more control over SBBS-initiated netmail source/origin addresses.

user choose which address to use as the source address when composing a netmail
message (the default being the most appropriate for the dest zone/net). This
change only works with SBBSecho v3.05 or later.
Some other incremental and safety improvements to sbbs_t::netmail() too.
FTN netmail file attachments needs some more work however (the "FA:" subject
prefix trick) - just remove it?

If the SMB header contains a source FTN address, use that as the origin address
of the netmail and do not look-up a local AKA match for the destination address.
Also, fix what appears to have been a (currently harmless) bug in
smsg_to_echostat_msg(): the msg.from_net.addr is not an ASCIIZ string when
net.type == NET_FIDO. Currently, source FTN addresses aren't set in echomail
message headers, where the echostats come from. Could just remove
these 2 lines.

When the local system has multiple AKAs for the same zone and we are picking a
origin/source address suitable for the destination address, pick the AKA that
matches both the zone and net of the destination address first (if there is
such a local AKA). It's the same logic used in sbbs_t::netmail() to display
the originating address, so the AKA picking logic now matches what is shown
to the netmail author and what SBBSecho will actually use.

(apparently this is a pseudo-standard supported by fastecho and other
"old-school" tossers) - per mark lewis' request.

existing tear/origin lines and don't add a new origin line.

"expand fields" could be misinterpretted (e.g. as 'false') if less than 3
args were passed to the function. Apparently you can NOT assume that argv[argc]
is undefined and would fail a JSVAL_IS_BOOLEAN test. In the reported problem,
MsgBase.get_msg_header() was being called with 2 arguments (from newslink.js)
and the if(JSVAL_IS_BOOLEAN(argv[n])) test, when n was 2, would eval to true
and then argv[n] evalulated as false, which would cause a message with no
message ID to not have one dynamically created, which would then cause the
message to fail to post to an NNTP server due to malformed Message-ID (a
missing message "id" property would end up being included in the newsgropu
article header as "Message-ID: undefined").

get_msg_index() had a similar potential issue, also fixed.

Also, use a stack pointer if the passed failed pointer is NULL.

length, just delete it (don't rename to *.?_?).

new contents,
Instead truncate to the new length after writing the new contents. This should
make the files more tolerant of "out of disk space" situations and reduce
the risk of lost data (0-byte files) due to low/no free disk space.

instead truncate to the new length after writing the contents. This should
make mail.sid files more tolerant of "out of disk space" situations and reduce
the risk of lost data (0-byte mail.sid files) due to low/no free disk space.

instead truncate to the new length after writing the contents. This should
make .ini files more tolerate of "out of disk space" situations and reduce
the risk of lost data (0-byte .ini files) due to low/no disk space.

thus fixsmb) to fail, but just before exiting, fixsmb would delete the .lock
file (whoops). The msgbase.lock file means something else is rebuilding this
base (e.g. smbutil) and thus we can't just go deleting the .lock file, unless
we (fixmsb) was the one that created it.

left-justified server prefix), e.g. "term" now added for terminal server
errors.

- Using NOTCICE (not INFO) log level.
- Move the protocol name after Node X and remove the colon.

(e.g. warnings and errors).
Also, stripped trailing white-space from many lines (editor setting).

Terminal Server SSH ERROR 'Bad argument, parameter 1' (-1) ...
from output_thread
Don't set ssh_mode=true until the ssh_session has been created successfully.

Also added some more debug output with regards to Cryptlib session number.

parameter 1' (-1) ... from output_thread

Terminal Server SSH ERROR 'Bad argument, parameter 1' (-1)
 ... from output_thread

The bbs_thread() sets the global/server sbbs ssh_mode to false and ssh_session
to 0 (the "parameter 1" value used in the cryptlib function calls in
output_thread) but was doing this without owning the ssh_mutex, so the
output_thread had a race condition where it would check ssh_mode=true and
then use grab the ssh_mutex and use ssh_session in a few cryptlib function
calls. The fix for the bbs_thread() to grab the ssh_mutex before setting
ssh_mode to false and ssh_session to 0 and have the output_thread() re-check
the ssh_mode after grabbing the ssh_mutex and not call any cryptlib
functions if ssh_mode was set to false while waiting for the mutex.

The cause would have been more obvious if the various cryptlib error/log
messages contained the cryptlib session ID value (which was 0 in this case).

WARNING.
Also fixed a couple of typos in strings.

switch to CP437 before any uifc stuff and everythings going to be happy.

missing, but also now logging when renaming (or failing to rename) bad
received .REP packet files.

packet errors, no need to save such REP packets as .bad files.

This should fix the issue with non-CP437 fonts reported by Kirkman.

error messages (in the POP3 server) as the msg.hdr.number is zeroed by
smb_getmsghdr() and potentially invalid when returning an error.

error log message as the header is zeroed in this function and thus the
message number previously read from the index, lost (zeroed) and useless.

screen of rows.

*nix-builds (for stdio WWIV doors).

Use drive mappings (e.g. D:, F:, G:, H:) rathe than absolute paths to node_dir,
ctrl_dir, data_dir, and exec_dir in drop files.
Also: fixed what appears to have been a copy/paste error in the node_dir
and ctrl_dir mapping to Win32-shortnames for non-native doors on Windows.

attachment is attempted but unsupported.

- writes outbound netmail to SMB rather than FTS-1 Stored Messages (*.msg)
  this means you must have SBBSecho v2.30 or later in order to export FTN
  netmail (for file attachments, you'll need SBBSecho v3.04 rev 3.78 or later)
- removed support for the now deprecated "default address" feature
  (nowadays, people are much more likely to gate FidoNet over Internet rather
  than the reverse!)

The QWK-to-NetMail function still needs a similar overhaul.

(e.g. "C:"), skip it on *nix and if the path contains backslashes, convert
to forward-slashes (for all OSes). This resolves an issue of sending a
netmail file-attachment that was created on a Windows/MS-DOS type system
and then scanned/exported on a *nix system. The FTN "stored message"
convention is to have the full path of the attached file in the message
subject.

Hopefully nobody wants to use outbound directory names beginning with "x:"
on *nix. :-)

included an "old-style" (not MIME-encoded) file attachment (where the file path
is in the subject and the FILEATTACH aux-attribute flag is set), the
corresponding netmail attribute flag would not be set in the stored msg (*.msg)
file. Also simplified the create_netmail() function prototype/usage, removing
the "file_attached" argument (use the hdr.auxattr instead).

1: Wrap SSH cryptDestroySession() calls in a wrapper that tracks the number of
    open sessions and logs errors upon failure (and remaining session count).
    This is how I found the SSH session leak in sbbs_t::hangup().
    Also found multiple cals to SSH_END() which were invalid (no session to
    destroy) and removed them.
2: If errors are detected when unpacking a REP packet, save the .rep file
    (in data/file/<filename.rep>.<timestamp>.bad) for later inspection.
    Also added some VOTING.DAT debug output (to be removed later).
    Trying to the get to the bottom of the
    "smb_addvote thread_back field missing" errors.

pause and CLS after.