Settings are overwritten on startup for OpenFlexure camera

[rwb27]

Describe the bug
The settings for openflexure_microscope_server.things.camera.picamera.StreamingPiCameraV2 are overwritten with defaults when I run ofm restart or otherwise restart the server

To Reproduce

• Start up the OpenFlexure Microscope server (I'm using latest v3 branch: !636 6ae8f8d970c7b7e18bc0637d2afd6fc37855be32)
• Change the camera's exposure time e.g. curl http://localhost:5000/api/v3/camera/exposure_time --request PUT -d 2000
• Check the change went through e.g. curl http://localhost:5000/api/v3/camera/exposure_time
• Check the change is reflected in settings/camera/Settings-StreamingPiCameraV2.json
• Restart the server with ofm restart
• Observe that the settings file has reverted
• Observe that the exposure is back to default curl http://localhost:5000/api/v3/camera/exposure_time

It's also possible to observe in the GUI, but this method was easier to use when finding the offending commit.

Expected behaviour
The settings should be remembered across a restart

Additional context
This problem was introduced by #256. The previous merge commit does not exhibit the problem.

This is not simply a blanket failure to remember settings. That would have been caught by unit tests. Other Things on the OpenFlexure server (e.g. the Camera Stage Mapping thing) do correctly remember their settings. The camera hardware is a bit badly-behaved when it comes to settings, so the camera Thing has some complicated logic embedded in various functional settings:

• When a setting is read, it will attempt to read the value from hardware (if the hardware is running). If the hardware is not running, it will return the cached value (which is initialised to a default.
• If a setting is read and the cached value does not match the value from the hardware, the cached value and the settings file will be updated.
• When a setting is written, it will update the cached value and the settings file.
• When a setting is written, if the camera is running it will also update the hardware.

My guess is that the problem appeared because of some subtle change in how the settings are initialised. This is resulting in the settings file being written during start-up of the Thing, using cached values for (most?) properties.

I think we can probably restructure the settings to get rid of the problem. However, I think it might be preferred to fix it in LabThings if we can. Restructuring the OFM code to be easier to understand would still, I think, be useful - but might be a longer-term change.

A log from before the bug, with a bunch of extra stuff added to the camera class.

Filtered log:

[2026-07-02 10:56:30,303] [INFO] <openflexure_microscope_server.logging> OFM server root logger has been set up at INFO level
[2026-07-02 10:56:30,336] [INFO] <labthings_fastapi.things.camera> Starting picamera __init__
[2026-07-02 10:56:30,341] [INFO] <labthings_fastapi.things.camera> stopping __init__
[2026-07-02 10:56:30,372] [INFO] <labthings_fastapi.things.camera> Setting exposure time to 1984, previous was 500.
[2026-07-02 10:56:30,373] [INFO] <labthings_fastapi.things.camera> Current streaming status: False
[2026-07-02 10:56:31,623] [INFO] <labthings_fastapi.things.camera> starting StreamingPiCamera2.__enter__
[2026-07-02 10:56:31,624] [INFO] <labthings_fastapi.things.camera> starting _initialise_picamera
[2026-07-02 10:56:31,628] [INFO] <openflexure_microscope_server.things.camera.picamera> Creating new Picamera2 object
[2026-07-02 10:56:31,683] [INFO] <labthings_fastapi.things.camera> stopping _initialise_picamera
[2026-07-02 10:56:31,685] [INFO] <labthings_fastapi.things.camera> Getting exposure time. Cached value 1984.
[2026-07-02 10:56:31,685] [INFO] <labthings_fastapi.things.camera> Returning 1984
[2026-07-02 10:56:31,727] [INFO] <openflexure_microscope_server.things.camera.picamera> Starting picamera MJPEG stream...
[2026-07-02 10:56:31,796] [INFO] <labthings_fastapi.things.camera> stopping StreamingPiCamera2.__enter__

