chytil@vlsivie.tuwien.ac.at (Chytil Georg) (11/15/90)
I'm currently in the frustrating process of installing bind 4.8.3 from ucbarpa.berkeley.edu on an Apollo DN3500 running Domain/OS 10.2.0.5, more or less BSD4.3 . After some minor hacks (thanks, Apollonians !) it finally compiled fine, and I'm now stuck in a runtime-problem : When configuring itself through getnetconf() named tries to open a socket on the net-interface (ok) and then on the local-loopback too --- and then fails with an 'Can't assign requested adress' . Sigh. While named's code is very readable as far as the how is concerned, I'm not always sure about the 'why' (Well, this may also be due to my inexperience, but I'm not that daring to admit this ). Why is the check_for_local_loopback situated _after_ the opensocket() in the code ? Why should local loopback interfere with the physical interface ? (Why can't I join the happy and drunk VMS-guys in the pub at 4pm sharp 1/2:-( 1/2;-) ). some debugging-information follows : Upon coming up named announces to syslog : Nov 13 17:11:55 walter named[287]: restarted Nov 13 17:11:55 walter named[287]: bind(dfd 6, 127.0.0.1[53]): Can't assign requested address /usr/tmp/named.run shows : Debug turned ON, Level 1000 Version = named 4.8.3 Wed Nov 7 19:07:03 MST 1990 chytil@vlsivie://vlsivie/users/chytil/named/named.new/named bootfile = /etc/named.boot dqp->dq_addr 128.130.40.5 d_dfd 5 dqp->dq_addr 127.0.0.1 d_dfd 6 point of failure is the call to bind() in opensocket -- should I add REUSE_ADDR (sp?) to socket-options or is this left out with purpose ? Thanx for Your help Georg