At the same time, fix an old error in prestel_handle_escaped() that
didn't pass the ctrl to prestel_apply_ctrl_after() properly, and
always flush the print buffer before calling prestel_handle_escaped()

We still shouldn't actually get C1 characters on a Prestel connection,
but if we do, at least we're doing the right thing now.

As for G1 characters, the default repetoire for these terminals per
ITU-T REC. T0101 is the mosaic character set with smoothed moasics,
something there's no evidence of a Prestel terminal ever supporting.
In order to implement those properly, I would need special handling
of "separated" in the renderer, and would need to fiddle with the
font to include the contiguous smoothed mosaics (and exclide the
separated ones).  Since this would end up allowing wildly un-Pretel
behaviour (ie: mosaics without a mosaic escape) I'm inclined to
just treat them all as spaces and move on util/unless I do full
support for all the Videotex standards.  Nobody has asked me to
open this can of worms, so I'm just going to leave it on the shelf.

Blinking text is opposite of the cursor

For "PC" modes, cursor blinked at 1/16th of the VSYNC rate, so use
the CGA timings, where were generally the slowest of the bunch...
This means cursor blink at 3.745Hz, and character blink at 1.8725Hz.

For Prestel, character blink was 0.75Hz with a 3:1 On:Off ratio.
The cursor on the other hand was 1/32th the VSYNC and was generally
PAL, so 1.5625Hz.

I still need to dig into C64 and Atari modes, because they're likely
way off now (assuming either supports blinking).

Only 0-31 and 128-255 are left needing CP437 love.

Also hack together a 12x20 "CP437" font that's just the Prestel font
with the characters uifc32 uses from cp437 in it.

Update for v1.4

Move reveal/dirty status into background colour.
Should fix ticket 182.

No change in functionality, just code comprehension

If enabled, this option prevents logins to this service from changing the
protocol (aka modem) field for a user logging-in.

This is a work around for Deon who has some message areas set to require
"PROT NOT IMAP" for access requirements and when his IMAP client connects
and logs-in, that changes the protocol field in his user account (to "IMAP")
which means that user account cannot see/read those message areas (until the
protocol field is changed by another login to another server/service) and
thus won't receive echomail notifications for message received for them in
those areas.

Defeat the Windows getaddrinfo feature:
If the pNodeName parameter contains an empty string, all registered addresses on the local computer are returned.

at least for me, "fe80::5dfc:b29c:85d6:8de7%21"

at least, "vert.search.charter.net" in my case

This is just the ones mentioned in the jsdocs.  I'll add more
if/when I want them. :D

corresponding wth new user DELDATE field

alert() doesn't throw exception on null/undefined, just no arg

deny() throws exception on no args or null/undefined arg

So don't check for null/undefined argv[0] if not passed one

Hopefully this will fix the occasional crash in the testsuites
(and presumably, occasional crash in other things that use background
threads with js.exec(), if there are any).

Except this time, in the right place.

This reverts commit ed8c30da.

This reverts commit 7eb498aa.

Hopefully this will fix the occasional crash in the testsuites
(and presumably, occasional crash in other things that use background
threads with js.exec(), if there are any).

User records (aka slots) will be preserved this many days since the user was
deleted (if the DELDATE field is non-zero) otherwise, days since last user
activity.

The initial purpose is to support new user field: DELDATE (user record
deletion date) - this allow user record retention for a period after
deletion rather than after period after last user activity.

The Win32 usereditor needs updating to use this function too.

Reported in unmonitor, which always passes NULL.

Previously, different files had different sets of wrappers around
pthread_mutex_* and rwlock_* functions.  This commit removes those
and puts assert_ prefixed ones in threadwrap.h and rwlockwrap.h.

It also uses them in all the bits that SyncTERM uses.

This means debug builds are much more likely to unexpectedly
terminate with an assertion... when it does, let us know!

While we're here, add debug assertions to the pthread stuff.