So use js_argc() to report an error if there's fewer than 1 argument, but
don't just fail silently when 2 arguments are provided.

Avoids crash when adding a msg to a non-sub-board msgbase.

It's possible to use savemsg() (from JS MsgBase.save_msg()) to add a message
to a message base that's not a sub-board (not configured in SCFG->Message
Areas and not the "mail" base), but in that case, savemsg() would dereference
an invalid sub in the scfg->sub array and crash. So use is_valid_subum()
to insure the subnum is a valid sub before using as an index.

Nonrelocatable const data goes in __TEXT,__const apparently.
I could put it in __DATA,__const and get an entry in the GOT, by
why bother?

If/when anyone wants this functionality again, write a short JS util.
For reference:
https://synchro.net/docs/utility_reference.html#SMBACTIV

This fixes issue #384.

This should fix issue #559, for example:
  Content-Type: multipart/alternative;
    boundary=mk3-ebfa33c1cd454cc2b1c618f5d74b41af; charset=UTF-8

We were assuming all boundary parameter values are quoted (apparently not
true, see https://www.rfc-editor.org/rfc/rfc2046#section-5.1.1) - so support
non-quoted boundary values which also may have a semicolon delimiter.

Also, truncate any white-space from a boundary value (shouldn't be there per
RFC2046).

Use this time-out value for calls to recv() and recvline() rather than a mix of
default and hard-coded values. If unspecified, uses a default of 60 seconds.

This fixes issue #562

This should resolve CID 453850

For graphical modes, this requests that ciolib be initialized in
the specified mode.  For text-based modes (ANSI, Curses, Win32 Console),
this is not used and "current" is used instead.

Now that vstat is what was last drawn on the screen we need to
save/restore the width/height around calls to bitmap_drv_init_mode().

Win32 rgbmap.s is different than other platforms, so they can't go
into the same place.

This is just to make it less noise to watch vstat.scaling from gdb.

Seriously, that's not possible.

My first idea was to have a per-external-program setting to enable input
translation, but Deuce suggested it would be better done in input_thread() and
after some thought, I agree. Translations are not done in data/file transfer
mode, so use the CON_RAW_IN console flag to indicate this condition. So even
JS console.getbyte() will return a translated char unless the console is set
raw input mode (this could be surprising behavior for some!).

I considered saving/restoring the console mode when performing a file transfer
but will leave that to a time when obviously needed.

This fixes issue #497 in automatic way (no option needed).

Should fix curses-only build.

Pretty sure this is what intptr_t was created for.

Stupid 1-based coordinates.

e.g. when using Ctrl-K or when being prompted for key-press due to inactivity

A bare '\r' doesn't move the cursor to the first column in PETSCII terminals.
The '[' Ctrl-A code moves the cursor to the first column with multiple
cursor-left sequences, if necessary.

Ideally this sequence would be part of the text.dat string instead of
hard-coded here.

This fixes issue #561

Note, the CI job is needs to be updated with AS=mingw32-as... need
to find where that lives.

Instead, generate these as part of the build and use the .incbin
macro on GNU-compatible assemblers to build the .o file quickly.

No good solution for Win32, but genmap could write the C file again
and take forever building it if/when we need to build rgbmap.obj
on Win32.

Rename to ciolib_[ry]2[ry]ptr and put in ciolib.c/h
Should fix curses-only build.

One or both of these should fix weird GDI zombie issues.

If the codepage tables have holes where US-ASCII chars go, don't
send the character under the assumption that it won't be on the
keyboard (and certainly not in the "same" place).