1 |
On Tuesday 28 January 2003 11:34, Dylan Carlson wrote: |
2 |
> On Monday 27 January 2003 10:57 pm, Stephan Hermann wrote: |
3 |
> > I don't think this is a good idea. |
4 |
> > On RedHat distribution you got this foolish system (of course binary |
5 |
> > packages). |
6 |
> > You don't know which application is in the normal kde distribution or |
7 |
> > just a plain application like whatever. |
8 |
> > |
9 |
> > If you see kde as one complete desktop enviroment, then you let kde as |
10 |
> > is (compiling all applications in one package the same time), there |
11 |
> > aren't a lot of patches for kde in the last few months, so we can live |
12 |
> > with it as is. |
13 |
> |
14 |
> You're arguing against something I don't think you understand at all. |
15 |
> |
16 |
> If KDE were broken out to smaller ebuilds, there would still be a master |
17 |
> ebuild to bring them all together as one bundle, so therefore users would |
18 |
> not have to remember what pieces make up a complete KDE bundle. |
19 |
|
20 |
And that's the problem. |
21 |
No user knows, what belongs to KDE as enviroment and what does not belong to |
22 |
KDE e.g. kportage. |
23 |
RHs build system works like "taking the complete source, configure one time, |
24 |
and call make <application>/<lib> several times, after that put the binary |
25 |
things into several packages". |
26 |
|
27 |
Now we're speaking about Gentoo. |
28 |
In the actual issue there is a comment from gentoo-dev published: |
29 |
---------------------------><------------------------------------------------------------- |
30 |
Kim Nielsen replied with "The gentoo kernel is quite stable but Gentoo was |
31 |
never ment as a server distribution even though it serves just as well as |
32 |
others like Redhat or Debian. It was intedned for network/developer use. |
33 |
---------------------------><------------------------------------------------------------- |
34 |
|
35 |
So, we're talking about developer business and not "end-customer" business. |
36 |
|
37 |
Yes, it burns cpu power and time, but if you want bleeding edge, so please, |
38 |
leave the configure/make/make install method as is. |
39 |
|
40 |
If someone wants to build an endcustomer distribution with gentoo, so please, |
41 |
use binary tbz packages and ship them directly to the customer. |
42 |
|
43 |
The difference is developer/power-user seeing and end-customer seeing. |
44 |
|
45 |
> The end result would be absolutely no different. People who want the whole |
46 |
> KDE bundle would still get the whole KDE bundle installed the same as it |
47 |
> would be otherwise. |
48 |
> |
49 |
> What it gives us is the ability to update one small piece of KDE without |
50 |
> having to update and recompile the whole thing in the event of a patch. |
51 |
> And between releases there are in fact a ton of patches that go out. |
52 |
|
53 |
And when those "patch-releases" (kde 3.0.1/3.0.2 etc.) will come out, not only |
54 |
one application is patched. |
55 |
All other patches are not directly for the public. |
56 |
And if someone wants those bleeding edge, he can spare space for putting a |
57 |
self compiled (without portage) version of kde in his/her homedir. |
58 |
|
59 |
|
60 |
> The only reason this hasn't been done already is because there isn't any |
61 |
> obvious solution to get around the increase in compile time -- which is a |
62 |
> result of having ./configure run over and over for each separate ebuild. |
63 |
|
64 |
You can't do it normally. If you do it, you have to split up the source |
65 |
packages. |
66 |
|
67 |
Did you ever take look on kdextragear-(1|2) ? |
68 |
Why do you think there is one admin dir for automake/autoconf ? |
69 |
And what happens, when the developer of kapplication8937 decide to release a |
70 |
early alpha software ? he cvs co the admin dir and his own application cvs |
71 |
dir, and put it together. |
72 |
So, if you want to have split ebuilds for kde env apps, do it as I explained. |
73 |
|
74 |
|
75 |
|
76 |
> Once that's eliminated or mitigated, the users wouldn't know any |
77 |
> difference. But I guarantee everyone would know the difference when it |
78 |
> comes time to patch KDE up for inevitable serious bugs or security flaws. |
79 |
|
80 |
Ahhhh....so, you want to have kdenetwork-kmail-3.1-patchlvl-99_2_rc7.ebuild |
81 |
and kdenetwork-knode-patchlvl-98_1_rc5.ebuild ? |
82 |
How would you handle all this patch things? |
83 |
|
84 |
It's easier to do something like this: |
85 |
|
86 |
mkdir bloody-kde-src |
87 |
cd bloody-kde-src |
88 |
cvs -d :pserver:anonymous@×××××××××××.org:/home/kde co qt-copy arts kdelibs |
89 |
kdebase |
90 |
and after this a cvs -d .... diff |
91 |
then cd kdebase/<application-name>/ |
92 |
make |
93 |
make install (if it's configured) |
94 |
|
95 |
easier then to write new ebuilds for kde patch level 666 rc9 |
96 |
|
97 |
> It would save a lot of people a lot of download and recompile time if it's |
98 |
> between major.minor release versions. |
99 |
|
100 |
if you don't have time to download and to recompile, use redhat, suse, |
101 |
mandrake,debian. |
102 |
|
103 |
At least, if you find a system to split up kde, you can find a good solution |
104 |
to split up binutils, when a patch for "ar" is out ,-) |
105 |
|
106 |
My 0.2€ ;) |
107 |
|
108 |
\sh |
109 |
|
110 |
-- |
111 |
gentoo-dev@g.o mailing list |