Before this fix there sometimes in the last iteration of the while loop
elapsed was grater than fade_time and that resulted in a tiny full
brightness blink at the end of the effect (when the key should be almost
turned off).

Fixes #1337

This is required for several new devices.

Fixes #1379

Signed-off-by: Marcus Weinhold - umwerk systems GmbH <marcus.weinhold@umwerk-systems.com>

Includes DeathAdder 2013. Fixes #1350.

The daemon would throw an exception upon next run as it was reading a
bytes when it expected an integer.

The starlight effect and some brightnesses were affected.

Fixes #1351.
Fixes #1354.

Fixes #1352

This patch adds several D-Bus methods for accessing the current state of
the devices (plus pylib API for the new functionality):
* get*Effect
* get*EffectColors
* get*EffectSpeed
* get*WaveDir
* restoreLastEffect

The star means e.g. getEffect, getLogoEffect, getScrollEffect, etc.

The last one can be used to reset to the previous effect after setting a
custom effect (which won't be persisted).

Co-authored-by: Luke Horwell <code@horwell.me>
Co-authored-by: Luca Weiss <luca@z3ntu.xyz>

Devices were throwing an exception that was:
- "embedded null byte"
- or "setCustom is not a valid method of interface"

This was due to devices missing a sysfs file in the driver. Or rather,
they had one (matrix_effect_custom), but was incorrectly specified here.

For devices that are missing a sysfs file, it's possible custom frames
are unsupported or was simply left out in the driver when the device was
added. Users of these devices should try amending their device in the
driver, as we don't have the hardware to check.

Fixes #1252
Addresses polychromatic/polychromatic#266

Closes #1304

https://github.com/bithatch/snake

This doesn't work well when the task is backgrounded and is not really
useful.

This option creates problem with the event files.

Reason is that the created event files are fifo pipes which - if not
opened on this side - will block on the other side (the daemon) and hang
forever. So we need to keep these pipes open to not make the daemon
misbehave.

This should make it much easier to just create all fake devices
that are available.

Previously we would always read the serial number and sleep 0.1 seconds
even if querying the serial number has succeeded. Move the sleep to only
sleep when really needed.

This decreases the init time of the daemon for ~80 devices from ~10
seconds to ~2 seconds.