(from stored messages in .msg files into packets), the origin net/node within
the  *packed message* header would be set to the local system's address
net/node (for the destination zone), always. This is the appropriate action
when packing EchoMail, but not for NetMail.

Now, there are lot of other header fields in the packed message that contain
the source/origin address (e.g. @FMPT, @INTL kludges) and those would reflect
the original origin address still. And these fields take precedence over the
orign net/node fields of the packed message header. So I'm not sure this bug
would actually cause any problems anywhere, but it was a bug all the same.

letter).
Also, REALNAME is now an alias for COMPANY and MEMO2. Man these legacy
@-code names are terrible.

This makes the config files more portable/migratable.

to not work if the second byte of a stack-allocated but unintiallized variable
(str) was non-zero. This was only observable on Win32-Release builds
(in debug builds, stack variables are initialized to non-zero). Reported by
Sneaky (Ian) who was patient and kept with me through the remote debugging
effort. Thanks!

(directories, real or theoretical).

change the window size and it should work out alright.

into any other files.

and Frame.cleartoeol()'s instead.

on rPi3) by using Frame.home() and Frame.cleartoeol() rather than
Frame.clear().

where Display.prototype.__getUpdateList__ was taking > 1 second on an rPi3:
We need to avoid using Frame.clear() which invalidates the entire frame
and requires an update check of every cell (e.g. thousands in a typical
frame). So we're going to depend on Frame.cleartoeol() which needed its
own optimization to invalidate every cell of the current line even when
the cells were already blank (undefined). So only mark the cells for
update if they actually have data in them that is being cleared.

Also, the putmsg() method was taking on a CRLF to the output string when
word-wrap was enabled. So removing the trailing whitespace after word
wrap does its thing. If this breaks other scripts, then perhaps we need
to restore the original CRLF on the end of the word-wrapped string, but
*only* if there was a CRLF on the end of the original string to begin
with.

for the json-service.ini "dir" key. The code that parses that value is not
very tolerant of the absence of a trailing slash.

Add CSI I support.
Fix CSI SP d support.