This shouldn't be necessary but may work around graphics driver bugs.

Beware: this patch may cause flicker when screen saver is in use. Have
not made sure it won't.

May fix #80.

Makes following semantic versioning easier.

Behaves somewhat better if the value is low; still could not get a zero value
to work, as X11 then still wakes up the screen on key-release events.

Note that after this change, SIGUSR1 is only sent if
XSECURELOCK_SAVER_RESET_ON_AUTH_CLOSE is enabled, and savers
should ideally handle it by soft-resetting the saver (implemented in
saver_xscreensaver only for now); not handling it will cause a hard-reset
of the saver by restarting it from scratch.

Fixes #75.

That script absolutely works from a release tarball.

Make it run at 60fps by default now, no longer 112fps when dithering.

Fix the displaymarker sometimes only moving by one character when starting from column zero (empty PW).

Without this, an attacker might infer the password length based on a
screenshot and the time of day, as rand() is a deterministic PRNG without
any runtime entropy sources.

After this change, the attacker will additionally need the exact
microsecond-accurate time of day, and the process ID.

We may alternatively move to using /dev/urandom or getrandom here, but this
does not seem worth the effort, given an attacker can learn the same info by
pointing a video camera at screen or keyboard, or a microphone at the keyboard.

Also stop accidentally including some .o files in the tarball; found in the
checking done by the script.

Not very fancy yet, but should do for now.

Contribute the example saver for displaying livestream feeds.

Internal reference: b/130370819.

Now the colors are only hardcoded in one place, making it easier to override them.

Not doing so but running xsecurelock via shell expansion means that the shell
will keep holding on to the sleep lock FD, preventing suspend.

(cleaner)

The easiest way to do that was to execl() a separate binary for the
placeholder (thereby closing all close-on-exec file descriptors, even
those we don't control, e.g. X11's), also making "ps" more descriptive
about what this process is for, and making sure any e.g. internal mutexes
by libraries we use are freed.

Fixes "systemd does not suspend laptop on closing the lid".

Otherwise the sigsuspend() call to wait for SIGCHLD does nothing and just hangs.

Also, refactor to prevent this issue from happening in the future.

$0 didn't contain the full path, so exec failed.

Fixes killing saver children twice twice.