Hi Jason, On 03/08/2012 08:06 PM, Jason Gerecke wrote: > While I agree that [0, 127] isn't obvious, changing it to be > 0-symmetric isn't going to happen any time soon. The X driver > unfortunately expects that range -- it doesn't ask the kernel. Any > kernel change would have to come quite a while after the X driver was > modified to ask the kernel instead (otherwise we risk breaking tilt > for anyone with an old Xorg driver)
Yes, I understand that. Can I suggest a plan? Have the X driver always center the range around zero. If axis resolution isn't specified, assume the range covers [-60,60] degrees. Otherwise, calculate the physical range from resolution (which is points/radian), clip it to values corresponding to [-60,60] physical degrees and report without scaling to the applications. This is, of course, assuming the Wacom driver would want to handle non-Wacom tablets. If not, then, for the sake of uniform evdev protocol, only centering needs to be added. After some time passes, say a year, update the kernel driver to center the range around zero, remove centering from the X driver and assume 0 is 0 degrees from normal. BTW, could it be that the driver does centering in a suitable way already? I'm not sure if anything should be done about the direction mismatch. > Is the range of values you get out of the kernel clipped to [-64,64], > or is evdev doing the clipping? I'm sorry, I didn't collect enough data. I've collected more input, which pushed the range to [-67,73]. > If the hardware is only sending data in that range there isn't much you > can do about it (aside from double-checking that you're reading the tilt > packets correctly). Well, I could still scale it. > Otherwise, I'd make sure the range you tell the kernel to expect is > correct -- I hope evdev doesn't have [-64,64] hardcoded, but we do the > same thing so.... It seems evdev doesn't do any special processing to the tilt axis, which I think is good. Now, aside from the strange range, the value change rate really looks more like the angle sine, than the angle itself. I.e. it seems the tablet reports fractions of the pen projection length. I did some measurements with a protractor and the values doesn't change at the same rate with angles. Even more, the negative values, seem to be offset somewhat and don't match the scale of the positive values closely, but they're still closer to a sine, than to an angle. This actually makes sense, since the tablet has no way to measure the angle directly and calculating it isn't a very suitable task for its CPU. > I usually use gimp for real-world testing, though I think you'll only > get tilt-sensitivity in gimp-2.7 (which is an unstable release...) I never used another tablet with tilt sensitivity with gimp. Could you suggest what output I should expect and with which tools? Thanks a lot for your comments and suggestions! Sincerely, Nick ------------------------------------------------------------------------------ Virtualization & Cloud Management Using Capacity Planning Cloud computing makes use of virtualization - but cloud computing also focuses on allowing computing to be delivered as a service. http://www.accelacomm.com/jaw/sfnl/114/51521223/ _______________________________________________ Linuxwacom-devel mailing list Linuxwacom-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linuxwacom-devel
