devi-elo problem

QNX Newsgroups (Archives) qdn.public.ddk.input
1

Hi,

I am evaluating QNX6 in an Advantech PPC-150T industrial PC that functions
perfectly under QNX4. I am informed that the touch screen controller is an
Elo TouchSystems E271-2202 and has no problems with QNX4 ‘Input smartset
uart/fd …’.The controller is hard wired to the COM4 serial port.

I have transferred the calibration data from the ‘absf’ file to the ‘acalib’
file, which is

0x0:1023x767: 3578 536 667 3419 0

I can run devi-elo using the uart or fd options and the cursor responds to
my touches, but there appears to be an error in the translation of the touch
screen → graphics co-ordinates that has all the hallmarks of a
signed/unsigned problem or integer overflow.

If I probe around I can find areas where the cursor aligns to my finger. If
I then drag my finger horizontally across the screen the cursor will follow
for a short distance. The cursor then goes over to the right hand side of
the screen and stays there as I continue to drag, and then it shoots back to
meet my finger which it then follows for a short distance…

There is a similar effect in the vertical direction. In this case the cursor
shoots up to the top periodically.

Before I start hacking the published code to try and investigate further, I
thought I would ask if anyone knows anything about this problem. Also, does
the code match the issue of drivers currently in Patch B?

Any help will be much appreciated.

Jim Douglas

2

I’m pretty sure that there is no problem in code. To clarify situation,
could you please start devi-elo in debug print mode (use -v option for
that). In case you are not aware, the most convinient way to do this is to
start driver from other computer using telnet. Just touch four points: left
top, left bottom, right bottom and right top (exactly in this order). Copy
screen output (there should be messages “Raw coords…” and “XFormed
coords…”) into reply to this message. It just lets me to estimate a
problem more precisely.


Thanks,

Alex Chapiro (R&D)
QNX Software Systems, Ltd.

Jim Douglas wrote in message <99486c$t5n$1@inn.qnx.com>…

Hi,

I am evaluating QNX6 in an Advantech PPC-150T industrial PC that functions
perfectly under QNX4. I am informed that the touch screen controller is an
Elo TouchSystems E271-2202 and has no problems with QNX4 ‘Input smartset
uart/fd …’.The controller is hard wired to the COM4 serial port.

I have transferred the calibration data from the ‘absf’ file to the
‘acalib’
file, which is

0x0:1023x767: 3578 536 667 3419 0

I can run devi-elo using the uart or fd options and the cursor responds to
my touches, but there appears to be an error in the translation of the
touch
screen → graphics co-ordinates that has all the hallmarks of a
signed/unsigned problem or integer overflow.

If I probe around I can find areas where the cursor aligns to my finger. If
I then drag my finger horizontally across the screen the cursor will follow
for a short distance. The cursor then goes over to the right hand side of
the screen and stays there as I continue to drag, and then it shoots back
to
meet my finger which it then follows for a short distance…

There is a similar effect in the vertical direction. In this case the
cursor
shoots up to the top periodically.

Before I start hacking the published code to try and investigate further, I
thought I would ask if anyone knows anything about this problem. Also, does
the code match the issue of drivers currently in Patch B?

Any help will be much appreciated.

Jim Douglas

3

As requested…

Raw coords: (-61, -114)
Xformed coords: (1023, 0)
Raw coords: (-58, -117)
Xformed coords: (1023, 0)
Raw coords: (-59, -114)
Xformed coords: (1023, 0)
Raw coords: (-59, -114)
Xformed coords: (1023, 0)
Raw coords: (-41, 3332)
Xformed coords: (1023, 742)
Raw coords: (-35, 3346)
Xformed coords: (1023, 746)
Raw coords: (-32, 3347)
Xformed coords: (1023, 746)
Raw coords: (-28, 3354)
Xformed coords: (1023, 748)
Raw coords: (566, -22)
Xformed coords: (1012, 0)
Raw coords: (566, -24)
Xformed coords: (1012, 0)
Raw coords: (563, -22)
Xformed coords: (1013, 0)
Raw coords: (567, -22)
Xformed coords: (1012, 0)
Raw coords: (580, -18)
Xformed coords: (1008, 0)
Raw coords: (580, -18)
Xformed coords: (1008, 0)
Raw coords: (534, -93)
Xformed coords: (1023, 0)
Raw coords: (537, -90)
Xformed coords: (1022, 0)
Raw coords: (535, -92)
Xformed coords: (1023, 0)
Raw coords: (538, -100)
Xformed coords: (1022, 0)
Raw coords: (526, -118)
Xformed coords: (1023, 0)
Raw coords: (518, -121)
Xformed coords: (1023, 0)

“Alex Chapiro” <achapiro@qnx.com> wrote in message
news:995kgn$mhu$1@nntp.qnx.com…

I’m pretty sure that there is no problem in code. To clarify situation,
could you please start devi-elo in debug print mode (use -v option for
that). In case you are not aware, the most convinient way to do this is to
start driver from other computer using telnet. Just touch four points:
left
top, left bottom, right bottom and right top (exactly in this order). Copy
screen output (there should be messages “Raw coords…” and “XFormed
coords…”) into reply to this message. It just lets me to estimate a
problem more precisely.


Thanks,

Alex Chapiro (R&D)
QNX Software Systems, Ltd.

Jim Douglas wrote in message <99486c$t5n$> 1@inn.qnx.com> >…
Hi,

I am evaluating QNX6 in an Advantech PPC-150T industrial PC that
functions
perfectly under QNX4. I am informed that the touch screen controller is
an
Elo TouchSystems E271-2202 and has no problems with QNX4 ‘Input smartset
uart/fd …’.The controller is hard wired to the COM4 serial port.

I have transferred the calibration data from the ‘absf’ file to the
‘acalib’
file, which is

0x0:1023x767: 3578 536 667 3419 0

I can run devi-elo using the uart or fd options and the cursor responds
to
my touches, but there appears to be an error in the translation of the
touch
screen → graphics co-ordinates that has all the hallmarks of a
signed/unsigned problem or integer overflow.

If I probe around I can find areas where the cursor aligns to my finger.
If
I then drag my finger horizontally across the screen the cursor will
follow
for a short distance. The cursor then goes over to the right hand side of
the screen and stays there as I continue to drag, and then it shoots back
to
meet my finger which it then follows for a short distance…

There is a similar effect in the vertical direction. In this case the
cursor
shoots up to the top periodically.

Before I start hacking the published code to try and investigate further,
I
thought I would ask if anyone knows anything about this problem. Also,
does
the code match the issue of drivers currently in Patch B?

Any help will be much appreciated.

Jim Douglas

\

4

This sounds like the eternal axis switching problem.
I suggest you try this. Edit your config file and
change the 0 at the end to 1. Restart the driver and
see what happens.

If this works, well let QSSL know that there is one more
user waiting for acalib to be fixed.


Previously, Jim Douglas wrote in qdn.public.ddk.input:

Hi,

I am evaluating QNX6 in an Advantech PPC-150T industrial PC that functions
perfectly under QNX4. I am informed that the touch screen controller is an
Elo TouchSystems E271-2202 and has no problems with QNX4 ‘Input smartset
uart/fd …’.The controller is hard wired to the COM4 serial port.

I have transferred the calibration data from the ‘absf’ file to the ‘acalib’
file, which is

0x0:1023x767: 3578 536 667 3419 0

I can run devi-elo using the uart or fd options and the cursor responds to
my touches, but there appears to be an error in the translation of the touch
screen → graphics co-ordinates that has all the hallmarks of a
signed/unsigned problem or integer overflow.

If I probe around I can find areas where the cursor aligns to my finger. If
I then drag my finger horizontally across the screen the cursor will follow
for a short distance. The cursor then goes over to the right hand side of
the screen and stays there as I continue to drag, and then it shoots back to
meet my finger which it then follows for a short distance…

There is a similar effect in the vertical direction. In this case the cursor
shoots up to the top periodically.

Before I start hacking the published code to try and investigate further, I
thought I would ask if anyone knows anything about this problem. Also, does
the code match the issue of drivers currently in Patch B?

Any help will be much appreciated.

Jim Douglas

\

–
Mitchell Schoenbrun --------- maschoen@pobox.com

5

Mitchell,

This sounds like the eternal axis switching problem.
I suggest you try this. Edit your config file and
change the 0 at the end to 1. Restart the driver and
see what happens.

Thanks. I didn’t change the axis switch because I assumed (bad thing) that
it would be the same as QNX4. Besides, the cursor will track correctly over
sections of the screen.

If this works, well let QSSL know that there is one more
user waiting for acalib to be fixed.

Don’t worry, myself and one or two others gave QSSL a very hard time in the
beta test group when we discovered that the touch screen drivers had been
released without the ‘calib’ (not ‘acalib’ anymore) utility. They have got
the message loud and clear.

Jim Douglas

Previously, Jim Douglas wrote in qdn.public.ddk.input:
Hi,

I am evaluating QNX6 in an Advantech PPC-150T industrial PC that
functions
perfectly under QNX4. I am informed that the touch screen controller is
an
Elo TouchSystems E271-2202 and has no problems with QNX4 ‘Input smartset
uart/fd …’.The controller is hard wired to the COM4 serial port.

I have transferred the calibration data from the ‘absf’ file to the
‘acalib’
file, which is

0x0:1023x767: 3578 536 667 3419 0

I can run devi-elo using the uart or fd options and the cursor responds
to
my touches, but there appears to be an error in the translation of the
touch
screen → graphics co-ordinates that has all the hallmarks of a
signed/unsigned problem or integer overflow.

If I probe around I can find areas where the cursor aligns to my finger.
If
I then drag my finger horizontally across the screen the cursor will
follow
for a short distance. The cursor then goes over to the right hand side
of
the screen and stays there as I continue to drag, and then it shoots
back to
meet my finger which it then follows for a short distance…

There is a similar effect in the vertical direction. In this case the
cursor
shoots up to the top periodically.

Before I start hacking the published code to try and investigate
further, I
thought I would ask if anyone knows anything about this problem. Also,
does
the code match the issue of drivers currently in Patch B?

Any help will be much appreciated.

Jim Douglas




\

Mitchell Schoenbrun --------- > maschoen@pobox.com

6

This data seams to be quite strange. Raw data cannot be negative! Are you
sure that you are using the last version of driver?

Actually, I think I know what’s that. Certanly this is an obsolete version.
Tomorrow morning I’ll answer you what to do.

Alex Chapiro (R&D)
QNX Software Systems, Ltd.

7

FYI it does seem somewhat old. The date on the file is 24th August 2000. I’m
unclear as to which version/patch introduced the touch screen drivers.

“Alex Chapiro” <achapiro@qnx.com> wrote in message
news:996507$264$1@nntp.qnx.com…

This data seams to be quite strange. Raw data cannot be negative! Are you
sure that you are using the last version of driver?

Actually, I think I know what’s that. Certanly this is an obsolete
version.
Tomorrow morning I’ll answer you what to do.

Alex Chapiro (R&D)
QNX Software Systems, Ltd.

\

8

Yes, I’m very positive now that you are working with old version. This bug
was fixed in Sep 2000. This is a problem of packaging. The only thing that I
can offer you now is to build yourself Elo input module from input DDK. A
question about the way of shipping the latest versions togeter with fixed
calib utility is now under discussion and I hope the problem will be solved
quite soon.

Sorry about that,

Alex Chapiro (R&D)
QNX Software Systems, Ltd.

9

The shipped devi-elo with patch B doesn’t work. But if you compile
the driver and use that instead it should work (at least it does for me)

  • Mario

“Alex Chapiro” <achapiro@qnx.com> wrote in message
news:997m33$svb$1@nntp.qnx.com…

Yes, I’m very positive now that you are working with old version. This bug
was fixed in Sep 2000. This is a problem of packaging. The only thing that
I
can offer you now is to build yourself Elo input module from input DDK. A
question about the way of shipping the latest versions togeter with fixed
calib utility is now under discussion and I hope the problem will be
solved
quite soon.

Sorry about that,

Alex Chapiro (R&D)
QNX Software Systems, Ltd.
\

10

Thanks to all.

I have picked up the new version of devi-elo from the beta site and it works
fine. ‘calib’ produces the correct numbers, but in my case it kills the
mouse and keyboard. Is that your responsibility Alex, or should I post the
problem to the beta newsgroup?

Jim Douglas

“Mario Charest” <mcharest@void_zinformatic.com> wrote in message
news:997n4d$ck$1@nntp.qnx.com…

The shipped devi-elo with patch B doesn’t work. But if you compile
the driver and use that instead it should work (at least it does for me)

  • Mario

“Alex Chapiro” <> achapiro@qnx.com> > wrote in message
news:997m33$svb$> 1@nntp.qnx.com> …
Yes, I’m very positive now that you are working with old version. This
bug
was fixed in Sep 2000. This is a problem of packaging. The only thing
that
I
can offer you now is to build yourself Elo input module from input DDK.
A
question about the way of shipping the latest versions togeter with
fixed
calib utility is now under discussion and I hope the problem will be
solved
quite soon.

Sorry about that,

Alex Chapiro (R&D)
QNX Software Systems, Ltd.


\

11

It is also an obsolet version of devi-hirun that you are running now. I
guess the version on the beta site should be newer (though it is not not the
newest one) and you shouldn’t experience this problem any more.

Alex Chapiro (R&D)
QNX Software Systems, Ltd.

Jim Douglas wrote in message <999bq6$7sf$1@inn.qnx.com>…

Thanks to all.

I have picked up the new version of devi-elo from the beta site and it
works
fine. ‘calib’ produces the correct numbers, but in my case it kills the
mouse and keyboard. Is that your responsibility Alex, or should I post the
problem to the beta newsgroup?

Jim Douglas

“Mario Charest” <mcharest@void_zinformatic.com> wrote in message
news:997n4d$ck$> 1@nntp.qnx.com> …

The shipped devi-elo with patch B doesn’t work. But if you compile
the driver and use that instead it should work (at least it does for me)

  • Mario

“Alex Chapiro” <> achapiro@qnx.com> > wrote in message
news:997m33$svb$> 1@nntp.qnx.com> …
Yes, I’m very positive now that you are working with old version. This
bug
was fixed in Sep 2000. This is a problem of packaging. The only thing
that
I
can offer you now is to build yourself Elo input module from input DDK.
A
question about the way of shipping the latest versions togeter with
fixed
calib utility is now under discussion and I hope the problem will be
solved
quite soon.

Sorry about that,

Alex Chapiro (R&D)
QNX Software Systems, Ltd.




\

12

Jim Douglas <jim@dramatec.co.uk> wrote:

Don’t worry, myself and one or two others gave QSSL a very hard time in the
beta test group when we discovered that the touch screen drivers had been
released without the ‘calib’ (not ‘acalib’ anymore) utility. They have got
the message loud and clear.

There was a package posted on the repository last week that contains a
matched set of devi- drivers and calib, including a devi-hirun
that doesn’t crash when calib is run.

13

Hello

Where can I find the package with the updated devi-* drivers
and calib that does not crash devi-hirun?

Michel Benoit