lirc serial_ir homebrew breakage

lirc serial_ir homebrew breakage

[tom]

Hi,

An upgrade from linux-5.9.16 to linux-5.10.4 results in lirc not working 
with my homebrew IR receiver.  Running mode2 and pressing the same 
button on the IR remote shows timing differences:

5.9.16 works, first few lines of output:

pulse 2676
space 371
pulse 1429
space 371
pulse 827
space 372
pulse 1426
space 372
pulse 1425
space 376
pulse 1424
space 374

5.10.4 fail
pulse 2670
space 380
pulse 1428
space 372
pulse 827
pulse 125391  <---
space 127128  <---
pulse 1406
space 394
pulse 1407
pulse 125394  <---
space 126726  <---
pulse 1400
space 398
pulse 1406
space 394
pulse 801

The reporting of two pulses in a row as well as the large time reported 
make it so the button presses are not recognized.

Is this a known issue?
I am a software developer and would be happy to help in any way.  I can 
gather more data or attempt to fix it myself.

Thanks, Tom Daley

Re: lirc serial_ir homebrew breakage

[Sean Young]

Hi Tom,

On Wed, Jan 06, 2021 at 09:16:31AM -0700, tom@tomdaley.org wrote:
> Hi,
> 
> An upgrade from linux-5.9.16 to linux-5.10.4 results in lirc not working
> with my homebrew IR receiver.  Running mode2 and pressing the same button on
> the IR remote shows timing differences:
> 
> 5.9.16 works, first few lines of output:
> 
> pulse 2676
> space 371
> pulse 1429
> space 371
> pulse 827
> space 372
> pulse 1426
> space 372
> pulse 1425
> space 376
> pulse 1424
> space 374
> 
> 5.10.4 fail
> pulse 2670
> space 380
> pulse 1428
> space 372
> pulse 827
> pulse 125391  <---
> space 127128  <---
> pulse 1406
> space 394
> pulse 1407
> pulse 125394  <---
> space 126726  <---
> pulse 1400
> space 398
> pulse 1406
> space 394
> pulse 801
> 
> The reporting of two pulses in a row as well as the large time reported make
> it so the button presses are not recognized.
> 
> Is this a known issue?

No, this is not known issue.

> I am a software developer and would be happy to help in any way.  I can
> gather more data or attempt to fix it myself.

Thanks for your bug report. In 5.10, there was a change which converted
the processing from nanoseconds to microsecondes, see:

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=528222d853f9283110f0118dd71d9f0ad686d586

One issue has already been found in this change, which hasn't been upstreamed
yet:

https://patchwork.linuxtv.org/project/linux-media/patch/20210105093023.5212-1-hias@horus.com/

Would you mind testing this patch out please?

Admittedly I'm not convinced that this is the problem though. I'm not sure
what is; the function frbwrite() contains two if statements to filter out
noise, maybe something is wrong there. You could try by commenting those
two if statements out and see if there is a problem there.

Your help is appreciated!

Thanks,

Sean

Re: lirc serial_ir homebrew breakage

[tom]

On 2021-01-07 03:25, Sean Young wrote:
> Hi Tom,
> 
> On Wed, Jan 06, 2021 at 09:16:31AM -0700, tom@tomdaley.org wrote:
>> Hi,
>> 
>> An upgrade from linux-5.9.16 to linux-5.10.4 results in lirc not 
>> working
>> with my homebrew IR receiver.  Running mode2 and pressing the same 
>> button on
>> the IR remote shows timing differences:
>> 
>> 5.9.16 works, first few lines of output:
>> 
>> pulse 2676
>> space 371
>> pulse 1429
>> space 371
>> pulse 827
>> space 372
>> pulse 1426
>> space 372
>> pulse 1425
>> space 376
>> pulse 1424
>> space 374
>> 
>> 5.10.4 fail
>> pulse 2670
>> space 380
>> pulse 1428
>> space 372
>> pulse 827
>> pulse 125391  <---
>> space 127128  <---
>> pulse 1406
>> space 394
>> pulse 1407
>> pulse 125394  <---
>> space 126726  <---
>> pulse 1400
>> space 398
>> pulse 1406
>> space 394
>> pulse 801
>> 
>> The reporting of two pulses in a row as well as the large time 
>> reported make
>> it so the button presses are not recognized.
>> 
>> Is this a known issue?
> 
> No, this is not known issue.
> 
>> I am a software developer and would be happy to help in any way.  I 
>> can
>> gather more data or attempt to fix it myself.
> 
> Thanks for your bug report. In 5.10, there was a change which converted
> the processing from nanoseconds to microsecondes, see:
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=528222d853f9283110f0118dd71d9f0ad686d586
> 
> One issue has already been found in this change, which hasn't been 
> upstreamed
> yet:
> 
> https://patchwork.linuxtv.org/project/linux-media/patch/20210105093023.5212-1-hias@horus.com/
> 
> Would you mind testing this patch out please?
> 
> Admittedly I'm not convinced that this is the problem though. I'm not 
> sure
> what is; the function frbwrite() contains two if statements to filter 
> out
> noise, maybe something is wrong there. You could try by commenting 
> those
> two if statements out and see if there is a problem there.
> 
> Your help is appreciated!
> 
> Thanks,
> 
> Sean

Hi Sean,

The patch fixes the problem I was seeing.  I have tested both receive 
and transmit on my homebrew IR, both are working.

Thanks, Tom

Re: lirc serial_ir homebrew breakage

[Sean Young]

On Thu, Jan 07, 2021 at 07:47:02AM -0700, tom@tomdaley.org wrote:
> On 2021-01-07 03:25, Sean Young wrote:
> > One issue has already been found in this change, which hasn't been
> > upstreamed
> > yet:
> > 
> > https://patchwork.linuxtv.org/project/linux-media/patch/20210105093023.5212-1-hias@horus.com/
> > 
> > Would you mind testing this patch out please?
> > 
> > Admittedly I'm not convinced that this is the problem though. I'm not
> > sure
> > what is; the function frbwrite() contains two if statements to filter
> > out
> > noise, maybe something is wrong there. You could try by commenting those
> > two if statements out and see if there is a problem there.
> > 
> > Your help is appreciated!
> > 
> > Thanks,
> > 
> > Sean
> 
> Hi Sean,
> 
> The patch fixes the problem I was seeing.  I have tested both receive and
> transmit on my homebrew IR, both are working.

Perfect, thank you very much. The patch should make its way into the stable
tree (at some point).


Sean