| 1 |
On Wed, Oct 23, 2013 at 9:13 AM, hasufell <hasufell@g.o> wrote: |
| 2 |
> Since I maintain blender I have come across quite a few frustrated |
| 3 |
> users already: https://bugs.gentoo.org/show_bug.cgi?id=488976#c7 |
| 4 |
> |
| 5 |
> I am not sure myself. On one hand we don't need python-updater anymore |
| 6 |
> and have very tight dependencies that ensure that all needed modules |
| 7 |
> are always available for the desired implementation. |
| 8 |
> |
| 9 |
> On the other hand it seems to give a lot of users trouble with |
| 10 |
> blockers, general configuration and mass-updates on things like |
| 11 |
> removing python:2.5. |
| 12 |
> |
| 13 |
> What are your opinions? Did it improve user experience? What could be |
| 14 |
> improved? |
| 15 |
|
| 16 |
As one of the lead devs on the python team, here are my thoughts. |
| 17 |
|
| 18 |
I think we have made things more "correct". As a developer, it is much |
| 19 |
easier for me to tell when a package has incomplete or simply broken |
| 20 |
python dependencies. |
| 21 |
|
| 22 |
On the user side, I think we have traded occasional/random build |
| 23 |
failures due to mismatched python versions for some barely |
| 24 |
comprehensible portage dependency conflict messages. This is certainly |
| 25 |
not ideal, but I think it is always better to have portage fail during |
| 26 |
dependency resolution than at build time. |
| 27 |
|
| 28 |
The (non-)relationship between eselect python and PYTHON_TARGETS is |
| 29 |
something that would be nice to resolve, but I don't know how to do |
| 30 |
it. PYTHON_SINGLE_TARGET will probably cause problems if/when packages |
| 31 |
start supporting python3 only. |
Archives