Full log

``` [2026-07-02 10:56:30,303] [INFO] OFM server root logger has been set up at INFO level [2026-07-02 10:56:30,336] [INFO] Starting picamera __init__ [2026-07-02 10:56:30,341] [INFO] stopping __init__ [2026-07-02 10:56:30,372] [INFO] Setting exposure time to 1984, previous was 500. [2026-07-02 10:56:30,373] [INFO] Current streaming status: False [2026-07-02 10:56:31,609] [INFO] Started server process [79330] [2026-07-02 10:56:31,610] [INFO] Waiting for application startup. [2026-07-02 10:56:31,623] [INFO] starting StreamingPiCamera2.__enter__ [2026-07-02 10:56:31,624] [INFO] starting _initialise_picamera [2026-07-02 10:56:31,628] [INFO] Creating new Picamera2 object [2026-07-02 10:56:31,680] [INFO] Initialization successful. [2026-07-02 10:56:31,680] [INFO] Camera now open. [2026-07-02 10:56:31,683] [INFO] stopping _initialise_picamera [2026-07-02 10:56:31,685] [INFO] Getting exposure time. Cached value 1984. [2026-07-02 10:56:31,685] [INFO] Returning 1984 [2026-07-02 10:56:31,688] [INFO] Camera configuration has been adjusted! [2026-07-02 10:56:31,690] [INFO] Configuration successful! [2026-07-02 10:56:31,727] [INFO] Starting picamera MJPEG stream... [2026-07-02 10:56:31,761] [INFO] Camera started [2026-07-02 10:56:31,796] [INFO] stopping StreamingPiCamera2.__enter__ [2026-07-02 10:56:31,797] [INFO] Initialising ExtensibleSerialInstrument on port None [2026-07-02 10:56:31,798] [INFO] Updating ESI port settings [2026-07-02 10:56:31,798] [INFO] Opening ESI connection to port None [2026-07-02 10:56:31,806] [INFO] Creating serial.Serial instance... [2026-07-02 10:56:31,808] [INFO] Created Serial(port='/dev/ttyAMA0', baudrate=115200, bytesize=8, parity='N', stopbits=1, timeout=None, xonxoff=False, rtscts=False, dsrdtr=False) [2026-07-02 10:56:31,808] [INFO] Testing communication to SangaBoard [2026-07-02 10:56:31,808] [INFO] Running firmware checks... [2026-07-02 10:56:31,814] [INFO] Firmware response: Sangaboard Firmware v1.0.4 [2026-07-02 10:56:31,816] [INFO] Success! [2026-07-02 10:56:31,817] [INFO] Creating serial.Serial instance... [2026-07-02 10:56:31,818] [INFO] Created Serial(port='/dev/ttyAMA0', baudrate=115200, bytesize=8, parity='N', stopbits=1, timeout=None, xonxoff=False, rtscts=False, dsrdtr=False) [2026-07-02 10:56:31,818] [INFO] Testing communication to SangaBoard [2026-07-02 10:56:31,819] [INFO] Running firmware checks... [2026-07-02 10:56:31,824] [INFO] Firmware response: Sangaboard Firmware v1.0.4 [2026-07-02 10:56:31,825] [INFO] Opened ESI connection to port None [2026-07-02 10:56:31,826] [INFO] Checking valid firmware... [2026-07-02 10:56:31,826] [INFO] Running firmware checks... [2026-07-02 10:56:31,831] [INFO] Firmware response: Sangaboard Firmware v1.0.4 [2026-07-02 10:56:31,832] [INFO] Loading modules... [2026-07-02 10:56:31,848] [INFO] Enabling backwards-compatible behaviour on new Sangaboard firmware [2026-07-02 10:56:31,872] [INFO] Application startup complete. [2026-07-02 10:56:31,877] [INFO] Uvicorn running on http://0.0.0.0:5000 (Press CTRL+C to quit) ```

Log showing the problem

Filtered log:

[2026-07-02 11:02:37,670] [INFO] <openflexure_microscope_server.logging> OFM server root logger has been set up at INFO level
[2026-07-02 11:02:37,702] [INFO] <labthings_fastapi.things.camera> Starting picamera __init__
[2026-07-02 11:02:37,713] [INFO] <labthings_fastapi.things.camera> Getting exposure time. Cached value 500.
[2026-07-02 11:02:37,714] [INFO] <labthings_fastapi.things.camera> Returning 500
[2026-07-02 11:02:37,724] [INFO] <labthings_fastapi.things.camera> stopping __init__
[2026-07-02 11:02:37,748] [WARNING] <labthings_fastapi.things.camera> None is not a valid background detector name.
