1 |
On Tue, 2020-09-22 at 07:51 -0500, Georg Rudoy wrote: |
2 |
> The PR ( https://github.com/gentoo/gentoo/pull/15938 ) contains |
3 |
> non-live snapshot ebuilds. Although, indeed, it's been last updated a |
4 |
> few months ago. |
5 |
> |
6 |
> What's ultimately blocking there is splitting the commit adding those |
7 |
> ebuilds into multiple commits, one per package, while keeping the tree |
8 |
> consistent. I admit I hadn't had (and most likely won't have in the |
9 |
> foreseeable future) the resources to do these |
10 |
> consistent-per-package-commits — doing this (effectively a toposort, |
11 |
> counting in the dependencies between ebuilds) by hand is painful, and |
12 |
> there's no good automation I'm aware of that'll do that for me. On the |
13 |
> other hand, |
14 |
> 1. I know at least some other multipackage groups don't always do |
15 |
> per-package updates and instead commit everything in one fell swoop |
16 |
> (like qt, for instance). |
17 |
> 2. I think I've been told on #gentoo-proxy-dev that if this indeed |
18 |
> proves to be a burdensome thing, adding all the ebuilds in one commit |
19 |
> is feasible. But it's indeed been a while ago, keeping track of time |
20 |
> is hard. |
21 |
> |
22 |
> If (2) is correct, I'm more than happy to fix whatever other issues |
23 |
> are pointed out in the PR. |
24 |
> |
25 |
> But, of course, all this is ultimately up to you folks. I'm just a |
26 |
> proxy. |
27 |
> |
28 |
> вт, 22 сент. 2020 г. в 07:12, Joonas Niilola <juippis@g.o>: |
29 |
> |
30 |
> > I'd also like to point out something regarding "-9999 packages |
31 |
> > only"; It |
32 |
> > may be buildable one day for users, and broken the next. And some of |
33 |
> > the |
34 |
> > deps may be unbuildable, it's really random and up to state of |
35 |
> > upstream |
36 |
> > instead of state of ::gentoo repo. This was the case with leechcraft |
37 |
> > for |
38 |
> > example, check bug #693328. You should always have a keyworded |
39 |
> > static |
40 |
> > version available. |
41 |
> > |
42 |
> > -- juippis |
43 |
> > |
44 |
> > On 9/22/20 2:23 PM, Michał Górny wrote: |
45 |
> > > # Michał Górny <mgorny@g.o> (2020-09-22) |
46 |
> > > # Poorly maintained suite of NIH packages. Only live ebuilds left |
47 |
> > > # for over a year. This really belongs in an overlay. Some of |
48 |
> > > them |
49 |
> > > # depend on deprecated dev-qt/qtwebkit (#684672). |
50 |
> > > # Removal in 14 days. Bug #693328. |
51 |
> > > app-leechcraft/laretz |
52 |
> |
53 |
> -- |
54 |
> Georg Rudoy |
55 |
> |
56 |
|
57 |
No, the original point really stands. The energy we have invested in |
58 |
EAPI bumping and what not is in no relation to the actual gain. The ROI |
59 |
on leechcraft has been negative, and not just by a small bit. The back |
60 |
and forth has been tiring, and the last-ditch efforts always come in at |
61 |
the last minute. This time it's going for sure. |