Created attachment 161151 [details]
Patch that pipes svscan into logger(1)
I like this. From discussion in email we need to test and ensure nothing bad happens if syslogd stops.
From the mails I sent to Bryan, here's a possible caveat I identified with my current solution:
I’m not completely sure what happens to the stdout pipe if syslogd crashes/is just not running. Does logger(1) buffer input until it can send it to syslogd or does it simply give up if the connection breaks? Might be worth finding out, because otherwise it might be safer to wrap it in a “while true” loop on the sh side…
This could also lead to a relatively ugly feedback-loop if svscan is supposed to supervise syslogd.
I guess my solution works for me. Also, syslogd has never crashed for me so far.
Is this still relevant?
No maintainer feedback!