1 |
On Fri, Mar 6, 2015 at 3:18 AM, Ben de Groot <yngwin@g.o> wrote: |
2 |
> As a follow-up, I did a qgrep in the tree for python support the other |
3 |
> day, and it turns out we have: |
4 |
> |
5 |
> - almost 2500 ebuilds with 2.7 support only |
6 |
|
7 |
My gut tells me this number is somewhat inflated. A large portion of |
8 |
these may be ebuilds that very few people use. Also, something which |
9 |
only supports python2 is more likely to be old/unmaintained than |
10 |
something supporting python3. |
11 |
|
12 |
However, I do agree that there are currently many ebuilds which |
13 |
support python2 only, that much is undeniable. |
14 |
|
15 |
> - less than 1500 ebuilds with support for both 2.7 and 3.* |
16 |
> - only just over 100 ebuilds that support 3.* only |
17 |
> |
18 |
> Based on these numbers, it is clear to me that the most reasonable default is: |
19 |
> |
20 |
> PYTHON_TARGETS="python2_7" |
21 |
> PYTHON_SINGLE_TARGET="python2_7" |
22 |
> |
23 |
> and leave it to the user to change that if they really want multiple |
24 |
> python versions. |
25 |
|
26 |
If the goal of the profile default is to provide *minimal* usable |
27 |
settings, then I agree that would be appropriate. I'm not willing to |
28 |
call it the most "reasonable". |
29 |
|
30 |
I would guess that most users don't really care about having two |
31 |
versions installed. So long as their applications work, I think most |
32 |
people don't care about python much at all. |
33 |
|
34 |
Personally, I look at it as enabling a modern version by default, and |
35 |
keeping a legacy version enabled so things don't break. Ideally, we |
36 |
would only enable python2 if the user needs some app that only works |
37 |
with python2, but there's no way to do that with use flags. |
38 |
|
39 |
> This would also significantly reduce the size of our stage3 tarballs. |
40 |
|
41 |
This would actually be a nice side effect of reducing the default |
42 |
PYTHON_TARGETS to either python2 or python3. |
43 |
|
44 |
If you really want to shrink the tarball, there is a bug open about |
45 |
stripping tests from dev-lang/python which would have an even more |
46 |
significant effect. |