[2026-07-02 11:02:37,749] [INFO] <labthings_fastapi.things.camera> Setting exposure time to 500, previous was 500.
[2026-07-02 11:02:37,749] [INFO] <labthings_fastapi.things.camera> Current streaming status: False
[2026-07-02 11:02:38,873] [INFO] <labthings_fastapi.things.camera> starting StreamingPiCamera2.__enter__
[2026-07-02 11:02:38,874] [INFO] <labthings_fastapi.things.camera> starting _initialise_picamera
[2026-07-02 11:02:38,877] [INFO] <openflexure_microscope_server.things.camera.picamera> Creating new Picamera2 object
[2026-07-02 11:02:38,933] [INFO] <labthings_fastapi.things.camera> stopping _initialise_picamera
[2026-07-02 11:02:38,934] [INFO] <labthings_fastapi.things.camera> Getting exposure time. Cached value 500.
[2026-07-02 11:02:38,935] [INFO] <labthings_fastapi.things.camera> Returning 500
[2026-07-02 11:02:38,975] [INFO] <openflexure_microscope_server.things.camera.picamera> Starting picamera MJPEG stream...
[2026-07-02 11:02:39,032] [INFO] <labthings_fastapi.things.camera> stopping StreamingPiCamera2.__enter__

Full log

[2026-07-02 11:02:37,670] [INFO] <openflexure_microscope_server.logging> OFM server root logger has been set up at INFO level
[2026-07-02 11:02:37,702] [INFO] <labthings_fastapi.things.camera> Starting picamera __init__
[2026-07-02 11:02:37,713] [INFO] <labthings_fastapi.things.camera> Getting exposure time. Cached value 500.
[2026-07-02 11:02:37,714] [INFO] <labthings_fastapi.things.camera> Returning 500
[2026-07-02 11:02:37,724] [INFO] <labthings_fastapi.things.camera> stopping __init__
[2026-07-02 11:02:37,748] [WARNING] <labthings_fastapi.things.camera> None is not a valid background detector name.
[2026-07-02 11:02:37,749] [INFO] <labthings_fastapi.things.camera> Setting exposure time to 500, previous was 500.
[2026-07-02 11:02:37,749] [INFO] <labthings_fastapi.things.camera> Current streaming status: False
[2026-07-02 11:02:38,859] [INFO] <uvicorn> Started server process [80579]
[2026-07-02 11:02:38,860] [INFO] <uvicorn> Waiting for application startup.
[2026-07-02 11:02:38,873] [INFO] <labthings_fastapi.things.camera> starting StreamingPiCamera2.__enter__
[2026-07-02 11:02:38,874] [INFO] <labthings_fastapi.things.camera> starting _initialise_picamera
[2026-07-02 11:02:38,877] [INFO] <openflexure_microscope_server.things.camera.picamera> Creating new Picamera2 object
[2026-07-02 11:02:38,930] [INFO] <picamera2.picamera2> Initialization successful.
[2026-07-02 11:02:38,930] [INFO] <picamera2.picamera2> Camera now open.
[2026-07-02 11:02:38,933] [INFO] <labthings_fastapi.things.camera> stopping _initialise_picamera
[2026-07-02 11:02:38,934] [INFO] <labthings_fastapi.things.camera> Getting exposure time. Cached value 500.
[2026-07-02 11:02:38,935] [INFO] <labthings_fastapi.things.camera> Returning 500
[2026-07-02 11:02:38,938] [INFO] <picamera2.picamera2> Camera configuration has been adjusted!
[2026-07-02 11:02:38,941] [INFO] <picamera2.picamera2> Configuration successful!
[2026-07-02 11:02:38,975] [INFO] <openflexure_microscope_server.things.camera.picamera> Starting picamera MJPEG stream...
[2026-07-02 11:02:39,012] [INFO] <picamera2.picamera2> Camera started
[2026-07-02 11:02:39,032] [INFO] <labthings_fastapi.things.camera> stopping StreamingPiCamera2.__enter__
[2026-07-02 11:02:39,033] [INFO] <sangaboard.sangaboard> Initialising ExtensibleSerialInstrument on port None
[2026-07-02 11:02:39,034] [INFO] <sangaboard.extensible_serial_instrument> Updating ESI port settings
[2026-07-02 11:02:39,035] [INFO] <sangaboard.extensible_serial_instrument> Opening ESI connection to port None
[2026-07-02 11:02:39,042] [INFO] <sangaboard.extensible_serial_instrument> Creating serial.Serial instance...
