1 |
> On 8/3/19 7:49 PM, Michael Orlitzky wrote: |
2 |
>> |
3 |
>> That makes me think that we should set |
4 |
>> |
5 |
>> ACCT_USER_HOME=/var/lib/amavis |
6 |
> |
7 |
> We'll do this during the next version/revision bump, keeping everything |
8 |
> else the same. |
9 |
> |
10 |
|
11 |
The recent homedir problems changed my thoughts on this. The |
12 |
mail-filter/amavisd-new package depends on acct-user/amavis because it |
13 |
needs the user to exist when we call e.g. fowners. But, within the |
14 |
amavisd-new ebuild, we have no way to depend on the value of |
15 |
ACCT_USER_HOME inside the acct-user ebuild. If amavisd-new needs |
16 |
/var/lib/amavis to exist and be owned by "amavis" at runtime because a |
17 |
configuration file points there, then the amavisd-new ebuild should take |
18 |
care of that, and not rely on it happening transitively. |
19 |
|
20 |
This is especially true for users shared by more than one package. It |
21 |
also makes the guidelines easier: if your user doesn't log in, leave |
22 |
ACCT_USER_HOME unset. I have a feeling that would be cleaner in almost |
23 |
all cases, and it would be easy to put into repoman as a warning so that |
24 |
we don't have to manually check it a thousand times. |