1 |
On 08/05/2016 01:11 PM, Chí-Thanh Christopher Nguyễn wrote: |
2 |
> |
3 |
> I disagree. ACCEPT_LICENSE does not change the files which the ebuild |
4 |
> installs on the user's filesystem, nor does it interfere with subslot |
5 |
> dependency calculations like EAPI changes do. Therefore a new revision |
6 |
> is not needed. |
7 |
|
8 |
Those two are sufficient for a new revision, but not necessary. Use |
9 |
common sense. If not doing a revision is going to negatively affect |
10 |
people, then you should do a revision. |
11 |
|
12 |
|
13 |
>> If you installed something whose EULA says it can hijack your webcam and |
14 |
>> post naked pictures of you to slashdot, but it incorrectly had |
15 |
>> LICENSE="GPL-2", wouldn't you want to find out that I corrected it? |
16 |
> |
17 |
> That is the package manager's task. I do get reminded if a package which |
18 |
> I have installed is now masked by package.mask, maybe something like |
19 |
> this would work for licenses too. |
20 |
> |
21 |
|
22 |
The package manager's task is well-defined, and that isn't part of it. |
23 |
Saying "maybe it's possible somehow some day" is not a good reason to |
24 |
screw up ACCEPT_LICENSE handling with certainty right now. |
25 |
|
26 |
To bring this back on topic. If I change LICENSE, I'm going to do it in |
27 |
a new revision, because I want it to work right -- to each his own. At |
28 |
the moment, doing a revision on top of a stable ebuild is not optimal, |
29 |
because little bug fixes like that can remain in unstable for a year. A |
30 |
carefully-worded exception could alleviate that, when the change could |
31 |
in no way affect architecture stability. |