1 |
>>>>> On Sat, 13 Aug 2011, Ciaran McCreesh wrote: |
2 |
|
3 |
> On Sat, 13 Aug 2011 12:10:45 +0200 |
4 |
> Ulrich Mueller <ulm@g.o> wrote: |
5 |
>> >> Except that large parts of the tree rely on packages in RDEPEND |
6 |
>> >> being available in pkg_*. |
7 |
>> |
8 |
>> > Then those packages are broken. |
9 |
>> |
10 |
>> Welcome to reality. ;-) |
11 |
|
12 |
> Reality is that RDEPEND cycle breaking happens, so those packages |
13 |
> only work if by fluke they're not in an RDEPEND cycle. |
14 |
|
15 |
> Now, we *could* weasel our way out of it by saying that it's illegal |
16 |
> for any repository to include a package that would induce such a |
17 |
> cycle between packages that rely upon the intersection of DEPEND and |
18 |
> RDEPEND being available in pkg_*. But that's pretty horrible, so if |
19 |
> we do that then we really need to fix things in future EAPIs. |
20 |
|
21 |
I'd still suggest that for existing EAPIs we should go with mgorny's |
22 |
latest patch, simply because "packages common to DEPEND and RDEPEND |
23 |
(but see below)" is a more accurate description than "none". |
24 |
|
25 |
However, we should keep in mind that the situation with respect to |
26 |
pkg_* is not well defined. Therefore a new dependency type could be |
27 |
beneficial for future EAPIs. |
28 |
|
29 |
Ulrich |