do just that.