Asus EeePC T91MT multitouch touchscreen [Lucid 10.04]

[Ozone77]

Thanks everyone for your work on this, trying to get my touchscreen working on UNR 10.04 and after following the most recent instructions here am getting a response from the touch but it always puts the cursor in the top left corner.
Running the egalax calibration registers the first touch but after that it ignores input and times out.
Any ideas?

[jtjs]

Thanks everyone for your work on this, trying to get my touchscreen working on UNR 10.04 and after following the most recent instructions here am getting a response from the touch but it always puts the cursor in the top left corner.
Running the egalax calibration registers the first touch but after that it ignores input and times out.
Any ideas?

It seems that you missed a step, you need to add the multitouch quirk:

Edit /usr/src/multitouch-1.5/drivers/hid/usbhid/hid-quirks.c
Add the following line after "} hid_blacklist[] = {"

{ USB_VENDOR_ID_ASUS, USB_DEVICE_ID_ASUS_T91MT, HID_QUIRK_MULTI_INPUT },

[Ozone77]

It seems that you missed a step, you need to add the multitouch quirk: ...

Ah, right you are. I made a typo in that line. Thanks!

[jtjs]

Ok, it's working great for me here too.

No udev rules, hal fdi, /etc/modules or xorg.conf files are needed.

The only problem was that its line 1554 not 1556 as mentioned in jtjs's post.

1554 (correct device for me):

//{ HID_USB_DEVICE(USB_VENDOR_ID_ASUS, USB_DEVICE_ID_ASUS_T91MT)},

1556:

//{ HID_USB_DEVICE(USB_VENDOR_ID_ASUSTEK, USB_DEVICE_ID_ASUSTEK_LCM2)},

Was this just a typo or are we possibly dealing with 2 different models of touchscreen?

Also doesn't seem to be using the mosart module, but whatever works...

hid-mosart doesn't get used, seems quite strange to me, you would think that this should be used?

Well, maybe it's a mistake? Looking through hid-core.c, the t91mt was added to the ignore list originally, that was done by Stephane Chatty, but that excludes the module from getting loaded at all...

There is another list in hid-core.c(about 200 lines above the ignore list), one for devices that do have a specialized driver. I added the entry for the t91mt to that list, now the hid-mosart module does get loaded(before usbhid, the driver that would be used by default. The touch screen continues to function as before, but presumably through the hid-mosart driver?

I am not familiar with hiddev though, so what I am doing is mostly guessing.

One note though, MyPaint now works, it didn't work for me before. (note, I have adjusted the pressure sensitivity mapping.. I only get one level of pressure)

H3g3m0n: do you have any thoughts?

Update:
I have also tried the above with the multi input quirk disabled.. It gives me just one touchscreen device instead of two, and the calibration is.. Odd.. The pointer doesn't sit in the top left corner anymore, but I wouldn't recommend using it.. atleast not until it can be calibrated, and without the multi input quirk enabled, that calibrate script does not work.

Update 2: I don't think calibration would work on this.. It definitely seems like dragging a mouse around the screen rather than using a touchscreen. It seems that the co-ordinates are all relative, nothing that relates to any specific spot on the screen. Perhaps this is why we're all waiting for xi2?

Update 3: Using the mypaint that is modified to work with resistive touch screens(from the t101mt thread), the pressure mapping seems to be taken from the x-axis.. it is really quite funny to use...

[H3g3m0n]

hid-mosart doesn't get used, seems quite strange to me, you would think that this should be used?

Well, maybe it's a mistake? Looking through hid-core.c, the t91mt was added to the ignore list originally, that was done by Stephane Chatty, but that excludes the module from getting loaded at all...

There is another list in hid-core.c(about 200 lines above the ignore list), one for devices that do have a specialized driver. I added the entry for the t91mt to that list, now the hid-mosart module does get loaded(before usbhid, the driver that would be used by default. The touch screen continues to function as before, but presumably through the hid-mosart driver?

I am not familiar with hiddev though, so what I am doing is mostly guessing.

One note though, MyPaint now works, it didn't work for me before. (note, I have adjusted the pressure sensitivity mapping.. I only get one level of pressure)

H3g3m0n: do you have any thoughts?

Update:
I have also tried the above with the multi input quirk disabled.. It gives me just one touchscreen device instead of two, and the calibration is.. Odd.. The pointer doesn't sit in the top left corner anymore, but I wouldn't recommend using it.. atleast not until it can be calibrated, and without the multi input quirk enabled, that calibrate script does not work.

