sftp_finish() which ends the communication, disables further
commands, and can explicitly be called more than once, and
sftp_end() which frees the state.

This prevents a race around sftp_finish() calls and sftpc_*() calls
which is likely responsibe for various weirdness in SSH connections,
and should fix ticket 135.

This eliminates the hack of using ssh_mutex to protect members of
sftp_state (which has its own mutex).

This doesn't fix the problem of sftp_state itself though.

This is not consistent with __DATE__ macro expansion, but I like it better

Heck, now that we're not really logging the build date much of anywhere, we
could go with any other more sensible date format here.

Wasn't sufficiently quoting, but we don't need to use echo in the first place

Will reconsider using %e instead of %d later

The Git commit date/time is more useful information.

Display the date of last git commit used in this bulid as part of git branch
and hash information. This helps to inform how up-to/out-of date the current
running build is, regardless of the build date/time.

There's some GNU make sorcery still needed here to address build errors
(though it does generate a working executable):
/usr/bin/ld: cannot find gcc.linux.x64.obj.debug: file format not recognized
/usr/bin/ld: cannot find gcc.linux.x64.lib.debug: file format not recognized
/usr/bin/ld: cannot find gcc.linux.x64.exe.debug: file format not recognized

DDFileLister: Removed unnecessary comments & stuff

See merge request !420

This one has caught me up before, for the CTX constructors didn't actually
accept any initial property values (as arguments).

mlong even documented it this way at
https://wiki.synchro.net/ref:uifc#javascript_interface

And construted his input.CTXs this way (using all 0 values, which were
ignored) in his xtrnmenucfg.js

POSIX requires a .. directory in the root of a filesystem, and
some SFTP clients (at least Filezilla and WinSCP) assume it's there
even if the remote doesn't list it.  Per POSIX, /. and /.. are the
same directory.

Translation from mouse coordinates to screen coordinates had issues
with internal vs. external scaling and for when there are gutters
(ie: black bars) on the edges.

This change snaps all gutter clicks to the nearest position inside
the window instead of discarding them, and ensures the calculations
are correct regardless of scaling mode.

Win32GDI mode was not tested, the only change there was to stop
ignoring mouse positions in the gutters.

This doesn't make anything better or worse for the C API, but doing
variadic pointer things in C++ is fraught.

If the sysop wants their login module to always execute for SSH connections,
set to this "Any" rather than the default ("Valid Key or Username").

And make the PETSCII ports more aligned with the other protocol ports
visually. If set to 0, say "<disabled>".

The stock text.dat FileAlreadyOnline string doesn't actually contain a '%s',
but if it did, it would've worked as intended in some cases, and broken badly
in others. Now fixed for if/when we do actually include the filenaeme (%s)
in that string.

... the library and directory (short) names

Making it more obvious *where* the referenced file has already been uploaded.

This allows me to shorten "Informational" to "Info" and "Debugging" to "Debug"
and be backward compatible with existing .ini files. I think "Info" and
"Debug" are universally known and take up less space (e.g. in SCFG).

Should only be run on official sbbs builds anyway

Some BBS programs (e.g. Synchronet v2) or possibly old offline message readers
don't deal well with long-line messages. So add a per-user option to word/line
wrap message text.

Addresses issue (feature request) #428
Thanks for the reminder fusion @ CFBBS!

As Nelgin pointed out, mouse hot-spots were still active when running external
programs (door games) - so clear any active hot-spots before running external
programs.

This could possibly be done in external() instead, but this sufficient for
now. I wouldn't want a menu that EXEC's something (via @-code) to lose all its
hot-spots as a result.

This is for sysops that never want mouse hot-spots in their BBS menus and
prompts (e.g. Kirkman, possibly Deuce) - set to "No".

Setting to "No" also disables the function of the JS console.add_hotspot()
method. Perhaps it should not. <shrug>

Setting to "No" does not disable the terminal mouse support question when
users are setting their terminal defaults. Perhaps it should.

This option should default to enabled (sysops that depend on mouse hot-spots
should double check) - it's enabled in new/stock installs.

Because the session is activated on the *next* time through the
loop, if the third attempt to authenticate (including public key)
fails, it would leave the session inactive and fail to log in with
a confusing error about an obsolete cryptlib.

On success, decrement ssh_failed to ensure another pass through
the loop.

Also, add more debug logging around the auth so you can clearly
see each failure, and log client_socket each time so it's clearly
grouped.