- Move closing of servlink control fd to close_connection()
instead of doing it in exit_local_server(), and make sure
we first close the data fd and then the control fd.
- Have servlink process ready fds in order net, data, ctrl
instead of ctrl, data, net. This seems to fix the problem
that squit reasons do not show up on the other side of a
ziplink (by making it send any final SQUIT and/or ERROR
before noticing the closed control fd).
WHOIS: Show the given nick in ERR_NOSUCHNICK even if it
starts with a digit. This is safe because the last
parameter is a nick and never a UID. (The first parameter
in a server-server two-param whois is a UID/SID.)
If we are connecting outward to a server, check if the
server name they sent is the same as what we tried to
connect to. Previously such a connection could succeed
if there existed connect blocks with the same IP and
passwords for the other server name.
Another handling of SJOINs without nicks:
Propagate them if the channel is +P or the channel
already existed, otherwise remove the channel again
and do not propagate the SJOIN.
addition to +i. As before, a restrictive mode must be in
place at /invite time for the invite to have an effect;
+r does not count as a restrictive mode if the user is
logged in; +l and +j always count as restrictive modes to
allow for cases where they would allow join at /invite
time but not when the user tries to join.
overflowing the client exiting server notice (from
TOPICLEN to 260). kill reasons become shorter accordingly.
kline/dline/gline reasons become 390.
away messages stay at TOPICLEN for now.
Change TS6 JOIN processing
- don't send out simple modes in TS6 JOIN and TS5 SJOIN when
a local user joins an existing channel
- don't send out simple modes in TS6 JOIN and TS5 SJOIN when
propagating a TS6 JOIN
- don't interpret simple modes in an incoming TS6 JOIN
This is to avoid desyncs when certain mode changes (e.g. -im)
cross with joins. A downside is that simple modes will be
more desynched when a JOIN creates a channel or lowers TS,
but that's less important.
Update the TS6 specification to include this, and clarify
that TMODE can come from a server and that MODE must be
translated into TMODE from other servers too.
- keep the bitmask reserved forever to the letter, fixing
the problems when loading multiple umode modules,
unloading them and then loading them in a different order
- don't allow local users to change umodes which have
been unloaded and don't set them on new users via
default_umodes
server name existed taking hostmasking into account
and just check with find_server(); admittedly
this checks if the name is a SID but that's not
a real problem.
to specify the full (unmasked) name of a server behind
a hostmask. As a result find_any_client() (for prefixes)
becomes equal to find_client(), so remove that too.
This makes it impossible to connect hostmasked servers.
(This support didn't work well anyway, was incompatible
with TS6 and we never masked ourselves.)
jilles