1 |
On Sat, Sep 24, 2005 at 11:29:56AM +0800, Linux GNUbie wrote: |
2 |
> The beauty of running a binary based GNU/Linux distribution not |
3 |
> particularly on CentOS alone but in general (includes Debian, Red Hat, |
4 |
> SuSE, Mandriva, etc.) is before the updates are released to the public |
5 |
> it has been tested and compiled for use in enterprise production use. |
6 |
|
7 |
You have something quite similar with Gentoo. After all, in what way does |
8 |
testing an ebuild or testing a binary package differ? The ebuild uses the |
9 |
same source for all testers. The only difference is that ebuilds can behave |
10 |
differently based on your USE flags (and a few other variables in your |
11 |
make.conf - but USE is the most notable one). |
12 |
|
13 |
Some people ask me if their mega-size-USE-flag affects performance in a |
14 |
negative way. It doesn't, assuming that you still understand why you set |
15 |
certain USE flags in that variable. Don't forget, the behavior of most |
16 |
ebuilds doesn't change with each USE flag change - only to those the ebuild |
17 |
listens to. |
18 |
|
19 |
For instance, if you check the USE flags for mysql, you'll notice that it |
20 |
only is affected by: |
21 |
berkdb, big-tables, debug, doc, minimal, perl, readline, ssl, static, tcpd |
22 |
|
23 |
How many MySQL users do you think have tested the result of the MySQL ebuild |
24 |
using your USE flags? That'll be quite a lot - and we are not only talking |
25 |
about Gentoo users, but also general MySQL users who manually build the |
26 |
sources (many MySQL production users do this this way) using the configure |
27 |
--with-blabla/--without-blabla tags that are mapped onto the USE flags in |
28 |
Gentoo. |
29 |
|
30 |
Anyway, what I wanted to state was that you can easily test packages |
31 |
yourself. Build the package on your buildserver, deploy that package on your |
32 |
development machine, stage it through the necessary pillars (like testing, |
33 |
staging, simulation) to eventually deploy the new package on your production |
34 |
system. |
35 |
|
36 |
This is a pattern used on many environments, even by those using pure binary |
37 |
packages. The difference with Gentoo is that they are immediately pinned to |
38 |
that binary package while Gentoo allows you to improve the package without |
39 |
much effort, slim down the installation to what you need (effectively |
40 |
decreasing the possibility of a security flaw or software bug affecting you) |
41 |
and even optimize the build to your system environment. |
42 |
|
43 |
Wkr, |
44 |
Sven Vermeulen |
45 |
|
46 |
-- |
47 |
Gentoo Foundation Trustee | http://foundation.gentoo.org |
48 |
Gentoo Documentation Project Lead | http://www.gentoo.org/proj/en/gdp |
49 |
Gentoo Council Member |
50 |
|
51 |
The Gentoo Project <<< http://www.gentoo.org >>> |