1 |
On Mon, Jul 30, 2012 at 12:53 PM, Dirkjan Ochtman <djc@g.o> wrote: |
2 |
> On Mon, Jul 30, 2012 at 5:56 PM, Mike Gilbert <floppym@g.o> wrote: |
3 |
>> Where else in the tree could this concept be applied? |
4 |
> |
5 |
> I think php5 was first handled this way? Maybe apache2? |
6 |
> dev-python/jinja2, for sure. |
7 |
> |
8 |
>> I would prefer to avoid going through the EAPI process if it is just |
9 |
>> going to be used for python. And even then, I'm not really excited |
10 |
>> about the prospect of explaining it. |
11 |
> |
12 |
> Nor am I, really. And may be we should pursue your idea at the same |
13 |
> time. But I think it would sure be nice if we can prevent painting |
14 |
> ourselves into a similar corner a few years down the road. |
15 |
> |
16 |
|
17 |
Ok. I think we will need something a bit more formally defined before |
18 |
we take such a proposal to a wider audience. |
19 |
|
20 |
>> A solution that works within the confines of the current EAPI spec |
21 |
>> should be greatly preferred. |
22 |
> |
23 |
> So do you know how many ebuilds we'd have to update to get it right? |
24 |
> |
25 |
|
26 |
At the moment, no. If someone could help me write a script to identify |
27 |
affected packages, that would be great. |
28 |
|
29 |
If said script uses the portage api to determine the dependency tree, |
30 |
we will want to modify python.eclass first to reduce the number of |
31 |
hits. This could be done in a local copy of the tree. |