globally and per-module. Command-line options still take precedence.
For jsdoor builds, jsdoor.ini is supported in whatever directory jsdoor is run
from.
Changed the default log level from Debug to Informational.
(If you need the old default, set LogLevel=debug in your ctrl/jsexec.ini file).

Update Copyright year to 2020.

text to an empty strings ("") in the ctrl/text.dat file, even the Name/Alias
prompt! Now this does mean that only terminal protocols where the username
is provided as part of the negotiation could be used to create new user
accounts (i.e. not Telnet), which I suppose could be fine for an RLogin-only
game server
.
If no user name is supplied by either protocol or prompt, an error is logged
and the function fails.

Fixed bug: for RLogin and SSH, if an invalid (e.g. blank) passsword was
provided during the handshake, an "Invalid Logon" would be displayed followed
by a "Passowrord: " prompt, giving the user an opportunity to reenter the
correct password. Only problem: this only worked for sysop accounts. Normal
users would just be disconnected after they entere (any) password, correct
or not.

This bug is very old in the RLogin support and appears to have been copy/pasted
for the SSH Login bit too.

and RLogin attempts.
It looks like someone copy/pasted the wrong lines from login.cpp.

an invalid login for a random (the last read?) user account with a user number
of 0.

configuration data.

Track and log the packet-export exceptions (link packets that were *not*
exported-to) and detail those exceptions (number of seen-bys, paths, origin
address, etc.) which were triggered.
This requires latest xpdev/str_list.*

- auto-increment the next sub position
- save and auto-increment the last conference number

This reduces the task of adding of subs to a QWKnet hub to just a bunch of
jamming on the ENTER key (usually).

<sys_zone>:<sys_net>/0. That's just surprising and undesired behavior.
Leave the old address in place in that case.

command-line (uisng -misc=<value> in addition to -import=<filename>). The
value can be decimal or 0x-prefixed hexadecimal, but you need to know the
property bit-flag value (SUB_* in sbbsdefs.h). For use with init-fidonet.js

why his messages aren't being exported to packets.
Don't increment the per-area-export counter when write_to_pkts() returns false.

JS numbers are double-precision floating points (doubles) and doubles cannot
be converted to unsigned integers uniformly across all architectures (e.g.
specifically on ARM, negative numbers would mostly get converted to 0U).

For details, see
https://www.embeddeduse.com/2013/08/25/casting-a-negative-float-to-an-unsigned-int/

yesno() and noyes() methods:
We need to check the argument count before using the second (optional)
argument or else it has some garbage/left-over value from some previous JS
function call. And no, just increasing the argument count in the method table
isn't enough. :-(
Reported by echicken and DaiTengu, thanks!

Moved some dbprintf() lines around a bit.

It would create 3 0-byte files (*.shd, *.sdt, *.sid), but more is actually
needed for a message base to be "created" (i.e. a call to smb_create()).
So, MsgBase.open() now uses smb_open_sub() rather than smb_open() to initialize
theSMB  status fields with the proper default values (based on the sysop
configuration) and calls smb_crate() if the header file is empty.
Yes, normally, SCFG creates message bases, but it shouldn't have to
(e.g. a fresh install on *nix, doesn't actually start with any files in
data/subs) and now that we have JavaScript-based message lister/readers, we
really needed this support.

- incomplete, it didn't account for the CLEAR @-code which is only handled here
- caused side-effects in that any Baja PRINT function call with a single line
  that contained any  ^A or @-codes would reset the console abort indication
  flag (SS_ABORT).
  Observed: hitting Ctrl-C at while executing the yesnobar.bin or noyesbar.bin
  would leave "Yes Yes No" on the console, for example.
  Solution: use the tos (top-of-screen) state before/after the call to ctrl_a()
  or show_atcode() to determine if the reset-abort hack should be applied.

"No" in yesno() and noyes(), unless the P_SAVEATR mode flag was specified
(which may be used for the old behavior).

a body.  Broken by HSTS enhancement.

There's a (small) possibility that the HSTS enhancement is now broken.

While the FastCGI read_wait_timeout function should never return
CGI_*_READY with CGI_PROCESS_TERMINATED, it looks like somehow on some
systems, there's an additional select() happening after the process is
done.

Explicitly track this, and try to prevent it.

Should fix the POST/GET hangs on the wiki once DigitalMan updates.

(SMTPREVERSEPATH a.k.a. "reverse_path") if it didn't already have one.
That's because smb_getmsghdr() will point the msg.reverse_path convenience
pointer to the from_net_addr if there was no explicit reverse-path (e.g. RFC822's
"return-path" header field). This could manifest itself in *any* change to a
message header via JS failing with an "illegal header length increase" error if
the added header field just happen to put the total header length over the
allocation threshold of the pre-existing msg header.

Fix: only model a msg header "reverse_path" property if the header field
actually existed (not based on the SMBLIB convenience pointer).
When the expand_fields option is used, the old behavior remains but expanded
headers cannot be written back to the base, so no harm there.

Reported by Coz in #synchronet from failed runs of scrubmsgs.js. Thanks!

Apparently some USENET articles are >= 1MB these days. Thanks Coz.

Treat \n the same as \r (next msg cmd)when reading msgs.

- don't indent keys
- separate sections with a blank line (to improve readability without idents)

A nice tip from Ragnarok (DOCKSUD)!

files were external to the executable (to conserve that precious 640KB of RAM!)
Some more use of SAFEPRINTF for good mesaure.

Use the COPYRIGHT_NOTICE macro rather than the build year.