config loading errors should report loudly instead of failing silently with exit code 2

[klaernie]

I spent yesterday trying to figure out why lmd was not starting up. It took quite some hours to realize, that the config format is not ini, but toml - and toml requires quoting strings always. (my inner monk now needs to rename the config to lmd-local.toml)

The confusing part is, that I tried to run the binary (./lmd -c repro.ini) and it did not print any error message, just silently exited. My next step was checking strace, and I could see that it exited just after reading the config file. Without any error message to guide I the scope was way too big and I started looking through the code to understand where it could have exit(2), but didn't find anything. I read over the config reading part a couple times, and only hours later it popped into my head, that there was no true ini-reader, but a toml-reader. So I read the wikipedia page and it mentioned the need to quote every string.

Here is a small reproduction:

~/repro % ll
total 11376
-rwxr-xr-x 1 kandre kandre 11567104 Mar 17 09:03 lmd
-rw-r--r-- 1 kandre kandre      105 Mar 17 09:03 repro.ini
-rw-r--r-- 1 kandre kandre      107 Mar 17 09:03 repro.toml
~/repro % tail repro*
==> repro.ini <==
Listen = [ '127.0.0.1:9999' ]
[[Connections]]
name = localhost
Source = [ '/var/run/nagios/livestatus' ]

==> repro.toml <==
Listen = [ '127.0.0.1:9999' ]
[[Connections]]
name = 'localhost'
Source = [ '/var/run/nagios/livestatus' ]
~/repro % ./lmd -c repro.ini; echo $?
2
~/repro % ./lmd -c repro.toml; echo $?
lmd - version 2.1.4 (Build: 15be67e, go1.19.6) started with config [repro.toml]
[2023-03-17 09:04:47.641][Info][pid:17184][main:293] lmd - version 2.1.4 (Build: 15be67e, go1.19.6) started with config [repro.toml]
[2023-03-17 09:04:47.685][Info][pid:17184][listener:136] listening for incoming queries on tcp 127.0.0.1:9999
[2023-03-17 09:04:47.685][Info][pid:17184][peer:300] [localhost] starting connection
[2023-03-17 09:04:47.688][Info][pid:17184][peer:1508] [localhost][r:be8c09] site went offline: dial unix /var/run/nagios/livestatus: connect: no such file or directory
[2023-03-17 09:04:47.688][Warn][pid:17184][peer:356] [localhost] initializing objects failed: dial unix /var/run/nagios/livestatus: connect: no such file or directory
^C
[2023-03-17 09:04:50.527][Info][pid:17184][main:736] got sigint, quitting
[2023-03-17 09:04:50.528][Info][pid:17184][listener:146] stopping tcp listener on 127.0.0.1:9999
[2023-03-17 09:04:50.528][Info][pid:17184][listener:147] tcp listener 127.0.0.1:9999 shutdown complete
[2023-03-17 09:04:50.528][Info][pid:17184][main:246] lmd shutdown complete
1
~/repro %

Now I'd have some small wishes:

• Could lmd please exit with an error message when parsing the config file failed?
• Could lmd please have a command line option to test the config and then exit? (like nagios or naemon do) I'd use this in our build system to prevent builds with obviously wrong configs.

Thanks and best regards, Andre

[klaernie]

@sni Would you consider adding a --test-config to test the config file and exit accordingly?

[sni]

well, usually the lmd.ini is pretty static not not changed all the time, but why not...

[klaernie]

Well, I'm just asking to protect from stupid - with me being the usually stupid one ;)