1 |
2008/1/30, Duncan <1i5t5.duncan@×××.net>: |
2 |
> |
3 |
> Volker Armin Hemmann <volker.armin.hemmann@××××××××××××.de> posted |
4 |
> 200801300220.21430.volker.armin.hemmann@××××××××××××.de, excerpted below, |
5 |
> on Wed, 30 Jan 2008 02:20:21 +0100: |
6 |
> |
7 |
> >> also adding --as-needed as LDFLAGS should help you save some time in |
8 |
> >> recompiling stuff.... |
9 |
> > |
10 |
> > yeah - no. Don't do it. It breaks stuff. |
11 |
> |
12 |
> I think the breakage in most of the common stuff Gentoo devs anyway use |
13 |
> has been fixed by now. I know I've had surprisingly few problems (read, |
14 |
> ZERO problems) with it here. Surprising, as I expected at least a few, |
15 |
> but I've seen exactly ZERO. |
16 |
> |
17 |
> That said, especially for those who just want things to work, without |
18 |
> having to futz with LDFLAGS and remerge something occasionally, I'd still |
19 |
> not recommend it. For those that enjoy the challenge of such things, |
20 |
> however, I'd say great! Go for it! And for those in the middle, well, |
21 |
> YMMV, as the saying goes. You probably lean one way or the other, so |
22 |
> take your pick. |
23 |
|
24 |
|
25 |
for me it has worked as a charm and as i've said: the ebuilds that are |
26 |
broken by it usually have ldflags removed so there's no need to worry |
27 |
anymore for it. i found it something awesome after the libexpat-2 update. |
28 |
without it i would have to emerge abour 300 packages (kde stuff, gcc, |
29 |
mime-shared, firefox ecc) but with it i only needed something like 10-15 |
30 |
packages of which 7 were updates and 2 or 3 had use flags changed. the doc |
31 |
about that flag is about 2 and a half years old and wasn't updated since |
32 |
then. i bet that every single dev would advise to us the flag if you happen |
33 |
to do have some packages that need to recompile broken stuff (x264-svn, |
34 |
ffmpeg and xine-lib would push some package recompile). |
35 |
i'd also advise to use the the --buildpkg feature for some high compiling |
36 |
time packages like kdebase, gcc, openoffice or mplayer. since these packages |
37 |
are likely to be the same for quite some time and they might be pushed into |
38 |
recompilation by some other updates having them packaged on disk (they'd |
39 |
require disk space when packaged) would mean to skip all the compilation |
40 |
part and jump directly to the install one (of course if they would need to |
41 |
be recompiled portage would do it). |
42 |
|
43 |
As for amd64 vs. ~amd64, I'm 100% ~amd64 here, and have been from when I |
44 |
> started on Gentoo. In fact, I've read suggestions that Gentoo tends to |
45 |
> work better at ~arch than at stable, because ~ is where most developers |
46 |
> are, and it's not uncommon for certain incompatibilities with "old" |
47 |
> software, that is, the crufty stable stuff from months or years ago |
48 |
> that's common in stable, to be overlooked until some poor stable keyword |
49 |
> user files a bug. Yes, before stabilizing, the arch-devs and arch- |
50 |
> testers normally test a package against a full-stable system, but it's |
51 |
> simply not possible to test against every permutation of USE flags and |
52 |
> mix of merged apps. While it's certainly true that ~arch packages have |
53 |
> the same issue, at least there there's a decently active community of |
54 |
> testers actively reporting bugs and devs fixing them. |
55 |
|
56 |
|
57 |
well, i'm more pleased by live ebuilds. i personally use a good ammount of |
58 |
live packages, especially multimedia, xorg and some network stuff like |
59 |
networkmanager and knm. |
60 |
|
61 |
Were it conveniently possible, I'd say the most trouble-free scenario |
62 |
> would be to take only ~arch packages that had been ~arch for say a week, |
63 |
> minimum, after verifying that nobody had run into and filed any serious |
64 |
> bugs on them. That'd be after the initial test wave had done its |
65 |
> installation and testing, but before the cruft that often attaches to |
66 |
> stable had set in. |
67 |
> |
68 |
> <brainstorming> What would be great would be a keyword system that would |
69 |
> allow just this, say ~ for initial testing, automatically upgraded to / |
70 |
> after the week UNLESS they've been marked ~~, with the extra ~ |
71 |
> automatically added to ~ packages by a script if a bug has been filed, |
72 |
> blocking the automatic upgrade to /, and a bugzilla keyword that a dev |
73 |
> could add to put the package back on automated / track if they've decided |
74 |
> the bug isn't worth derailing the automated / upgrade over. Then people |
75 |
> could go full testing ~ mode if they wanted, / mode if they wanted almost |
76 |
> ~ but wanted to be spared the pain of the most obvious bugs as found in |
77 |
> the initial testing wave, and full stable arch if they wanted crufty old |
78 |
> packages, say for a server only upgraded for security issues or the like, |
79 |
> somewhere. </brainstorming> |
80 |
|
81 |
|
82 |
usually that's the point of hard masking. when a package doesn't work then |
83 |
it goes hardmasked. |
84 |
|
85 |
Of course, YMMV, but ~ for the entire system, with appropriate site based |
86 |
> masking as Gentoo already makes possible with /etc/portage/package.mask |
87 |
> and the like, isn't as terrible or system breaking as some folks like to |
88 |
> make it out to be. By policy, ~ is only for stable track packages in the |
89 |
> first place. Obviously broken packages and those not considered stable |
90 |
> candidates normally never get even the ~ keyword, as they are kept in |
91 |
> development overlays or in the tree but without keywords or fully hard |
92 |
> masked, so ~ packages aren't the broken things a lot of people make them |
93 |
> out to be. |
94 |
|
95 |
|
96 |
i still think that having the base system on the unstable arch isn't a very |
97 |
good idea. i've used the same stuff for quite some time, but i found out |
98 |
that it isn't a very good stuff. or at least 6 months ago wasn't a good |
99 |
idea. |
100 |
|
101 |
-- |
102 |
> Duncan - List replies preferred. No HTML msgs. |
103 |
> "Every nonfree program has a lord, a master -- |
104 |
> and if you use the program, he is your master." Richard Stallman |
105 |
> |
106 |
> -- |
107 |
> gentoo-amd64@l.g.o mailing list |
108 |
> |
109 |
> |
110 |
|
111 |
|
112 |
-- |
113 |
dott. ing. beso |