Touchspots lag behind high performance clients

[Daniel van Vugt]

If a client is careful with how it processes input, takes the raw input stream (MIR_CLIENT_INPUT_RATE=0) and uses swap interval zero, then that client will actually render faster than (ahead of) the touchspots.

This is probably because old input data is being used for the touchspots. A framedropping client has the benefit of only being a couple of milliseconds old when it appears on screen, at most. Whereas the touchspots are likely being drawn based on input information that's 16ms old.

To achieve minimal lag (for touchspots or any client), you need to resample the input coordinates just before scan-out. As the client is in this case, but touchspots are not.

[Michał Sawicz (saviq)]

Syncing task from Mir.