Hi Derek,
On 19.10.2015 22:29, Derek Foreman wrote:
On 19/10/15 10:32 AM, Leon Anavi wrote:
Hi Derek,
On 19.10.2015 17:16, Derek Foreman wrote:
On 19/10/15 04:44 AM, Leon Anavi wrote:
Hi Derek,
On 14.10.2015 23:37, Leon Anavi wrote:
Hi Derek,
On 14.10.2015 20:11, Derek Foreman wrote:
On 14/10/15 05:31 AM, Leon Anavi wrote:
Hi Derek, All,
Derek, I am experiencing similar issues with weston-keyboard as in
your
article "When is a Keyboard Not a Keyboard?" which has been
published at
Samsung OSG blog.
Could you please provide me help how to get Weston 1.6 working
with the
virtual weston-keyboard in the HTML5 apps from the AGL app suite when
hardware keyboard is not present?
Looks like a bug I fixed last December, actually:
http://cgit.freedesktop.org/wayland/weston/commit/?id=f0aaa417
Fairly small so it might be an easy forward port...
Thank you for the provided feedback.
I applied a patch with this bug fix to Weston 1.6 but I am still
experiencing the same issue (as described in my previous email). It is
a bit late here so I will continue debugging it tomorrow.
After further exploration of various log files I noticed that the issues
while running AGL app suite on Tizen with Crosswalk, Wayland and Weston
1.6 do not appear if an event from /dev/input is detected as a keyboard.
As a work around a colleague from Konsulko Group defined a GPIO as a key
at the device tree. This way the GPIO is detected by Weston as a
keyboard:
That's quite clever, and I'll try to remember it. :)
Yes, thank you, I am lucky to work with experienced kernel developers at
Konsulko Group :)
Last time someone asked me for a way to create fake keyboard presence, I
suggested writing a program with uinput - but if you've got control over
your device tree your approach is much simpler.
Yes, uinput was next on my list for a work around.
[00:00:08.831] input device 'gpio-keys.22', /dev/input/event1 is a
keyboard
The configuration of Weston is:
[input-method]
path=/usr/libexec/weston-keyboard
[keyboard]
keymap_layout=us
Derek, in this case we are using the behavior described in your article
"When is a Keyboard Not a Keyboard" for the work around. Please let me
know how can I help you find and fix the root cause of the problem as it
seems to be a Weston issue.
It's definitely a weston issue and not one of configuration, but it's
hard to tell if it's already been fixed upstream.
Do you have (or can you easily build) weston-editor for your platform?
It's part of the weston repository... If you can reproduce the crash
with that then I can reproduce it on my desktop here and fix it promptly.
I've tried weston-editor on my rpi2 and it will crash when no real
keyboard is present. The virtual keyboard pops up, but switching
between numeric and alphabetic entry crashes. This is the bug I gave
you the patch for previously, so I think whatever you've found can't be
triggered with weston-editor.
This afternoon I also included weston-editor in my image with Weston
1.6. Exactly as you described, it crashed when a real keyboard is not
present (of course, for the test I used a device tree without a fake
keyboard).
After that I applied the patch that you pointed out from the upstream
and with it weston-editor is working fine. Furthermore, the Browser from
AGL application suite is also working fine. So obviously it is the same
problem in both cases and the patch solves it. Somehow last week I did
not apply it properly. Thank you for the help.
A backtrace of weston when it fails would also be quite helpful.
Is it possible to build an image that can reproduce this problem that
runs on an intel machine? I have an old laptop I could re-image (I can
knock the keyboard device registration bits out of weston to mimic lack
of keyboard presence easily enough...)
Yes, I can build an image for Intel machines. I am not sure if the bug
will appear on them. In general it should be possible to boot this Tizen
image from USB stick. Btw, based on my experience Intel NUC and
MinnowBoard MAX are useful for testing Tizen 3 on Intel. I will give it
a try.
I have a NUC here - it probably won't crash because the acpi power
button will be enough of a keyboard to mask the bug.
