Fixes #134.

The correct name should be XSECURELOCK_SAVER_STOP_ON_BLANK. We also
starts children when screensaver is on, but screen is not blanked, or
when screensaver is one and screen is blanked, but this option is not
enabled.

Should help with issue #48.

Got new features.

Useful for waking up the locker on system resume from suspend and
similar events not visible to X.

This enables children some time to initialize and prepare their
content. They are started before the main loop, we sleep
XSECURELOCK_SAVER_DELAY_MS, then map the windows. The saver window is
mapped first, then the background window and the obscurer window. This
order works fine for me and I don't get a black flash anymore, despite
the saver being written in Python.

On shutdown, the order is also a bit altered to reduce the flash due
to the obscurer window in white.

The multiplexer is left untouched as its parent window is not mapped.

Fix #94

Co-authored-by: Joshua Trees <me@jtrees.io>

* configurable background color with XSECURELOCK_BACKGROUND_COLOR

* Add docs regarding `XSECURELOCK_MULTIPLEX_BACKGROUND_COLOR`

Co-authored-by: Cullen Ross <cullenrss@gmail.com>

Upstream xscreensaver is using `/usr/libexec` instead of `/usr/lib` since version 4.21

This reverts commit 66abedd7.

Somehow I failed there.

No actual "contentful" changes expected here.

This also applies to `saver_mplayer`, hence the more generic name.

This changes avoids switching to black between mpv invocations when we reach the end of a playlist. This is quite notable when using only one image.

This will prompt PAM to refresh Kerberos credentials and similar tokens
based on the authentication that just finished.

added annotation to where to find date format

Prevents needlessly launching the saver at startup when launched by xss-lock and the screen is already blanked.

Related to issue #84.

This isn't quite what I want, so for now it's already marked deprecated.
Only there for testing a solution to issue #95.

The leak does not actually have any impact as it only occurs in the error case
in which the calling process (auth_x11) exits anyway; yet fixing this
anyway seems a good idea.

cat_authproto (the demo program) had an actual, real leak with this,
even - but it is a test program only.

Also remove a potential bad pointer dereference in case NULL isn't the
zero bit pattern.

Fixes #96.

It documents where the xbm file comes from.

Fixes #90.

Now it compiles as C89 again.

Fixes #85.

That way, when a user ever sees it, they see the cause of the problem:

INCOMPATIBLE COMPOSITOR,
PLEASE FIX!

Then we can safely show Ctrl+Alt+O on the screen :)

At least one user tried to press zero instead of oh; changed Ctrl-Alt-O
(upper case) to Ctrl-Alt-o (lower case) to disambiguate.

Should fix crash issues with nvidia driver.

Credit goes to tommie.