[2026-07-02 11:02:39,043] [INFO] <sangaboard.extensible_serial_instrument> Created Serial<id=0x7f625491b0, open=True>(port='/dev/ttyAMA0', baudrate=115200, bytesize=8, parity='N', stopbits=1, timeout=None, xonxoff=False, rtscts=False, dsrdtr=False)
[2026-07-02 11:02:39,044] [INFO] <sangaboard.sangaboard> Testing communication to SangaBoard
[2026-07-02 11:02:39,044] [INFO] <sangaboard.sangaboard> Running firmware checks...
[2026-07-02 11:02:39,049] [INFO] <sangaboard.sangaboard> Firmware response: Sangaboard Firmware v1.0.4
[2026-07-02 11:02:39,050] [INFO] <sangaboard.extensible_serial_instrument> Success!
[2026-07-02 11:02:39,051] [INFO] <sangaboard.extensible_serial_instrument> Creating serial.Serial instance...
[2026-07-02 11:02:39,052] [INFO] <sangaboard.extensible_serial_instrument> Created Serial<id=0x7f624e0d90, open=True>(port='/dev/ttyAMA0', baudrate=115200, bytesize=8, parity='N', stopbits=1, timeout=None, xonxoff=False, rtscts=False, dsrdtr=False)
[2026-07-02 11:02:39,052] [INFO] <sangaboard.sangaboard> Testing communication to SangaBoard
[2026-07-02 11:02:39,055] [INFO] <sangaboard.sangaboard> Running firmware checks...
[2026-07-02 11:02:39,060] [INFO] <sangaboard.sangaboard> Firmware response: Sangaboard Firmware v1.0.4
[2026-07-02 11:02:39,061] [INFO] <sangaboard.extensible_serial_instrument> Opened ESI connection to port None
[2026-07-02 11:02:39,061] [INFO] <sangaboard.sangaboard> Checking valid firmware...
[2026-07-02 11:02:39,061] [INFO] <sangaboard.sangaboard> Running firmware checks...
[2026-07-02 11:02:39,066] [INFO] <sangaboard.sangaboard> Firmware response: Sangaboard Firmware v1.0.4
[2026-07-02 11:02:39,067] [INFO] <sangaboard.sangaboard> Loading modules...
[2026-07-02 11:02:39,083] [INFO] <sangaboard.sangaboard> Enabling backwards-compatible behaviour on new Sangaboard firmware
[2026-07-02 11:02:39,108] [INFO] <uvicorn> Application startup complete.
[2026-07-02 11:02:39,112] [INFO] <uvicorn> Uvicorn running on http://0.0.0.0:5000 (Press CTRL+C to quit)

[rwb27]

So far I've determined:

• The setting is definitely read from the right file.
• There's an extra read happening during __init__ of the camera class. Note that this must be after super().__init__ given the placement of the logging lines.
• The settings file is being written during __init__, before the settings have been initialised.

Looking at the code, there's a write to self.tuning during __init__. This is wrapped in an exception handler, because writes to properties (before this bug was introduced) raised an exception when they tried to emit an event, but the event was not emitted.

I suspect the change in behaviour is a direct result of the exception disappearing. My guess is that the exception had the unintended side effect of preventing the settings write, which would otherwise have been triggered by that operation. When the exception was removed, this behaviour resulted.

I think this makes it a fairly general LabThings bug: writing to any setting during __init__ (a reasonable operation, to set the initial state) will overwrite the settings file. I think a reasonable fix would be to disable syncing to disk until the settings have been loaded. I can't see any reason that would be a bad idea.

In the future I'd like to give Things more control over when settings are loaded. For example, it might be nice to call self.load_settings during __enter__ so that the settings get directly written to the hardware. I think that would benefit from LabThings keeping track of whether load_settings has been called, and preventing accidental overwrites of the settings file before it's been loaded.