fix journalctl --since

[yalh76]

The problem

Difficult to explain, as it's more sensations and difficult to reproduce...

Sometime you try to install a package and even if the service is started and the line_match can be found in journalctl, ynh_systemd_action doesn't detect it.

Same during package_check with services fail to starts but start during manual installation

Solution

Diving in ynh_systemd_action and in journalctl arguments. It seems that --since=-0 is documented nowhere

PR Status

Ready

How to test

Make some install and check what happens :/

Validation

• Principle agreement 0/2 :
• Quick review 0/1 :
• Simple test 0/1 :
• Deep review 0/1 :

[Josue-T]

Hello,

I also detected sometime that this helper is sometime a little bit buggy. Mainly with gitea so without systemd.

Did you see that your patch fix your issue ?

[maniackcrudelis]

--since=-0 is since minus 0 seconds.
Just tried the command, it works the exact same way with -0 and now

Are you sure that's the issue you had ?

[yalh76]

I'm like 95% of the issue I had but no proof, no logs, no screenshot, nothing....

I can just tell the conditions...

I was working on YunoHost-Apps/distbin_ynh#52 => applying last example_ynh on distbin_ynh

As I worked like a barbarian, change_url script wasn't working.... And as I didn't took time to read all logs, I decided to test manually change_url on distbin_ynh

I applied a snapshot and started a virtual machine, launch three ssh windows with a command for each one:

1. $ yunohost app install distbin --debug
2. $ journalctl -u distbin -fe
3. $ journalctl -u nginx-fe

like that:

At the end of the install, ynh_systemd_action was waiting distbin service start... but in the $ journalctl -u distbin -fe windows I seen that the disbin service was really started. So I launch a fourth ssh to execute $ service distbin restart and the package installation ended well.

I have seen that a few times on different packages but wasn't able to reproduce systematically...

I will try to make additional tests. and let you know.

[maniackcrudelis]

The thing is, using --since=-0 as well as --since="now" will just start to follow the log with no lines before.

The behavior is the same, as said before, between now and 0s before now.
So I really think the problem is elsewhere.

You should probably try to follow $templog as well in a separate terminal to see if the helper really miss something.

Also, instead of a sleep 60, you can use the timeout option of ynh_systemd_action.

[yalh76]

I wasn't able to reproduce systematically. So closing for now