1 |
On 11 November 2012 00:43, Michał Górny <mgorny@g.o> wrote: |
2 |
|
3 |
> Hello, |
4 |
> |
5 |
> As some of you are aware already, there is a significant difference |
6 |
> in handling Python implementation choice between python.eclass |
7 |
> and python-r1 / p-d-ng. |
8 |
> |
9 |
> The python.eclass packages use USE_PYTHON (undocumented variable) |
10 |
> or the eselected Python2+3, and store that choice implicitly. |
11 |
> Everything works on the assumption that implementations don't change |
12 |
> or user runs python-updater. |
13 |
> |
14 |
> The python-r1 / p-d-ng packages use explicit USE flags to choose |
15 |
> implementation. The choice is thus explicitly exported and other |
16 |
> packages can use USE dependencies to enforce matching implementations. |
17 |
> |
18 |
> The problems arise whenever python-r1 package depends on package using |
19 |
> the python.eclass, or the other way round. There is no clear way to |
20 |
> enforce the same implementations being used, so everything falls down |
21 |
> to user. |
22 |
> |
23 |
> In order to solve the most common problems due to that, I have |
24 |
> published a news item explaining how PYTHON_TARGETS work and what needs |
25 |
> to be done with them. |
26 |
> |
27 |
> In the discussion with other Gentoo Python devs, it was assumed that |
28 |
> USE_PYTHON should not be considered public and thus not be described |
29 |
> in the message. Sadly, it seems that this resulted in people actually |
30 |
> trying to disable Python3 using PYTHON_TARGETS and seeing random |
31 |
> failures due to python.eclass enforcing it anyway. |
32 |
> |
33 |
> |
34 |
> I believe we should thoroughly think what should be done now in order |
35 |
> to solve the issues best for the transition time. Here are a few |
36 |
> solutions I can think of. |
37 |
> |
38 |
> 1) lard both eclasses with USE_PYTHON <-> PYTHON_TARGETS consistency |
39 |
> checks. This should be pretty easy for python-r1, harder for |
40 |
> python.eclass (anyone volunteering to touch it in thick gloves?). |
41 |
> |
42 |
> In other words, we just warn users whenever they need to do something |
43 |
> to ensure things work. |
44 |
> |
45 |
> |
46 |
> 2) make python.eclass respect PYTHON_TARGETS. |
47 |
> |
48 |
> Sadly, due to PMS limitations we can't tell if PYTHON_TARGETS was set |
49 |
> by user or has the default value. So we either can: |
50 |
> |
51 |
> a) stop using USE_PYTHON and eselect defaults, start using |
52 |
> PYTHON_TARGETS only, pull in correct Python versions, tell users to run |
53 |
> python-updater; |
54 |
> |
55 |
> b) prefer PYTHON_TARGETS over USE_PYTHON and the default (but only |
56 |
> if PYTHON_TARGETS != default PYTHON_TARGETS, so we're on thin ice here). |
57 |
> |
58 |
> |
59 |
> 3) make python.eclass respect and export PYTHON_TARGETS in IUSE. |
60 |
> |
61 |
> Like the above but instead of running the fragile and broken |
62 |
> python-updater, we simply force rebuild of all packages. People can |
63 |
> start using USE-deps in the meantime, everyone becomes a bit happier |
64 |
> in the further plane. Sadly, this means every Python package is going |
65 |
> to need being rebuilt. |
66 |
> |
67 |
> |
68 |
> Please also note that 2 & 3 mean mass-changing behavior of stable |
69 |
> packages. Not something I'd do in a short timeframe where the issue |
70 |
> should be best fixed. |
71 |
> |
72 |
> Any other ideas? Any +1, -1s? |
73 |
> |
74 |
> -- |
75 |
> Best regards, |
76 |
> Michał Górny |
77 |
> |
78 |
|
79 |
I'm sorry if I missed something, but I don't see why USE_PYTHON should be |
80 |
considered not public. That has been the only mechanism until now (in |
81 |
combination with package.mask) to get a consistent system without redundant |
82 |
multiple python versions. |
83 |
|
84 |
I think it is easiest to tell users to set USE_PYTHON as well, until |
85 |
python.eclass is finally phased out. |
86 |
|
87 |
-- |
88 |
Cheers, |
89 |
|
90 |
Ben | yngwin |
91 |
Gentoo developer |
92 |
Gentoo Qt project lead, Gentoo Wiki admin |