1 |
>>>>> On Sat, 20 Nov 2010, Donnie Berkholz wrote: |
2 |
|
3 |
> This is kind of a strange time to run these, since then you'll end |
4 |
> up with a new user and group even if the build fails or is killed. |
5 |
> You'll also end up with extraneous users and groups on build |
6 |
> servers. Does this user/group need to be present on the build system |
7 |
> for some reason? I'd much prefer to see it in pkg_{pre,post}inst(), |
8 |
> despite pkg_setup() being much more common in the tree (from a dumb |
9 |
> grep/sort/uniq): |
10 |
|
11 |
> 74 pkg_preinst() { |
12 |
> 101 pkg_postinst() { |
13 |
> 369 pkg_setup() { |
14 |
|
15 |
Don't forget the various eclasses creating users and groups, which |
16 |
probably account for even more ebuilds: |
17 |
|
18 |
apache-2.eclass |
19 |
games.eclass |
20 |
mysql.eclass |
21 |
qmail.eclass |
22 |
vmware-mod.eclass |
23 |
|
24 |
All of them call enew{user,group} from pkg_setup. (With the exception |
25 |
of mysql.eclass, which for some reason calls them from both pkg_setup |
26 |
and pkg_preinst.) |
27 |
|
28 |
Ulrich |