| 1 |
On Wed, Jul 31, 2013 at 10:26 AM, Tanstaafl <tanstaafl@×××××××××××.org> wrote: |
| 2 |
> On 2013-07-31 11:20 AM, Canek Peláez Valdés <caneko@×××××.com> wrote: |
| 3 |
>> |
| 4 |
>> If you don't use the systemd USE flag (and never install anything that |
| 5 |
>> depends on systemd), you will not get systemd installed, but many |
| 6 |
>> packages will install systemd unit files in /urs/lib/systemd/system. |
| 7 |
>> This unit files are little non-executable files which do nothing in |
| 8 |
>> your system, but some people feel really strongly about having |
| 9 |
>> anything in their machines with *systemd* in its path. If you want to |
| 10 |
>> exorcise those unit files, add /usr/lib/systemd/system to |
| 11 |
>> INSTALL_MASK. |
| 12 |
> |
| 13 |
> |
| 14 |
> Ok, thanks Canek... but my last question remains... if this really is going |
| 15 |
> to be the only and one true way to opt out of systemd, shouldn't this be |
| 16 |
> well documented in the man page, as opposed to just generic references to |
| 17 |
> masking 'files'...? |
| 18 |
|
| 19 |
No, because the *exact same* situation occurs for Bash completion |
| 20 |
scripts... and logrotate scripts... and cron jobs... and... |
| 21 |
|
| 22 |
The devs decided (and I agree with them) that the important thing is |
| 23 |
to cover the necessities of the majority of users and to have |
| 24 |
reasonable default settings. Therefore, having USE flags for |
| 25 |
bash_complete, and logrotate, and crond, and systemd, and OpenRC, and |
| 26 |
whatever else you want to throw in the mix is overkill and a |
| 27 |
maintenance nightmare. Not to mention that they will require a full |
| 28 |
rebuild every time you changed one of those flags. And the packages |
| 29 |
(in general) will not care about those tiny files; they will work fine |
| 30 |
with all of them installed, no matter if you don't use Bash |
| 31 |
completion, nor logrotate, nor crond, nor systemd nor OpenRC. |
| 32 |
|
| 33 |
So, those files are installed unconditionally. And that's the smart |
| 34 |
thing to do, since most users will not even care about any of them. |
| 35 |
|
| 36 |
There is no need to document nothing special about any of them |
| 37 |
(bash_complete, logrotate, crond, systemd, OpenRC, etc.), since that |
| 38 |
option is for really special cases (think embedded devices with really |
| 39 |
small disk space), or for really picky users (like myself some weeks |
| 40 |
ago, before I reached the conclusion that masking files in /etc/init.d |
| 41 |
is not worth it). |
| 42 |
|
| 43 |
>> It's the exact same situation with OpenRC: those of us who install |
| 44 |
>> systemd don't want nor need the files in /etc/init.d, but they get |
| 45 |
>> installed anyway. If we want to exorcise OpenRC init scripts from our |
| 46 |
>> systems, we need to add /etc/init.d to INSTALL_MASK. |
| 47 |
> |
| 48 |
> |
| 49 |
> And so *both* should be fully documented in the man page... |
| 50 |
|
| 51 |
No, see above. |
| 52 |
|
| 53 |
>> For the record, I now think it's a waste of time trying to stop the |
| 54 |
>> installation of tiny files that basically do nothing, either in |
| 55 |
>> /usr/lib/systemd/system or in /etc/init.d, but you have the option if |
| 56 |
>> you so desire. |
| 57 |
> |
| 58 |
> |
| 59 |
> Ok, and thanks again... |
| 60 |
|
| 61 |
Regards. |
| 62 |
-- |
| 63 |
Canek Peláez Valdés |
| 64 |
Posgrado en Ciencia e Ingeniería de la Computación |
| 65 |
Universidad Nacional Autónoma de México |
Archives