Re: Bug#443769: rt2500-source: Missing dependency on bzip2
Romain Beauxis wrote:
> Let's say that it's the quantitative approach. Other approaches are just
> chatty chatty.
Well, quantitative must not always be the best thing. And if it should
be an argument one should create *proper* stats.
> (This search is not adequate, it matches non-module packages like
> ocaml-search)
Yeah, you name it. I think we should not depend on too much half-baken
stats, that might not be true. I have refined your stats by searching
only with the list of m-a packages and yes the number of bz2 tarballs is
actually higher then the number of gz tarballs.
Stats are:
Total number of packages that could be analyzed: 43
bz2 Packages: 26
gz Packages: 17
So the real percentage seems to be 60% vs. 40%.
> So that makes roughtly 75% of bzip2 tarballs and 25 % of gz tarballs.
> That number would mean to me that it is more likely to have a recommends than
> a depends.
Yes, thats what I first said. And if that is the better way then
installing it from within module-assistant (what I agree with after
all), then thats the way it should be done with build-essential, too.
That is for consistency reasons.
> For instance, should module-source package also depends on module-assistant ?
That would not make much sense, but a "Suggests:" would make sense.
> This is the case for many packages but I don't think module-assistant is
> required to build a package, isn't it an *assistant* ?
It is not required. Yes it is an assistant.
> A simple policy like "source packages must recommend module-assistant and
> should provide gzip tarballs" would give a common answer, given that it's not
> a technical issue as far as I see it...
Agreed. A policy defining which tarball to choose and which recommends
to setup would ease a lot.
Regards,
Patrick
Reply to: