Salvatore, Am 06.01.20 um 22:46 schrieb Salvatore Bonaccorso: > I think the whole problem is placed at multiple levels. You seem to > use systemd as init, but when lgorotate comes into action, the > postrotate is
Yes. I use systemd. I updated my unit file to `Type=simple`, because I wanted tinyproxy to log to stdout (for the logs to be picked up by systemd-journald). While doing so I removed the PidFile setting I no longer needed. > invoke-rc.d --quiet tinyproxy reload > /dev/null > > As tinyproxy.service can't reload the service through the .service > file, it will fallback into reloading tinyproxy via the init-script: > > […] > > PIDDIR will be ".", and the chown and chgrp will be applied to the > current working directory. > > The current working directory is changed to /, and thus the chown and > chgrp call upon '.' will affect /. This matches my observations. >> Thus I feel the severity of `critical` is justified for this bug report. > > I'm not sure this is warranted, but I will leave this to the decision > of the maintainer :) At least the description within reportbug applied to my case. It was a nasty surprise, because it didn't happen immediately after the configuration change, but with a bit of delay. Best regards Tim Düsterhus
