Since that's the most used/assumed byte order for IP addresses, let's make
that the default and eliminate a lot of unnecessary byte-swapping.

This fixes the issue reported by Keyop in IRC, introduced in commit 29b93c32

It's a bit different than straight 4:3 due to Mode 7 not using the
full line period or all the lines.

For non-Guest users (without the 'U' restriction), if the FTP client
sends the MKD or XMKD command, just return the proper succcess result
and do nothing (don't log as a hack attempt or return a syntax error
result).

Some FTP clients (e.g. Reolink Doorbell camera) just always try to
blindly create a directory before uploading files, but if that command
fails (as we would), the upload is aborted. So pretend to create the
diretory whether it already exists or not, just so the client will
proceed with the upload.

For G or U restricted users, treat (log) as hack-attempt, as before.

<@Deuce> Filenames are always assumed to be DOS programs, despite not being able to run them.
<nelgin> If sbbs was smart, it would know that if it's not a dos program then to use jsexec or whatever.

There: sbbs is now more smarter

We need to provide an address family "hint" to getaddrinfo() to limit the
results to IPv4 addresses.

Regression introduced in commit 29b93c32

Reported by Keyop via IRC

That's what Commstar uses in Prestel mode after all.

This mostly just changes the arrow key bindings and adds support
for Ctrl+Fx and Shift+Fx, and makes End Copy.

The big change though is that Page Down, not Tab now toggles reveal
mode.  Sorry all you guys with 60% keyboards.

For use by SyncTERM in Mode 7/Prestel modes.

It seems some of the Shift+x codes were incorrect (had unshifted
as high byte and shifted as low byte).

Basically, the only thing the SAA5050 enforces is that characters
on a bottom row need to be double-height, or they're displayed as
spaces.

The Prestel terminal now has an extra bit that indicates it is a
prestel terminal, and the bitmap console hacks up the second row.

While we're poking this, use macros for magic bits in FG/BG colours.

Needs clean-up still

No other changes

Since we do Mode7

Replace with \xXX values

This hopefully fixes the new CI pipeline issue

The smooth mosaics aren't available with the SAA5050.
Nothing to see here.

This sets us up better for adding more platform support (e.g. x64) later.

This makes ctrl/modopts/*.ini more useful for multiple language support

I accidentally committed an experimental x64 build configuration. I want to do
that right when I do add the x64 build configuration.

How these files differ from modopts.ini (and by extension,
ctrl/modopts.d/*.ini):

- The root section is always parsed, part of the returned object
- [sections] denote ARS that if matched (current user meets requirements)
  then those key/values are added as properties of the returned object
  possibly overriding previously defined properties (options), this allows
  different users (or clients/hosts, terminals) to get different option
  values much more easily

Up from the top row scrolls down, does not wrap to bottom.

Backspace from the top-left position scrolls down and moves the
cursor to end end of the new top line, does not wrap to bottom.

(so, [modname:charset=mode7], not [modname:charset=MODE7])

... cause it won't (fix directory case in path).

fexistcase() only corrects the case of the filename, not the directories in
the path. So we'll just use the lowercase version of the charset string,
always.

Also, return false upon failure of any call to replace_text(), not just the
last one.

This was the cause of some very slight funkiness when displaying files
after a save/restore (e.g. Ctrl-T or other global hot key) operation.

This is a getting a little kludgey

I think doing something with ARS would be nicer, but then we get into the
area of constructing a 'options' object out of keys from multiple sections
and which sections should supercede which? This'll do for now for Nelgin's
BBC mode7 terminal project.

For Nelgin's mode7 terminal project

No point in going any further and confusing users.

... in File.iniGetSections() method

because iniReadSectionList() doesn't recognize !include directives.

This fixes issue #871 whereby modopts.js couldn't find sections in
modopts.d/*.ini files.

I just want this in the history in case I ever want to play with
the concept again.

The fg/bg values all have multiple representations, and some bits
have extra meanings, so inequality there doesn't mean they're not
the same colour.

This fixes an issue where the *second* drag select on a screen
would not highlight blue backgrounds correctly.  The background
colour gets the "pixels set" bit flown because of how
restorescreen() works, which caused this to "know" it's not blue
and use a blue background.

A (good) argument could be made that savescreen()/restorescreen()
shouldn't result in any changes, but since uifc32 doesn't directly
use colours that aren't represented by legacy attributes, there's
no need for it to get tricky here.

SlyEdit: Display a user inactivity warning message (configurable)

See merge request !498