| 1 |
On Wednesday 25 June 2003 13:22, Matt Thrailkill wrote: |
| 2 |
> Or heck, if I'm doing a large-scale complex deployment, there doesn't |
| 3 |
> seem to be a way for me to stick with what I know may be good (i.e. |
| 4 |
> 1.4-release) and still get bug fixes and security updates through |
| 5 |
> Portage short of me maintaining my own tree and having my machines pull |
| 6 |
> that down. |
| 7 |
|
| 8 |
If you record the bug fixes and use emerge <packagename> and never emerge -u |
| 9 |
world, things are quite stable, and only required updates are installed. |
| 10 |
|
| 11 |
> When 1.4 is done and 1.5 starts getting all the stuff that is considered |
| 12 |
> unstable now, is there going to be a new profile for 1.5 I guess? Or is |
| 13 |
> it going to be a little sloppier, with both 1.4-stable and 1.4-unstable |
| 14 |
> slipping forward more and more before a new profile gets made? |
| 15 |
> |
| 16 |
Version numbers only really matter for the installation CD's, everyone else is |
| 17 |
just running his/her own variant of the gentoo metadistribution. New profiles |
| 18 |
are introduced only for incompatible changes to the base system. |
| 19 |
|
| 20 |
> Seems like being really anal about profiles, i.e. when the dev team |
| 21 |
> decides to make a new profile for such and such feature set, could |
| 22 |
> accomplish most of this. Even then though, wouldn't the local Portage |
| 23 |
> tree have to contain all the ebuilds in all the profiles? That could |
| 24 |
> start getting fat after a while, since any one machine is only going to |
| 25 |
> use a subset of ebuilds. If strict version control was done with |
| 26 |
> profiles, it might be a good idea then to make rsyncs pull down just the |
| 27 |
> piece of the tree with the ebuilds for the current profile. |
| 28 |
|
| 29 |
A profile is more like a mask dissallowing certain versions of ebuilds. For |
| 30 |
that the number of "inactive" ebuilds is fairly small. This is esp. true for |
| 31 |
the x86 architecture. Many ebuilds don't have keywords for other |
| 32 |
architectures, but that does not mean they do not work, it just means they |
| 33 |
are untested for the architecture. So you'd still want them probably. |
| 34 |
> |
| 35 |
> Yes, I know, long long musing to go with such a small quote. Granular |
| 36 |
> version control makes me feel comfortable about what goes onto my |
| 37 |
> machines. |
| 38 |
> |
| 39 |
|
| 40 |
Do not run emerge -u world. I, personally, never use it except in pretend |
| 41 |
mode. As I'm currently working as a computing scientist, I don't trust |
| 42 |
computers ;-) |
| 43 |
|
| 44 |
Paul |
| 45 |
|
| 46 |
-- |
| 47 |
Paul de Vrieze |
| 48 |
Researcher |
| 49 |
Mail: pauldv@××××××.nl |
| 50 |
Homepage: http://www.devrieze.net |
Archives