🚀 go-pugleaf

The way the ISODE servers work is that if their stdout is
non-existent, they assume that they are running as background daemons
and write to their configured log file.  If they have a stdout, they
assume they are running from the command line (e.g. because they are
being debugged) and they write to stdout/stderr and to log files in
the connected directory.

Some genius at Sun decided that they should change the way /etc/rc*
scripts work.  Back in the good old days(:-) it used to be that they
had no controlling terminal and programs that wanted one had to
explicitly redirect output (e.g. to /dev/console).  So, you would see
things like:

	if [-f /etc/foo ]; then
		echo -n 'starting foo server: '		>/dev/console
		/etc/foo				>/dev/console
	fi

	if [ -f /usr/local/etc/tsapd ]; then
		echo -n 'starting ISODE: '		>/dev/console
		/usr/local/etc/tsapd &
		echo -n 'tsapd '			>/dev/console
	fi

This causes stdout to go to /dev/console and also makes /dev/console
the controlling tty for the /etc/foo process.  And the tsapd runs
without a controlling tty, figures this out, and writes to logdir.

Somewhere along the line, this behavior changed such that all the
/etc/rc scripts run with a controlling tty of /dev/console and
stdin/out/err redirected there.  So now you see scripts like this:

	if [-f /etc/foo ]; then
		echo -n 'starting foo server: '
		/etc/foo
	fi

	if [ -f /usr/local/etc/tsapd ]; then
		echo -n 'starting ISODE: '
		/usr/local/etc/tsapd >/dev/null 2>&1 &
		echo -n 'tsapd '
	fi

In other words, to get the "correct" behavior in an /etc/rc script you
have to explicitly redirect stdout/stderr to /dev/null.

We reported this to Sun as a bug and were rebuffed.  It was certainly
an undocumented change which broke many of our programs that relied on
not having a controlling tty.

/a