The "official" log line format is (as of right now at least):
"[socket] [protocol] [user/host-ID] [! if error]message"
Hopefully I can remember that when I add new log messages.

Also setting the protocol to SMTPS, POP3S, SEND/TLS, when appropriate and
passing it around to pretty much any function that can log a message.
Added debug-level log output when the mail sever is sending message body text
(every 100 lines).

The "max lines" parameter value passed to sockmsgtxt was using the cryptlib
session number - this would truncate sent email bodies to some seemingly
random number of lines. Introduced in rev 1.622.
Also added a couple of helpful log lines to let the log-reader know if/when
TLS was successfully negotiated for sendmail sessions.

(defaults to 10 megabytes, "10M").
When SBBSecho starts up, if this minimum amount of free space is not
available in any of the following directories, SBBSecho will display an
error and immediately exit:
- data directory (in SCFG)
- logs directory (in SCFG)
- netmail directory (in SCFG)
- outbound directory (from sbbsecho.ini)
- temp directory
- each linked-node's outbox, if configured

If the sysop runs out of disk space for some reason, at least SBBSecho
won't contribute to the problem.

so remove 'errno' value (and corresponding error strings) from generated
/logged error messages. fseek(), read(), write(), open(), etc. do set errno
to a meaningful value, so we still log errno/strerror() in those failure
cases.

prompts with the conditional beep enabled only with the '-b' option
(silent by default). Thanks to Nelgin.

with smb_t.status.

not null on failure.

Also, is that some line ending weirdness?  *shrug*

'l' (list messages), 'x' (dump index), 'r' (read msgs), and 'v' (view msgs)
commands by "days old", by specifying a negative number (in days).
(e.g. "smbutil r-30" would read messages posted within the past 30 days).
This was added to test the smb_getmsgidx_by_time() function, but has some
value otherwise I suppose.

don't try to (re)lock it, just return SMB_SUCCESS.

I honestly don't recall why I made this change, but it makes sense, so I'm
committing it finally.

algorithm now. Also, now returns the index offset of the matched index
(if found), negative error value otherwise.

I was testing something and noticed that the old smb_getmsgidx_by_time()
implementation could sometimes return the index of a message posted before
the requested time. That was fixed by re-writing the binary search logic.

Also, I found it useful in my testing to have this function return the
offset of the matching index (in addition to the index record itself), so
checks of the return value need to compare >= SMB_SUCCESS, not just ==
or != SMB_SUCCESS.

userdatdupe() to match the passed login-id (username):
*  This allows more permissive username matching, e.g. if your user name
   is "Analog Kid", you could login with "analogkid", "analog.kid" or
   "analog_kid", which is handy when passing a user name on the command
   line (e.g. to an ssh client) for example.
*  This is the same function used for Telnet logins via exec/login.js

so that the current sub/group and dir/lib short name is displayed as
part of the prompt.

@-codes within them. Previously, only if the entire mnemonic string was
comprised of an @-code would it be expanded, now any @-codes within will expand
as well.

experimental 'U' (View Unread-Only toggle) command. The 'U' command from
the e-mail menu (e.g. email_sec.js) works better than the toggle approach and
is probably all that is needed.
Also, the user number lookup in the User Edit bit is better at finding the
proper user account (e.g. for received netmail).

from the Chat menu would immediately clear the screen and redisplay the
menu after displaying the change in the node's state. This was due to the
enhancement in rev 1.77 of con_out.cpp whereby we no longer increment
lncntr if the first lines displayed after a key-press are blank (who
cares if blank lines scroll off the screen?). That's fine and all, but the
auto-pause check in outchar(FF) would only auto-pause if the line-counter
was greater than *one* (has been this way since forever). So with the
enhancement in 1.77, a single non-blank line displayed after a key-press
was now not enough to trigger the auto-pause ([Hit a Key] prompt).
So I've changed the line-counter trigger value to greater than *zero*,
which means we could have some new excessive auto-pauses, but other places
where auto-pause wasn't working, should again.

One new excessive auto-pause was after the system password was entered at
the SY: prompt, so that fix is included in this commit.

default location to find !include'd baja (*.inc) files. If you use the '-I'
option, it will override the environment variable setting. For Nelgin.

using SSH: the output from the program (file transfer driver, door game)
would reach the SSH client, but input from the client (or user) would not
reach the external program:
passthru_output_thread() - responsible for SSH input from the client/user
to be sent to a "raw" pass-through socket used by the external program,
had a reversed logic bug added around rev 1.686.
Problem was introduced as part of the cryptlib update/SSH overhaul of
early 2018. Thanks to Thumper @ THEWASTE for the bug report!

the sysop must now enable this feature in SCFG->Networks->FidoNet.
Also, widen the FidoNet configuration menu in SCFG, display more FTN
addresses and the default origin line (part of it, at least) in the menu.

the sysop must now enable this feature in SCFG->Networks->FidoNet.

for DOS programs (part of the USE_DOSEMU patch).

flags set manually in SCFG.

(messages which contain the RESCANNED control line) to downlinks.

Also renamed pkt_to_pkt() to write_to_pkts().

"Conditional jump or move depends on uninitialised value(s)"
Thanks, Nelgin.

If the ctrl/ssl.cert file did not exist, the various TLS/SSL services would
fail to create it and (sometimes) log the "creating SSL context" error.
This problem was introduced with rev 1.30 (Mar-9) and the fix/patch
provided by Deuce via irc. Thanks to Psi-Jack for the bug report.

(previuosly) silently fail - in the sendmail_negotiate() and pop3_thread().
If get_ssl_cert() can return -1 withOUT setting the estr argument to a
non-NULL value, then we could still have silent failures (TBD).

(out of disk space?).

properties to msg_area object. See NMAIL_* in sbbsdefs.js for the bit values.

- finduser() method no longer prints "Unknown user" upon failure (the calling
  script can do this if/when it wants).
- read_mail() method now accepts an optional 3rd argumnet (loadmail_mode)
  which may be used to load *only* un-read mail (for example), see LM_* in
  sbbsdefs.js

Pass the lm_mode as a 4th argument to any readmail mod (if one is installed) -
this correlates with the LM_* mode flag values (0 is normal/usual).

File attachments can also be in a file/####.out directory (e.g. outbound
Internet email attachments), so try deleting from there too when deleting
mail with attachments.

- similar to what we do in netmail() with the WM_NETMAIL flag.
If you call writemsg() without WM_EMAIL set, it tries to use the subnum as an
index into the sub-board array and will crash.

var text = load({}, 'text.js');