Re: lintian: r77 - in trunk: checks debian testset testset/binary/debian
On Sun, Feb 29, 2004 at 01:50:33AM +0100, Frank Lichtenheld wrote:
> On Fri, Feb 27, 2004 at 03:04:24PM +0100, jeroen@wolffelaar.nl wrote:
> > Log:
> > checks/huge-usr-share:
> > + Check for a big /usr/share (currently 250kB) in arch-dependent
> > packages (Closes: #232849)
> > Thanks Steve McIntyre for the idea
> >
> > Not sure what the treshold should be, and maybe make it also dependent on the
> > .deb size? I.e., a 100MB deb with 300kB /usr/share is maybe not that bad...
>
> Hmm, I would be very cautious with this check.
> As I also said on -devel (while Steve was obviously not listening): go
> for the obvious cases. We have and had already lintian checks that give
> many falso positives (the Perl module location one is a prominent
> example), which I see as very bad, because it hurts the confidence in
> lintian, which is very very bad, since we cannot _order_ people to use
> lintian. But I'm going off the topic... Perhaps my opinions are also
> influenced by this, uhmm, not-so-perfect mass bugs filing of Steve.
Personally, I find a lintian warning much different (less intrusive and
better ignorable if considered not applicable to this case) from a real
bug report. But indeed, care should be taken to not overdo it.
> I would go for something > 5 MB and/or > 95% and skip the check
> completely for packages smaller than 5 MB or something like that.
> Make it 4 or 3 MB and 90%, but 250k is IMHO way too low.
I think this is a too high limit. The goal of the warning is IMHO that
maintainers of packages where one could split some common package out
are presented with this warning, so that they are reminded of the
possibility to split. Being too conservative would make this warning not
presented to the intended audience.
But again, what the limit should be, I really don't know precisely, in
any case, let's first go for the most conservatist approach, making it
more strict is always possible. Feel free to change the limits to your
taste, re-evaluation can happen later.
> Before including this in a lintian upload I will run it on a full (or at
> least one arch) Debian mirror and investigate the output. I would have
> done it now, but there seems to be a problem with the svn repository:
> djpig@djpig:~/debian/lintian/nmu-svn/trunk$ svn update
> svn: PROPFIND request failed on '/lintian/trunk'
> svn: PROPFIND of '/lintian/trunk': timed out waiting for server
> (http://svn.wolffelaar.nl)
Solved, my computer went crazy, and I was visiting my parents.
--Jeroen
--
Jeroen van Wolffelaar
Jeroen@wolffelaar.nl (also for Jabber & MSN; ICQ: 33944357)
http://Jeroen.A-Eskwadraat.nl
Reply to: