Severity of VCPIOError errors

[mciupak]

Due to high level of errors severity it is impossible to use monitorcontrol (VCPIOError exception thrown) with monitors which do not follow the DDC/CI standard e.g. Philips 223V monitors returns corrupted data i.e. incorrect checksum and/or wrong length, despite that it is possible to control it.

Example output:
Reading 0x60...
Send: 51 82 01 60 dc | Q...
Recv: 6e | n
Invalid response, corrupted data - xor is 0x5e, length 0x08
6e 88 88 02 00 60 00 00 03 00 01 | n.........
Send: 51 82 01 60 dc | Q...
Recv: 6e | n
Invalid response, corrupted data - xor is 0x5e, length 0x08
6e 88 88 02 00 60 00 00 03 00 01 | n.........
Send: 51 82 01 60 dc | Q...
Recv: 6e | n
Invalid response, corrupted data - xor is 0x5e, length 0x08
6e 88 88 02 00 60 00 00 03 00 01 | n.........

If possible, it could be better to handle errors without throwing an exception and try to continue execution.

[newAM]

I have seen this behavior once before on a different monitor, what helped for that was increasing the timeout beyond the DDC-CI minimum values.

Can you try increasing these constants to see if it changes the behavior?

monitorcontrol.vcp.vcp_linux.LinuxVCP.GET_VCP_TIMEOUT = 0.2  # default 0.04
monitorcontrol.vcp.vcp_linux.LinuxVCP.CMD_RATE = 0.2 # default 0.05

[mciupak]

Unfortunately, it does not help:

File "/root/.local/lib/python3.7/site-packages/monitorcontrol/vcp/vcp_linux.py", line 178, in get_vcp_feature
raise VCPIOError(f"checksum does not match: {checksum_xor}")
monitorcontrol.vcp.vcp_abc.VCPIOError: checksum does not match: 80

[newAM]

Huh, interesting! This is the first monitor that I have heard of that violates the spec then.

I would like to keep the exceptions as the default behavior since they usually mean that the rest of the command will fail.
These exceptions are buried rather deep as you noted. I would like to ideally add some way for the user to handle the exceptions better, but I don't see an elegant way forward with that.

I am thinking of a static class variable to disable exceptions; something along the lines of:

monitorcontrol.vcp.vcp_linux.LinuxVCP.BYPASS_CHECKSUM: bool = True
# and others for the various locations where VCPIOError can be raised

Would that work for your application?

[mciupak]

Bypassing only checksum won't work as most probably remaining code will fail as well. I saw this already - not only checksum was incorrect but expected data was too short thus unpacking failed and it was not caught.

For my application it would be better to pass some argument during runtime down to vcp to bypass checksum and following operations and try to recover and/or use some predefined default values i.e. I have config file in which I can specify e.g. which bus number should have it bypassed and remaining one(s) should not.

As I have physical access to this monitor I can prototype and propose something. What do you think?

[newAM]

As I have physical access to this monitor I can prototype and propose something. What do you think?

That would be great! 👍 It is hard to develop for something that does not meet the spec.

[jiri-one]

I have latest Arch Linux, ddcutil works great, but I would like to use Python implementation, because I plan gui and small hw which will be set backlight value from ambient light (I finished HW, next part is SW in Python). But unfortunately latest monitorcontrol said:

monitorcontrol --get-luminance
Traceback (most recent call last):
  File "/usr/bin/monitorcontrol", line 8, in <module>
    sys.exit(main())
  File "/usr/lib/python3.8/site-packages/monitorcontrol/__main__.py", line 70, in main
    luminance = monitor.get_luminance()
  File "/usr/lib/python3.8/site-packages/monitorcontrol/monitorcontrol.py", line 180, in get_luminance
    return self._get_vcp_feature(code)
  File "/usr/lib/python3.8/site-packages/monitorcontrol/monitorcontrol.py", line 157, in _get_vcp_feature
    current, maximum = self.vcp.get_vcp_feature(code.value)
  File "/usr/lib/python3.8/site-packages/monitorcontrol/vcp/vcp_linux.py", line 178, in get_vcp_feature
    raise VCPIOError(f"checksum does not match: {checksum_xor}")
monitorcontrol.vcp.vcp_abc.VCPIOError: checksum does not match: 115

But the problem is not in monitor (it is HPE243i), because on Windows
monitorcontrol is working great :).
On Arch Linux I followed this https://monitorcontrol.readthedocs.io/en/latest/linux_setup.html and it is working for ddcutil too, but for monitorcontrol there is checksum bug.

[newAM]

...huh, if two people have this issue then this likely is a bug.

I'll add something this weekend to optionally disable the checksum.

@jiri-one does you monitor also work aside from the bad checksum?

[jiri-one]

On Win10 I tested only get and set luminance and it worked great. But on Archlinux everything ended with bad checksum.

[sebastianmaisel]

I also always get the VCPIOError which says: checksum does not match: 80, but I didn't get the display run on my Ubuntu properly with other tools while it works perfect on macOS with ddcutil.
There may also be a driver issue.

[newAM]

I pushed a 2.2.0 release which disables the checksums by default.

https://monitorcontrol.readthedocs.io/en/latest/api.html#checksum-behaviour

This does not fully address this issue since there were also problems with the payload unpacking mentioned here.

Let me know if this solves any of the problems you guys are having (or what new problems you now have).

[jiri-one]

I tested new version on Archlinux and it is not working.
https://pastebin.com/xNYnmCP1
And the last version is working nice on Windows .... with same monitor.

[sebastianmaisel]

Exact same issue at my side... :-(

…

Am 05.10.2020 um 07:00 schrieb jiri-one ***@***.***>:  I tested new version on Archlinux and it is not working. https://pastebin.com/xNYnmCP1 — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

[jiri-one]

@newAM Maybe you want to make some tests on my monitor? If you contact me directly on email jiri spot one in domain pm spot me (I hope I confused the spambots enough :)) then I can redirect some port on my router to my ArchLinux PC, and you can connect to it over SSH. Just if you want, it is just idea.