Update 2: I don't think calibration would work on this.. It definitely seems like dragging a mouse around the screen rather than using a touchscreen. It seems that the co-ordinates are all relative, nothing that relates to any specific spot on the screen. Perhaps this is why we're all waiting for xi2?

Update 3: Using the mypaint that is modified to work with resistive touch screens(from the t101mt thread), the pressure mapping seems to be taken from the x-axis.. it is really quite funny to use...

From what I can see the EGALAX touchscreen has the HID_QUIRK_MULTI_INPUT enabled so it should be the same for the mosart since they are both from ENAC.

Give GIMP a go, it has pressure sensitivity and allows you to specify which axis is used for the pressure. See how many axis are listed Not sure if it will allow multi-input though :/ I think you need to set the virtual X input thing to disabled first.
Edit>Prefrences>Input>Extended input

For me using the non-mosart drivers, Gimp lists 2 AsusTek, Inc. MultiTouch devices. The first has up to 8 numbers as choices for the input axis but doesn't seem to respond to any input. The second only has 2 number but seems to respond but it won't drag for me (it just moves the cursor to the position touched then doesn't do anything).

I also note that in /dev/input have event6, event7 and event8 all listed as Asus multitouch devices in lsinput (from the input-utils package). Although only the event7 responds to touchscreen events when 'cat'ing the output directly. It might be worth checking if you have the same number with mosart and if it outputs to any of the others.

Could be worth chmod 666ing the /dev/input/event* nodes to ensure there aren't any permissions problems. Not sure if these programs read from the directly or not.

Could do with finding some program capable to reading the input from the dev input/hid nodes directly to check for multitouch event input support. I do recall seeing one posted a while ago but have no idea where now.

Also worth trying twofing again and seeing if that works now.

As for the XI2, if its working with egalax, it should be working with masart as they are basically the same drivers. Not sure why calibration is screwed. Maybe it's combining multiple devices into 1 input, so your getting the result of both the single touch and the multitouch nodes. In that case we would just need to override with an xorg.conf.

If calibration is needed, it might be worth trying the manual xinput set-prop calibration.
https://help.ubuntu.com/community/T101MT#TOUCH SCREEN

I'll have a go with the mosart drivers myself later (but it probably won't be until tomorrow now).

[jtjs]

I gave twofing a whirl again yesterday(Note: you have to modify the udev rule to use the device at "0185" rather than the device id that is already specified).

Once calibration has been done, it loads, it sees the different windows on the screen, but it doesn't seem to understand the output from the t91mt's screen.

If I have time I may look at the source and possibly add some more debugging output, that is unless the things I want to add are already there..

I also tried gimp, but no luck so far.

Another thing to note, the other two asus devices listed after the t91mt in hid-core.c, those were also added by Stephane Chatty at the same time as the t91mt was added.

Also, if you look at the source to the hid-mosart driver, it was originally set to call the multi input quirk.. It seems it didn't give the desired results, and only caused problems with xi2..

Update:
Looking through twofingemu.c.. it grabs the first device that it finds in the xinput list.. However, as it stands, with the mosart driver loaded, I have 3 devices showing up in the xinput list.. I think the egalax folks only have one showing up. This might be causing the issues, I'll look into it more tonight.

[jtjs]

Looking at hid-mosart.c.. it doesn't emit any details about the pressure of the touch, that would explain the lack of input for that. What is odd is that the x axis of the input is being mapped by programs which listen for the sensitivity..

[thomi_ch]

Hello..

Have done all steps.. the first time with previous Kernel it worked.. but now with 2.6.32-25 i can't really use the touchscreen of my T91MT with lucid 10.04.1. It's like mirrored and other strange thing. Have tried to calibrate it, but no way...

Any idea?

Thanks for response
thomi

[H3g3m0n]

It's been working fine for me with 10.04.1.

Maybe you need to play around with the Axis inversion stuff:

xinput set-int-prop $INPUTDEV "Evdev Axis Inversion" 8 0 0

Change the 0's at the end to 1's depending on which axis you wan't to invert.

Otherwise give it another go from scratch.

[thomi_ch]

Hey

Have found the problem.. i had 9.10 before,.. and there evtouch was used for the touchscreen of the T91MT... A fdi file was available and the xserver-sorg-input-evtouch was installed. Have removed the fdi file and uninstalled the evtouch driver and rebooted...

Now, the axis are correct, but the calibration not. On top of the screen calibration is better but on middle and bottom of the screen the cursor is out of calibration. Have tried to calibrate it, but without success.

How can i check, which driver is used? Which input dev is used?

And how to fix that calibration issue?

Thanks for feedback
thomi