1 |
Piotr Pruszczak posted <4405835C.5080209@××××××××.pl>, excerpted below, |
2 |
on Wed, 01 Mar 2006 12:19:56 +0100: |
3 |
|
4 |
> I want to start the new gentoo installation on the workstation with |
5 |
> |
6 |
> ACCEPT_KEYWORDS=~amd64 |
7 |
> |
8 |
> at the make.conf |
9 |
> |
10 |
> my question is : did anybody tried this? |
11 |
> how stable is it for workstation? && should I expect any troubles ? |
12 |
|
13 |
It's generally stable, with a few occasional hickups. |
14 |
|
15 |
Keep in mind that ~arch normally indicates a package considered basically |
16 |
stable upstream -- enough for Gentoo testing with an eye toward Gentoo |
17 |
stability at a later time. You will see occasional -rcs and the like |
18 |
going ~arch, as a test for the final release going ~arch and then |
19 |
arch-stable. Still, even that doesn't happen that much, the -rcs remain |
20 |
masked by default Gentoo policy, and a dev should be pretty confident in |
21 |
the stability of the rc -- that it's stable enough that it itself could |
22 |
be called the release and eventually stabilized -- before it would get |
23 |
unmasked even to testing. Otherwise, the ebuilds are in the tree for devs |
24 |
and advanced users brave enough to try them to unmask them and do so. |
25 |
|
26 |
There are exceptions. A ~arch bash release a few months ago caused some |
27 |
havoc with booting as the (bash based) initscripts no longer behaved as |
28 |
expected, for example. If you are unprepared to cope with that by simply |
29 |
booting your backup installation on another partition, or booting the |
30 |
LiveCD or Knoppix or whatever, then chrooting if necessary and |
31 |
troubleshooting and fixing the problem, you shouldn't be running unstable. |
32 |
However, many would point out that every sysadmin should have at least one |
33 |
level of emergency boot and restoration backup available, regardless of |
34 |
whether they are running stable or not, and those "many" would be |
35 |
absolutely correct! It's just that statistically, you have a slightly |
36 |
higher chance of having to actually /rely/ on that backup, perhaps once or |
37 |
twice a year, compared to once every one to two years (and then only if |
38 |
you happened to sync at just the wrong moment before it was caught), |
39 |
maybe, on stable. Likewise with backups of your important documents. In |
40 |
theory, something /could/ go haywire enough to do an rm -rf .* in a less |
41 |
than well tested ebuild or something. I've never had that happen and |
42 |
doubt it ever will, but there's a non-zero chance of it, and a responsible |
43 |
sysadmin will have backup measures in place sufficient to minimize any |
44 |
damage. Again, /every/ sysadmin should be this responsible, but there's a |
45 |
slightly higher risk of having to actually rely on that backup, if you are |
46 |
running ~arch. Of course, the risk is STILL far higher that you'll simply |
47 |
fat-finger something yourself, than that the system will screw up that |
48 |
badly, which is itself another reason to make sure those backups are |
49 |
maintained AND tested. A good sysadmin knows well enough how easy it can |
50 |
be to be one's own worst enemy, and allows for it. That applies to |
51 |
firewalls and to backups equally, as well as to not taking the name of |
52 |
root in vain. |
53 |
|
54 |
Back to ~arch packages. Gentoo's own portage product, as well as |
55 |
baselayout, follow somewhat different rules than normal packages, since |
56 |
Gentoo /is/ the upstream. Thus, both of them routinely have -rc's or |
57 |
-pre's out for ~arch users. Again, it's no big deal for a sane |
58 |
administrator, that knows what he's doing, and has a sane backup system in |
59 |
place if something happens. Early in the current -rc cycle for baselayout, |
60 |
there were a couple kinks in the initscripts, for instance, but they |
61 |
continued to work well enough to get to a console prompt, which is in |
62 |
almost all cases enough, for a decent sysadmin. Most of the time, they |
63 |
won't even have to resort to the emergency boot and maintenance partition, |
64 |
as they'll have the tools they need at the command prompt to troubleshoot |
65 |
and get a fully working system once again. |
66 |
|
67 |
Here's a very good place to mention my favorite often-overlooked portage |
68 |
feature once again! =8^) FEATURES=buildpkg can be quite useful to |
69 |
someone running stable, but it can be **EXTREMELY** useful to someone |
70 |
running ~arch. If you've been using buildmpk, and you find the latest |
71 |
~arch version of whatever broken on your system, it's no sweat at all to |
72 |
revert to your previous version, since you have it in binary form all |
73 |
ready to emerge -K, no recompiling necessary. That means fixing gcc |
74 |
breakage is easy as well, since the binary package won't require gcc to |
75 |
recompile. Even fixing portage is easy, even if emerge -K won't work, |
76 |
because the binary packages are simply tar.bz2 tarballs, with a bit of |
77 |
additional portage metadata glued to the end. Thus, after taking care to |
78 |
backup any config files that come with the package (make.conf being the |
79 |
big one in this case), simply untar the latest working portage package to |
80 |
/, and it'll overwrite the broken one, giving you a working portage once |
81 |
again. Of course, if bzip2 or tar break, that won't work, but you can |
82 |
probably simply reemerge them then, or simply copy over tar and/or bzip2 |
83 |
from your emergency restore solution to the working system, and get back |
84 |
to work. |
85 |
|
86 |
Running ~arch, you ARE expected to run into a bit of minor breakage from |
87 |
time to time, however. Really, a big reason for ~arch in the first place |
88 |
is to find these minor breakages and fix them, but that only happens if |
89 |
the testers experiencing the breakage report it by filing bugs. Be a |
90 |
responsible bug reporter. If you don't know how already, learn how to |
91 |
file good bugs -- including doing reasonable searches for duplicates |
92 |
before you file your report -- it doesn't help get the bug fixed if the |
93 |
dev is spending the time they WOULD be working on fixing the bug, reading |
94 |
and marking all the duplicates, instead. |
95 |
|
96 |
That said, the usual breakage is just that, quite minor, and often won't |
97 |
affect you much at all beyond having to tweak a config a bit because the |
98 |
automated handling that would work before the package went stable, broke |
99 |
for some reason. There are usually plenty of workarounds, including the |
100 |
one of simply putting that particular version in your package.mask file |
101 |
and downgrading back to what you had before, until it's fixed. Again, |
102 |
that's actually minor enough to almost be boring, especially if you have |
103 |
buildpkg setup as described above. |
104 |
|
105 |
-- |
106 |
Duncan - List replies preferred. No HTML msgs. |
107 |
"Every nonfree program has a lord, a master -- |
108 |
and if you use the program, he is your master." Richard Stallman in |
109 |
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html |
110 |
|
111 |
|
112 |
-- |
113 |
gentoo-amd64@g.o mailing list |