Assign a timestamp to messages on creation and apply a timestamp to each
incoming message as we don't trust the client's timestamp anyway. This
avoids playing whack-a-mole trying to propagate the correct timestamp in
different situations.

Fixes #33

Makes sure multibyte chars aren't truncated.

Fixes #41

Just a quick fix until all the mwscript event handler stuff is mucked
out.

(Or anything that isn't .mwrc or a .js file)

function handler(ev)
{
	switch (ev.type) {
        case "mw_logon":
        case "mw_logoff":
        case "talker_join":
        case "talker_leave":
                onoff = ev.data;
		t = new Date(onoff.unixtime * 1000);
		mw.print("onoff event at: " + t);
		break;
	}
	return true;
}
mw.onevent.push(handler);

Don't send separate messages to announce these events, have mw display
them on receipt of onoff events, if scripted handlers allow it to.

If we dont alter the output string until we have a valid escape sequence
then we dont accidentally forget the current acting colour when bad
sequences are found

implements #38

Also fix and simplify the implementation of mw.command

Fixes #34

Fixes #36

Fixes #33

Seems a bit pointless. Also this is a portability improvement since it
doesn't work on *BSD.

Rearrange the order of the questions new users are asked, use more
welcoming words and separate out the data protection statement into a
config file (/etc/mw/data_usage.txt by default).

Should make it simpler to move logging into the server.

Apparently this is required since readline 6.3

Andrew Price authored Nov 07, 2015 and Imran Hussain committed Nov 17, 2018

(cherry picked from commit 711804a7)

    function handler(ev)
    {
        switch (ev.type) {
        case "message_received":
                msg = ev.data;
                ...
                mw.print("msg.channel.id: " + msg.channel.id);
                mw.print("msg.channel.name: " + msg.channel.name);
                mw.print("msg.channel.topic: " + msg.channel.topic);
                mw.print("msg.channel.soundproof: " + msg.channel.soundproof);
                ...
    }
    mw.onevent.push(handler);

Untested.

Configured with

-UDUK_USE_DUKTAPE_BUILTIN \
-DDUK_USE_INTERRUPT_COUNTER \
-DDUK_USE_EXEC_TIMEOUT_CHECK=check_exec_timeout

Turns out msg->bodylen doesn't include the '\0'. Luckily the json parser
doesn't go further than the first character after the outer '}'.

Lots of constification needed to make sure pointers into the allocated
memory aren't freed.