[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Spliting packages between pkg and pkg-data

Nicolas Boullis <nboullis@debian.org> writes:
> On Sun, Nov 20, 2005 at 12:39:24PM -0800, Russ Allbery wrote:

>> Well, one practical concern is that it makes it harder for other
>> utilities like lintian to analyze the package properly.

> Well, that's an argument I don't like. Those are tools that help us
> check our packages are correct, but I don't think it is reasonable to
> modify otherwise correct package only to make such tools happy. For the
> same reason, I'm not happy with setting lintian overrides, as I consider
> it bloats the packages for no good reason (although the "bloat" is very
> limited").

Sure, ideally lintian would be able to cope with any packaging practice
that actually works.  However, this is a lot like coding standards in a
programming language.  There's a lot of evidence to indicate that
following standards that are more restrictive than the language actually
requires can make it easier to find bugs and detect problems, even though
it isn't necessary and there are times for exceptions.

I think lintian is one of the best things about Debian since it lets one
catch a multitude of problems that are otherwise difficult to detect and
easy to forget to look for.  Using any sort of lint tool is always a bit
of a tradeoff, though, since there will always be things that are correct
but impossible for the lint tool to analyze for one reason or another.
lintian's particular flaw is that it can't do most forms of interpackage
analysis, so one has to provide all the details it needs to do its job in
a single package whenever possible.

In the absence of any compelling reason why something *needs* to go in
another package, it therefore makes sense to me to keep it together to
make life easier for QA tools.  But then, I'm a big believer in lint,
-Wall, and similar checking tools, up to and including minor modifications
to my coding or packaging style to avoid false positives.  I think the
tradeoff is more than worth it; the more effective I can help the lint
tools to be, the more consistently high-quality my packages are.

Russ Allbery (rra@debian.org)               <http://www.eyrie.org/~eagle/>

Reply to: