Unexpected Application Error in 3.7.1-dev

[tp1de]

Loading new dev versions of 3.7.1 I get the following error while selecting devices / customizations:

This is using Firefox

Using Edge it looks a bit different:

[tp1de]

When re-starting the device-list works for 1-2 minutes. Afterwards the error occurs.

[tp1de]

This error co-exists with api errors on .../api/system. API command works for 1-2 minutes afterwards not anymore.
Same as for web-ui errors on devices.

The error occurs on 3.7.1-dev.1 onwards.
@proddy Please have a look at api response for mixer mm100. This might be the cause of errors.

API.txt

[MichaelDvP]

Seems to be the name of the MM100 from telegram 1, can you give us the output of read 21 1

[tp1de]

ems-esp:$ read 21 1
000+00:03:10.932 N 3: [emsesp] mixer(0x21) -W-> me(0x0B), ?(0x01), data: 32 35 33 30 38 35 34 30 30 31 37 36 39 37 37 33 38 31 31 30 31 31 34 00 FF FF FF
000+00:03:11.001 N 4: [emsesp] mixer(0x21) -W-> me(0x0B), ?(0x01), data: 4D 04 1F 1E 1E 00 00 00 00 00 00 00 00 48 00 30 (offset 27)
ems-esp:$

[MichaelDvP]

Seems offset 27 is not always a null terminated ascii name.

[MichaelDvP]

Try to set customizations custom name for MM100.

[tp1de]

Try to set customizations custom name for MM100.

Yes this works.

[proddy]

seems so, well spot Michael. I hate that Bosch is so inconsistent. Another user (in #2073) with an MM100 shows the text "MM100" in the telegram 0x01 at offset 27. I haven't been able to decipher the logic of that telegram yet.

[MichaelDvP]

@tp1de Please update to my testbuild and check if it works without custom name. (I added a check for printable characters).
Also the thermostat timer-json should work again.

[tp1de]

yes it works with new testbuild but name is now M/M100 which does not make a lot of sense.
(During startup name is MM100 (old version) and afterwards replaced by M/MM100)

... and I recognized that Dashboard does not show entities anymore (maybe I switched too often between Versions)

[tp1de]

Update: Dashboard seems to be related to FireFox / Browser Update. Edge / Chrome shows the dashboard correctly

[proddy]

Just downloaded and installed Firefox 132.0, and Dashboard works

[MichaelDvP]

I always use firefox and have seen something like that often. It's a cache issue and ctrl-R sometimes only helps short. Clear website data and cache for this device.

[MichaelDvP]

yes it works with new testbuild but name is now M/M100 which does not make a lot of sense.

This is what is written in your MM100 telegram 01: only a singel M, then EOT marker (ascii 04). Maybe we should ignore the 01-telegram name if it is not a null-terminated string? On the other hand this gives info about different versions of device, some users have communication issues special with MM100 and emsesp, for others it works.

[tp1de]

I always use firefox and have seen something like that often. It's a cache issue and ctrl-R sometimes only helps short. Clear website data and cache for this device.

I was running the debugger (F12) and I have seen that there are http get commands every second.
Within reply there are JSON-errors as well. I do not know if this has an impact.

[tp1de]

This is what is written in your MM100 telegram 01: only a singel M, then EOT marker (ascii 04). Maybe we should ignore the 01-telegram name if it is not a null-terminated string? On the other hand this gives info about different versions of device, some users have communication issues special with MM100 and emsesp, for others it works.

It is up to you. I haven't really understood why you intend to read the name out of the telegram.

Within Bosch-Group products there are many similar products with different names and firmware releases.
I believe tat the productID is sometimes the same for different brands - just the product sticker changes.
Since the firmware is not flashable / upgradeable there are a lot of active versions existing.

[proddy]

I always use firefox and have seen something like that often. It's a cache issue and ctrl-R sometimes only helps short. Clear website data and cache for this device.

I was running the debugger (F12) and I have seen that there are http get commands every second. Within reply there are JSON-errors as well. I do not know if this has an impact.

Can you post some screenshots?

[tp1de]

[proddy]

thanks, the page should be refreshing every 3 seconds. Maybe that is too much. I also see the

content-type in the header is set to application/json which is incorrect as we're using binary compressed with msgpack for the response payload. I'll fix that.

[MichaelDvP]

I've checked the json error, it affects all msgpack sendings since a very long time and mny emsesp versions (devcies->deviceData, Customization).
Setting in AsyncJson.h: AsyncJsonResponse _contentType = isMsgPack ? MSGPACK_MIMETYPE : JSON_MIMETYPE;
with mimetype application/msgpack solves this. But appart from the error message in analyzter i can't see issues in webpage.
Also the 3 sec refresh works well with my really slow laptop, also while compiling emsesp in the background.

[proddy]

yes, it's harmless but I changed it anyway. I had already implemented it the same way you suggested before seeing your comment. Great minds think alike !

[proddy]

yes it works with new testbuild but name is now M/M100 which does not make a lot of sense.

This is what is written in your MM100 telegram 01: only a singel M, then EOT marker (ascii 04). Maybe we should ignore the 01-telegram name if it is not a null-terminated string? On the other hand this gives info about different versions of device, some users have communication issues special with MM100 and emsesp, for others it works.

We could read the bytes from offset 27 until we hit a non-ASCII value, and then null-terminate the string. And ignore anything that is less than 3 characters long (like just "M").

[MichaelDvP]

We could read the bytes from offset 27 until we hit a non-ASCII value, and then null-terminate the string. And ignore anything that is less than 3 characters long (like just "M").

Done. In my dev2-testbuild.

[tp1de]

This solution works good

[tp1de]

@MichaelDvP something wrong on your testbuild with mqtt:

[MichaelDvP]

Can not reproduce and have not changed anything in mqtt code. Collision with a second mqtt client?
Have you tried restarting mqtt broker?

[proddy]

using the same MQTT clientID perhaps?

[tp1de]

While using dev.3 from @proddy mqtt works - just the devicename is not implemented.
when using test.3 from @MichaelDvP mqtt disconnects and restarts every 3 seconds and repeats the last mqtt commands every time. Mqtt settings are kept.

[proddy]

on the same hardware or are you running multiple instances?

[tp1de]

same Hardware and just updating ems-esp firmware. Re-start of mqtt broker has no impact. There must be a difference in Michaels coding.

[MichaelDvP]

nd repeats the last mqtt commands every time.

The mqtt code is not changed, i've updated/merged latest changes to dev/dev2 and compared.
What commands do you see? Maybe it's a change in commands?

[MichaelDvP]

@tp1de The mqtt in my testbuild should work now, Was in HA config for the timer values, the default case results in a scrambled topic for the timer.

[tp1de]

@tp1de The mqtt in my testbuild should work now, Was in HA config for the timer values, the default case results in a scrambled topic for the timer.

Yes I can confirm it is working now.
I just added #2179 for a bug on ems-esp start. This is a general topic - not just for your testbuild.

[proddy]

great, we can close this one now then.