Bug#41232: debian-policy: [PROPOSAL] Build-time dependencies on binary packages

[Antti-Juhani Kaijanaho <gaia@iki.fi>]

On Mon, Jul 26, 1999 at 10:54:38AM +0200, Roman Hodek wrote:
> I see your point, and I can live with the Arch- variants if a majority
> wants them.

The majority?  There have been, what, probably less than ten people
involved in this discussion.  I don't think a majority vote among them
would be of any indication of what majority want.

> But I still think they just make more work, both for
> maintainers who have to define them, and for tools which read source
> dependencies.

The work for the tools is insignificant.  For maintainers - well, they
can always use Build-Depends as a catch-all and use it like they would
use it in your model.

Here's how the fields would map to the targets

Six-field:
  Build-(Depends|Conflicts): build, binary-arch, binary-indep
  Build-(Depends|Conflicts)-Arch: binary, binary-arch
  Build-(Depends|Conflicts)-Indep: binary, binary-indep

Four-field:  
  Build-(Depends|Conflicts): build, binary, binary-arch, binary-indep
  Build-(Depends|Conflicts)-Indep: binary, binary-indep

(I just noticed that we must not have build map to anything else than
the plain Build-Depends in either model; as you noted earlier, it would
just make the other fields void of purpose.)

If in the four-field model, a maintainer could say

Build-Depends: foo bar
Build-Depends-Indep: baz

where bar is being needed only for the arch-dependent packages, then in the six-field
model she can say either

Build-Depends: foo bar
Build-Depends-Indep: baz

(the same!) or

Build-Depends: foo
Build-Depends-Arch: bar
Build-Depends-Indep: baz

whichever suits her.  Of course, the two-field way in the six-field model
will be deprecated when (if) we allow building only arch-all packages.

Would the six-field model be okay with you in this light?

> > Build-time dependencies must specify version number(s) of package(s)
> > if the version in the current Debian stable distribution is not
> > adequate. If this is necessary usually a >= dependency should be
> > used.
> 
> This looks like an important point. Anybody against it?

I would be.  I don't like introducing the variable concept of "stable
distribution" in here.  I'd more like to use the rule we use with binary
dependencies: use versioned dependency if some version would not be
acceptable, independent of the distributions involved.

-- 
%%% Antti-Juhani Kaijanaho % gaia@iki.fi % http://www.iki.fi/gaia/ %%%

   "... memory leaks are quite acceptable in many applications ..."
    (Bjarne Stroustrup, The Design and Evolution of C++, page 220)