1 |
Hi, |
2 |
|
3 |
A small batch of patches intended to improve & fix |
4 |
python_wrapper_setup(), pretty much thanks to perfinion. |
5 |
|
6 |
Long story short, he noticed two issues: |
7 |
|
8 |
1. we are missing pythonN-config, both in wrappers and in eselect-python |
9 |
(addressed by patch 2), |
10 |
|
11 |
2. since python3.4, some symlink reading magic is applied that causes |
12 |
python to use wrong prefixes if python-config is symlinked somewhere |
13 |
(addressed by patch 1). |
14 |
|
15 |
While at it, I also added some 'unsupported' wrappers to ban trying to |
16 |
use pythonY (accidental fallback to system python) when pythonX is |
17 |
selected via wrapper, and changed the exit code for unsupp-scripts. |
18 |
|
19 |
Patches 1 & 4 should be harmless. |
20 |
|
21 |
Patch 2 should not cause issues either. Some scripts may start using |
22 |
pythonN-config when they used to fallback to python-config. However, |
23 |
fixing eselect-python side of this may actually cause issues if this |
24 |
causes configure scripts to prefer python3-config over wrapped |
25 |
python2-config. Patch 3 addresses this. |
26 |
|
27 |
Patch 3 is going to cause issues if configure scripts look for |
28 |
pythonN-config before python-config. For example, in python2.7 they are |
29 |
going to find python3-config and try (and fail) to use that, while right |
30 |
now they falled back to python-config. However, the worse would happen |
31 |
if we fixed eselect-python without committing this patch first. |