Am 31.12.2014 um 01:00 schrieb Tomas Carnecky:
I was trying to install postfix into an VM image which used an auto-generated hostname. It happened that the hostname was fully numeric (7593408), and the post-install script failed to execute properly. Here's an excerpt from the install log: installing postfix... /usr/bin/postconf: warning: valid_hostname: numeric hostname: 7593408 /usr/bin/postconf: fatal: unable to use my own hostname /usr/bin/postconf: warning: valid_hostname: numeric hostname: 7593408 /usr/bin/postconf: fatal: unable to use my own hostname [repeats about 10 times] /usr/lib/postfix/post-install: Error: "" should be an absolute path name. When later trying to run that VM, postfix failed to start up because the installation failed to correctly set the owner on the /var/spool/postfix folder. Running set-permissions fixed that. Using a non-numeric hostname during installation, too. Is there any reason why the post-install script needs a valid hostname?
not really *but* take not care about sane hostnames used for SMTP HELO and sane PTR-records brings a ton of other troubles in case of a MTA if you try to send mail to the outside world
smtp_helo_name = $myhostname invalid helo_name = reject mail here
