Re: [gentoo-server] portage suggestions

[Jason Stubbs]

Tomek Lutelmowski wrote:
> I think these two features would be very nice for server enviroments:
>
> 1. For stability reasons, relaying only on "x86" flag is often not
> sufficient. There are many cases when someone masked package as stable, then,
> few hours later (after users feedback) its reverted to "~x86". After syncinc
> portage tree, I never know which packages marked as stable are truly stable
> and well tested by community. My idea is to include new flag for emerge. For
> examle:
> # emerge -pu --stablesince=48h world
> Would list all packages for upgrade, which has not been changed since 48
> hours, so there is low possiblity that this list includes untested packages.

It won't achieve the desired affect. People using stable will just start 
doing the same thing
and you'd end up with packages being remasked after 2 days instead of 2 
hours. Packages
going arch and then being sent back to ~arch is not something that 
should happen often,
if at all. If you are finding that it happens regularly, you might want 
to open a bug for
qa@g.o and list up the offenders.

> 2. In all my server instalations, I like to keep portage tree as small as
> possible - for two reasons: syncinc speed and disk space ussage. Now I can
> only use RSYNC_EXCLUDEFROM flag in make.conf to exclude portage branches from
> syncinc. Much more convinient and efficent would be RSYNC_INLCUDEFROM flag,
> so I could define which branches of tree I want to sync. The portage tree
> will be much smaller, and I wouldnt have to remove new branches that I dont
> need to sync. Of course in longer term such flag would help to lower bandwich
> usage of rsync servers.

According to rsync's man page, you can do this using RSYNC_EXCLUDEFROM 
already.
Something like the following should work:

+ /profiles
+ /eclass
+ /sys-*
+ /dev-*
+ /metadata/glsa
- *

--
Jason Stubbs
-- 
gentoo-server@g.o mailing list