Error migrating calendar if connection_timeout is not set

[ccMatrix]

I've updated ics_calendar from latest 4 to 5 and had the following error in the log:

2024-09-11 07:33:12.463 ERROR (SyncWorker_3) [homeassistant.helpers.json] Failed to serialize to JSON: /config/.storage/core.config_entries. Bad data at $.data.entries[97].data.connection_timeout=<object object at 0xf6347e08>(<class 'object'>
2024-09-11 07:33:12.470 ERROR (MainThread) [homeassistant.helpers.storage] Error writing config for core.config_entries: Failed to serialize to JSON: /config/.storage/core.config_entries. Bad data at $.data.entries[97].data.connection_timeout=<object object at 0xf6347e08>(<class 'object'>

I did not have connection_timeout set in the yaml configuration. After I saw the error I added a line connection_timeout: 15 to the yaml and restarted home assistant which imported the yaml configuration into the core.config_entries file properly.

[Muenchen-Michi]

I had the same error, but did not recognize, that it's a problem with a missing settings in the old configuration, so I returned to version 4.2.0.

But this error is critical, because it stops HA from saving any new configuration. I added new integrations, that were all gone after a restart of HA. After going back to v4.2.0 and reinstalling my new integrations everything seems to work.
So I guess, v5.0.0 with the error writing the config is a severe problem, especially because you don't see in the log that this problem is caused by ics_calendar.

[wralb]

Thanks @ccMatrix, I also had to include the reference to connection_timout to get the migration to proceed. I also found that new entities with _2 appended were created and I had to delete theold, and rename the new so they would work with all my templates and automations.

[franc6]

I've not been able to reproduce this one so far -- my own import data never included connection_timeout, since it's not something I've ever needed to adjust. It'll probably be a while before I can get to this. :(

[ccMatrix]

I didn't have connection_timeout set before moving to 5.0.0 which is where the error came from. Setting it fixes it since then the default is no longer None which is an object it seems to have issues with serializing.

I'm running Home Assistant on a Raspberry Pi 4 via Docker
HA version: 2024.9.1
Docker Image: ghcr.io/home-assistant/raspberrypi4-homeassistant:stable
OS: Ubuntu 22.04 LTS on aarch64 architecture

It is possible that this might be different when on e.g. x86/amd64. Maybe others with the same problem also have it on arm ?

[a351must2]

Thanks @ccMatrix, I also had to include the reference to connection_timout to get the migration to proceed. I also found that new entities with _2 appended were created and I had to delete theold, and rename the new so they would work with all my templates and automations.

This is what I observed as well. I had to create new ICS_Calendar entries, then I was able to see the old entities listed saying they were no longer provided by ICS Calendar. I was able to delete the old entries and rename the new ones I added to get everything working. I noticed as I was about to roll back to 4.2.0.

[franc6]

I would recommend going back to 4.2.0, anyway for now. I am curious, though it shouldn’t make any difference for serializing, is anyone who sees this running HA on x86-64?

[a351must2]

I would recommend going back to 4.2.0, anyway for now. I am curious, though it shouldn’t make any difference for serializing, is anyone who sees this running HA on x86-64?

Yes, mine is on x86-64

[ccMatrix]

I see that in

ics_calendar/custom_components/ics_calendar/calendardata.py

Line 62 in d51df84

self.connection_timeout = _GLOBAL_DEFAULT_TIMEOUT

The default value for connection_timeout is set to _GLOBAL_DEFAULT_TIMEOUT which is imported from socket. IF I found the correct library then socket defines this as just object():
https://github.com/python/cpython/blob/e9eedf19c99475b1940bbbbdc8816b51da3968e7/Lib/socket.py#L826

So this might be the reason why it tries to serialize an object?

I didn't 100% understand yet where and how the serialization/conversion from the yaml config to the entity for HA happens in the migration process.

[franc6]

Oh! Great thought! Look at

ics_calendar/custom_components/ics_calendar/__init__.py

Line 259 in d51df84

data[CONF_CONNECTION_TIMEOUT] = None

. I think that’s the problem. Try changing it to use _GLOBAL_DEFAULT_TIMEOUT . If you have the time, feel free to make a PR if you can test it, too. That’s really what it should be, not None, anyway. If you (or someone) can test that, I’d really appreciate it.

[franc6]

I'm blind; I can reproduce this, and the change above doesn't help. If I can steal some time away today to work on this I will.

[ccMatrix]

My guess was that the object is initalized into a CalendarData object from the yaml which is then sent to HA for serialization into the entity store. There the connection_timeout is set to _GLOBAL_DEFAULT_TIMEOUT which probably isn't serializable by the json serializer. So you would have to set a different default there. None would work better since that can be serialized to null