Gree: Fix reporting vertical swing

[gentoo-root]

toCommonSwingV() is only called when SwingAuto is false, but it converts kGreeSwingLastPos to kAuto. It doesn't make sense, because:

1. kGreeSwingLastPos means that swinging is stopped (i.e. the shutter remains in its last position), which corresponds to kOff.
2. kAuto shouldn't be returned from this function at all, because it's handled separately in toCommon() when SwingAuto is true.
3. As can be seen in setSwingVertical(), when automatic is false, the valid set of positions includes kGreeSwingLastPos, but not kGreeSwingAuto.

Fix the logic by amending toCommonSwingV() according to the considerations above. It fixes parsing of received IR packets when the user disables vertical swinging from the remote (tested with YAP1FB).

For consistency and robustness, educate setSwingVertical() and convertSwingV() about the supported kGreeSwingLastPos mode.

Add a unit test for the described bug.

[NiKiZe]

I'm curious if there could be any case where the existing behavior is the wanted one. (Don't want to break anything, at least not without documentation)

[gentoo-root]

if there could be any case where the existing behavior is the wanted one

Well, given the considerations that I gave in the commit message, it doesn't look like it was intended. Too many things don't add up. Plus I added a unit test that sets kOff and checks if the reported state becomes kOff (it wasn't the case before my patch) — I don't think any other result should be expected after setting kOff :)

[NiKiZe]

Yes your test case is perfect. And at least also shows that there isn't any existing cases that we expect to behave differently.

To be honest I don't follow the explanation fully. (I don't disagree)
Reading again I do fully agree.