Yes, you are right. I did a test on Intel NUC with Tizen:Common (built
using Yocto/OE). Weston 1.6 detects the power and the sleep buttons as
keyboards:
[20:07:59.343] input device 'Power Button', /dev/input/event0 is a keyboard
[20:07:59.344] input device 'Sleep Button', /dev/input/event1 is a keyboard
Best regards,
Leon
Weston can be
tricked into ignoring keyboards if you comment out the
LIBINPUT_DEVICE_CAP_KEYBOARD clause in evdev_device_create() in
libinput-device.c
hitting a key on a real keyboard will cause a crash, but it should allow
mimicking a keyboardless situation on an intel device.
If you can supply an image or help me through the build process to get
an intel image I can reproduce this on, I can probably get it sorted
fairly quickly.
I also have a beaglebone black and an rpi2 on hand if either of those
could reproduce it, but the rpi2 tizen port has some crosswalk issues
that will likely make this untestable.
Yes, there are some issues with Crosswalk on Raspberry Pi 2 therefore
you cannot reproduce the issue on it.
The environment in my case is a bit more complicated because I am
working on i.MX6 devices and I had to apply some patches on Wayland /
Weston 1.6 to get it working with hardware graphics acceleration.
You should be able to reproduce the bug on HummingBoard if you had one
and if you disable gpio_ir_recv (cause Weston detects it as a keyboard):
https://wiki.tizen.org/wiki/HummingBoard#Building_Tizen:Common_with_Yocto
Unfortunately I don't have one of those.
Best regards,
Leon
Best regards,
Leon
Thanks,
Leon
Hope that helps a bit,
Derek
In my case Weston keeps crashing when I type in HTML5 input fields
from
the AGL apps through the virtual weston-keyboard if no hardware
keyboard
is attached. I am working on a Tizen image that contains Wayland and
Weston 1.6, Crosswalk 11.40.277.0 and libinput 0.6.0. Please have a
look
at my test cases for more details:
* Test case 1, weston.ini:
[input-method]
path=/usr/libexec/weston-keyboard
[keyboard]
keymap_layout=wkb
Results:
1.1. No hardware keyboard: FAIL
Weston crashes when the user tries to launch Browser from the
Homescreen
app:
[00:00:32.676] caught signal: 11
1.2. With hardware keyboard: FAIL
Weston detects the hardware keyboard:
[00:00:07.480] input device 'SEM USB Keyboard', /dev/input/event0
is a
keyboard
[00:00:07.481] input device 'SEM USB Keyboard', /dev/input/event1
is a
keyboard
Weston crashes when the user tries to launch Browser from the
Homescreen
app:
[00:00:49.925] caught signal: 11
* Test case 2, weston.ini:
[input-method]
path=
[keyboard]
keymap_layout=wkb
Results:
2.1. No hardware keyboard: OK with remark - there is no virtual
keyboard
(as expected) so there is no way to type.
2.2. With hardware keyboard: FAIL
Weston detects the hardware keyboard:
[00:00:06.586] input device 'SEM USB Keyboard', /dev/input/event0
is a
keyboard
[00:00:06.587] input device 'SEM USB Keyboard', /dev/input/event1
is a
keyboard
Weston fails as soon as the user starts to type in HTML5 input field
for
URL at the Browser app through the hardware keyboard:
[00:00:48.006] caught signal: 11
* Test case 3, weston.ini:
[input-method]
path=/usr/libexec/weston-keyboard
[keyboard]
keymap_layout=us
Results:
3.1. No hardware keyboard: FAIL
Weston crashes when the user tries to launch Browser from the
Homescreen
app:
[00:00:28.391] caught signal: 11
3.2. With hardware keyboard: OK
Weston detects the hardware keyboard and both the hardware and
virtual
weston-keyboard work fine:
[00:00:06.767] input device 'SEM USB Keyboard', /dev/input/event0
is a
keyboard
[00:00:06.771] input device 'SEM USB Keyboard', /dev/input/event1
is a
keyboard
I want to get weston-keyboard working fine in Weston even if there
is no
hardware keyboard.
Thanks,
Leon
--
Leon Anavi
Software Engineer
konsulko.com
_______________________________________________
Dev mailing list
[email protected]
https://lists.tizen.org/listinfo/dev
