this is how prompt()/print() behave.  Colours should be "right" now.

field (as this was its intended purpose).

Fix bug where-by if a non-pass-through mailproc was addressed by an e-mail and
another recipient was addressed, the message would not be routed to the
mailproc.

Spam related messages and web server session timeouts.

slash (ie: virtually in the same directory).

amymore.

Only runs as a door now.

work.

ansi_restore() in place of dumb macros that don't check the current terminal
support.

comment regarding better validation of ANSI cursor position report response.

expecting an ANSI cursor position report and we will use this value to
determine the user's screen dimensions (currently, lines only).

(these are not fidonet standard values).

server now automatically includes the mail proc name in the log() output).
Outbound listserver messages now have the SMTP reverse-path (translated to
Return-Path header) set to the list server's address (so bounces will go there
instead of to the author of the original message).
The outbound messages now have an agent-type of "PROCESS" which will be used
to allow this script to process bounce messages in the near future (and remove
invalid e-mail addresses from the subscription list).

- Fixed line counter/return value of sockmimetext() - did not count last line
if not \n terminated.
- If a POP3 "TOP" command is used to request more lines than there are bytes
in message body, just send all lines and don't log an error.
- JS mailproc error reporting and log() method now includes the mail-proc
name and the "SMTP" log prefix. "JavaScript:" is no longer included as part
of the log() output since that is implied by the mailproc name.
- js_mailproc() now reuses the same runtime, context, and global object to
speed up the initialization/execution of multiple JS mailprocs per message.
- Beginning support for messages to be bounced back to mailprocs (e.g.
ListServer), specifically by ignoring bounces of SMTP sys messages
(instead of all non-person agent types). More work in bounce() is needed.

super-wonderfull srandomdev() where available.

Where we can't, try to use /dev/urandom if possible...

If that fails, start making stuff up...

At some point, the Win32 build should use CryptGenRandom() which is slightly
better on Windows NT/2K/XP and vastly better on Vista... not present on 9x
apparently though.

HAS_DEV_URANDOM and URANDOM_DEV based on that

as well.

This fixes a bug where /dev/urandom is not a character special file
(on FreeBSD it's a symlink) and there is a random() function available.

In this case, the same sequence of random numbers was used for each process
(most noticable with jsexec).

pause if necessary is expected to be handled by the caller.

if true.

Call console.clear() in draw_main(true) so the screen is cleared rather than
overwritten.  On some terminal programs (ie: SyncTERM) this forces the current
screen into the scrollback rather than removing it.

1) When a log file does not exist, mark it as having been parsed rather
   than retrying next time.
2) Keep track of the oldest since date and do not parse after it