can reduce the timeout simply by removing a has sign.

(lyskomd_start): New args: confs, texts and nogarb.

	Contrary to my belief, a timeout pattern seems to refer to a
	specific spawn id.  If we add it here when there are no
	spawned processes, the timeout will refer to stdin, and this
	will cause expect to detect eof on stdin when run from
	Xenofarm.  Instead, everything that spawns a process must add
	timeout patterns to the expect_active variable.
(l2g_start): Add a timeout pattern to expect_active.
(lyskomd_start): Ditto.
(lyskomd_start_fail): Ditto.
(client_start): Ditto.
(client_start_fail): Ditto.

(kill_lyskomd): Ditto.

	want_stale.  Copy number.txt.  Don't copy the database if
	init_db is 0.  Expect a log message about a removal of a stale
	lock file if want_stale is 1.  Expect log_messages after the
	startup messages but before the garb messages.
(lyskomd_fail_start): Copy number.txt.

(lyskomd_start): Removed unused global declaration of srcdir.
(lyskomd_fail_start): Ditto.
(client_death): Ditto.

	for stray output.  This makes it easier to see which client
	that misbehaves.
(suspend_client): Expect a response from the client.
(resume_client): Ditto.

	fooled us to test for "buffer_full" instead; that doesn't
	work.
(simple_expect): Ditto.
(good_bad_expect): Ditto.
(extracting_expect): Ditto.
(unanchored_expect): Ditto.
(lyskomd_start): Ditto.
(lyskomd_fail_start): Ditto.
(client_start): Ditto.
(client_start_fail): Ditto.

	tests to fail.

(lyskomd_expect): New proc.
(spawn_lyskomd): Use -f, not -d.
(lyskomd_start): Check all output from lyskomd.
(lyskomd_fail_start): Ditto.
(lyskomd_death): New argument: reason. Check all output from lyskomd.
(talk_to): Set current_talk_what and current_talk_nr, so that
	lyskomd_expect can restore the old settings.  Set proper
	line_leader for lyskomd.

	contained for instance a dot.  So this failed if a FQDN name
	is used.

	locksuite.py.

(obtain_lock): New proc.
(release_lock): New proc.
(l2g_start): Get the lock.
(l2g_stop): Release it.
(lyskomd_start): Get the lock.
(lyskomd_fail_start): Get and release the lock.
(lyskomd_death): Release it.
(client_start_fail): Get and release the lock

(lyskomd_fail_start): Ditto.
(client_start): Ditto.
(client_start_fail): Ditto.

(dbck_run): Ditto.
(spawn_lyskomd): New argument: log file name.  All callers
	updated.  Use the lyskomd.supp valgrind suppressions file.
	Pass --show-reachable=yes and a log file name to valgrind.
(check_valgrind): New arguments: logfile, need_leaks and need_errs.
	All callers updated.

(l2g_stop): Check valgrind output.
(lyskomd_fail_start): Use valgrind.
(check_valgrind): Ensure that the memory report is seen.
(dbck_run): Use valgrind.

	name of the valgrind binary to use.
(spawn_lyskomd): Adjust accordingly.  Don't use -v.
(check_valgrind): New argument: expected_leaks.
(lyskomd_death): Pass the new optional argument "expected_leaks"
	to check_valgrind.

(spawn_lyskomd): New proc, that knows how to use valgrind.
(lyskomd_fail_start): Use spawn_lyskomd.
(parse_valgrind_leak): New proc.
(check_valgrind): New proc.
(lyskomd_death): Call check_valgrind.

	avoid some TCL errors.

(lyskomd_start, lyskomd_fail_start, dbck_run): Ditto.

(lyskomd_fail_start): Likewise.

	expect_after, and fail if it is detected.

    Eliminated calls to build_matching_info when creating, renaming
    and deleting conferences. Integrated the conf_table array into
    match_table.

    Name matcher can handle holes in the table.

Testing:
    Added some name lookup test cases.

Distribution:
    Distribute lyskomd tests 12.exp and 13.exp.

Documentation:
    sub-comment was incorrectly marked obsolete. Fixed.