treewide: start logger at priority 12

[pprindeville]

Maintainer: @blogic, @MikePetullo, @dubek, @psycho-nico
Compile tested: x86_64, generic, LEDE head (6be7010)
Run tested: same

Built and installed (via scp/opkg install) all three packages and tested them individually, rebooting each time.

Description:

Set START priority to 12 (instead of 20, same as network).

[champtar]

Hi @pprindeville
if network is at 12, maybe set logger at 13?

[pprindeville]

if network is at 12, maybe set logger at 13?

It's not, it's at 20. And that's the point. We want logging run before the network comes up. If the network fails to come up, we need some place to capture information (logs) about why.

[champtar]

I misunderstood your comment, then why not less than 12? (add a small explanation in the commit, we want to be as early as possible but after X)

[dangowrt]

imho neither 12 nor 20 make sense here. 41 would be appropriate, as that's the point where we are supposed to have mounted everything syslog-ng might be writing to

[pprindeville]

/var is a link to /tmp, so that gets mounted early on.

We could always add triggers so that if the network configuration changes or the filesystem changes that it gets reloaded...

[dangowrt]

same as for syslog-ng

[dangowrt]

ulogd may require to be started even much later as it may require a database server to be up and running... anyhow, it's really a different type of beast than rsyslog or syslog-ng.

[pprindeville]

Okay, dropping ulogd from changeset.

[pprindeville]

I'm made the requested changes. Can we please proceed?

[pprindeville]

(add a small explanation in the commit, we want to be as early as possible but after X)

Done.

[pprindeville]

Can we proceed without @dangowrt's approval as he seems to be MIA?

[pprindeville]

Abandoning... 12 is too early and indeed the timezone hasn't been configured by then (because S10system might not have completed yet).