1 |
On Thursday 10 April 2003 21:50, Matt Thrailkill wrote: |
2 |
> Biggest problem I see is putting them up for download. Making them |
3 |
shouldn't be that hard, just require ebuild maintainers to submit a binary |
4 |
pkg built with emerge -b when they submit their .ebuilds. |
5 |
|
6 |
Easy for you to say! I and other gentoo devs use 56k dialup, which tyipcally |
7 |
sends data at 33.6kbps. Ever tried uploading big stuff at that rate (which |
8 |
also blocks your other internet activity)? |
9 |
|
10 |
Besides building against GRP would mean a) maintaining a grp chroot |
11 |
(reasonable) and b) building every package twice - once for myself, once for |
12 |
grp. (No ccache because my cflags are different from grp's). So twice the |
13 |
build time. Ugh. |
14 |
|
15 |
The solution for these is a compile farm into which devs can log remotely and |
16 |
build packages. Or it could just build them automatically. But all this would |
17 |
be is an extended GRP set - I don't see the need for that. GRP for 1.4 is |
18 |
already slated to include all of kde and gnome, plus probably some of the |
19 |
other heaviest compiles. Other than that, 99% of systems only need a smallish |
20 |
amount of smallish apps compiled, you can finish that quite quickly even on a |
21 |
slow machine - quicker probably than it'll take you to donwload the |
22 |
precompiled GRP. |
23 |
|
24 |
(BTW with a fast machine and a 56k dialup, many packages probably take longer |
25 |
to download precompiled than to compile locally (that's if you discount time |
26 |
taken to download the sources though :-) ). |
27 |
|
28 |
Anyway if this is just an exnteded set of packages for GRP, why is everyone |
29 |
talking about it as if it were a radical new feature? |
30 |
|
31 |
Making grp include all or nearly all the source tree shouldn't be that big a |
32 |
problem. After all we already include most of the longest compiles (and have |
33 |
openoffice-bin), so we probably already have at least a third of the tree as |
34 |
far as compile time goes - and we can generate it for all archs in jsut a few |
35 |
days according to the release policy schedule. So if we wanted to make a |
36 |
complete GRP we probably could. |
37 |
|
38 |
Making it always include all new (stable) versions of packages (for all |
39 |
archs!) would need a lot of compile power though, or rather it would need a |
40 |
dedicated compile server (or piece of a cluster). Well, if you the proponents |
41 |
of this idea have such a server handy, you can easily enough set it up to do |
42 |
this and tell gentoo users they can get packages from there. Adding a feature |
43 |
to emerge that'll make it fetch binary packages from a specified server |
44 |
should be easy enough. (Although we need to improve the binary package format |
45 |
to make it specify metadata like CFLAGS, arch... I don't know if that's been |
46 |
done yet.) |
47 |
|
48 |
Well all that you can do you have such a server with a good uplink to spare. |
49 |
However we of the project (me personally too) don't think that would be good |
50 |
use of such a server, because we think the existing GRP is enough (it can be |
51 |
expanded in time, but not so radically as you propose, at least not right |
52 |
away. And it's not top priority with us.) So we would like to respectfully |
53 |
ask you, if you have such a server, to please donate it to the project :-) |
54 |
|
55 |
(Or at least hear our counter proposal of what we'd do with it. There are/were |
56 |
plans of a build farm for devs to log on - for testing though more than for |
57 |
building binary packages iirc - I don't know what the current status is on |
58 |
that.) |
59 |
|
60 |
-- |
61 |
Dan Armak |
62 |
Gentoo Linux developer (KDE) |
63 |
Matan, Israel |
64 |
Public GPG key: http://cvs.gentoo.org/~danarmak/danarmak-gpg-public.key |