Hi Adam! Adam Majer [2005-06-17 13:06 -0500]: > Ahh. So you are using tla as well. BTW, arch is a great source version > control system. I don't know why people are complaining about it not > being like CVS. :)
Well, not being like CVS is a feature IMHO :-) I use bazaar for development, which feels much nicer than tla and can handle tla archives. > A valid criticism would be the long paths which cause all sorts of > problems under Windows. Right, but that's nothing I really care about... > It is commented. All unix_socket_* are commented, > > #unix_socket_directory = '' > #unix_socket_group = '' > #unix_socket_permissions = 0777 # octal > > So I assumed that the socket would be placed in the default location. Right, and in recent versions the default location is /var/run/postgresql for clusters owned by "postgres", and /tmp for other cluster owners. So the client side seems to be okay, since it looks in /var/run/postgresql. I just found the reason why the server side is misbehaving. pg_ctlcluster never passes the "default" socket directory to the postmaster, so it will always be /tmp (the upstream default) if unix_socket_directory is not configured in postgresql.conf. My bad, I never noticed this since newly created clusters always set the option. I'll fix that now and will also add a testsuite check. Thanks for the report! Martin -- Martin Pitt http://www.piware.de Ubuntu Developer http://www.ubuntulinux.org Debian Developer http://www.debian.org
signature.asc
Description: Digital signature
