1 |
On Sunday 14 September 2008 22:37:21 Volker Armin Hemmann wrote: |
2 |
> > > Making the 'official' overlay paludis-only was a BIG mistake. |
3 |
> > |
4 |
> > It's not "paludis-only", it will work with any package manager whose |
5 |
> > developers care enough to support the useful features of the kdebuild-1 |
6 |
> > EAPI. |
7 |
> |
8 |
> and that was an official api? Yes? No? |
9 |
|
10 |
Official according to whom? It is a stable and well documented eapi that any |
11 |
package manager that regards those features to be useful can implement. |
12 |
|
13 |
> Was it needed? No - proven by kdesvn-portage |
14 |
|
15 |
And noone ever claimed otherwise. |
16 |
|
17 |
[...] |
18 |
> > The KDE overlay isn't produced by the "paludis-group". |
19 |
> |
20 |
> no, it was just made by vivid paludis fans. Hey, lets make an overlay a lot |
21 |
> of user want - and make it paludis only. That way we can push paludis! |
22 |
|
23 |
Wow. What do you base this nonsense on? |
24 |
|
25 |
> > No, to provide ebuilds that make use of useful features that benefit both |
26 |
> > users and developers. |
27 |
> |
28 |
> which one? Which features 'benefit both users and developers' and are sooo |
29 |
> important that the kde overlay had to be paludis only? |
30 |
> |
31 |
> Name them please. |
32 |
|
33 |
- '-scm' support (--dl-reinstall-scm for users) |
34 |
- use dependencies (no surprising interruptions mid-merge) |
35 |
- suggestions (you see them upfront rather than in elog messages afterwards) |
36 |
- sets |
37 |
|
38 |
The latter of these is not subject to eapi but incredibly useful when dealing |
39 |
with huge amounts of packages. It obsoletes meta packages and makes |
40 |
reinstalls or uninstalls of all packages in the sets trivial. For Paludis it |
41 |
also means that you can unmask/keyword two hundred packages just by addding |
42 |
the sets to your packages.{keywords,unmask} equivalents. Both Paludis and |
43 |
Portage 2.2 now has sets support although the details of their |
44 |
implementations vary greatly. |
45 |
|
46 |
> Oh, does paludis support and equivalent to 'keep-going' or |
47 |
> '--ignore-failures' or are people who wants this extremly usefull features |
48 |
> still attacked and insulted? |
49 |
|
50 |
Paludis had --continue-on-failure long before --keep-going was implemented in |
51 |
Portage (which is months after the creation of the kdebuild overlay). This is |
52 |
also one of the advantages that users of live KDE ebuilds got by using |
53 |
Paludis (or get if you consider the additional flexibility when compared |
54 |
to --keep-going to be useful). |
55 |
|
56 |
On Monday 15 September 2008 00:38:18 Volker Armin Hemmann wrote: |
57 |
> lets see - an overlay is setup to develop and test ebuilds for KDE4 that |
58 |
> should some day go into the tree. |
59 |
> |
60 |
> Deciding to use a feature that the official pm does not provide - and only |
61 |
> one of the three makes the 'testing' part and 'for the tree' pretty |
62 |
> superfluos. |
63 |
|
64 |
And again I wonder what you base this nonsense on. Perhaps you never ever |
65 |
bothered to look at this overlay, what it provides or what the intentions |
66 |
behind it was? |
67 |
|
68 |
As one of the original decision makers behind it I can tell you, that we never |
69 |
intended to put any of these ebuilds in the tree. For this reason it also |
70 |
never contained any release of KDE. Releases were maintained separately in |
71 |
the tree using eapi 1. It only contains live ebuilds. Which is where '-scm' |
72 |
and sets support provide the biggest advantages. |
73 |
|
74 |
And we didn't do it to harass users. We did it because we wanted to get some |
75 |
real world experience with some of the features that Paludis had provided for |
76 |
years yet there were no indications Portage would support any time soon. Live |
77 |
KDE packages was deemed the place where adding this requirement made the most |
78 |
sense. Managing two hundred packages without those features is pain anyway. |
79 |
|
80 |
We decided that the monthly KDE releases that we were packaging and adding to |
81 |
the main tree using eapi 1 were frequent enough for those who didn't want |
82 |
Paludis for whatever reason. If anybody disagreed with that they could |
83 |
maintain their own overlay (which they did/do). We also announced it over |
84 |
three weeks before we actually made it happen so anybody who cared about the |
85 |
live KDE ebuilds can't really complaim about having been caught by surprise. |
86 |
|
87 |
-- |
88 |
Bo Andresen |