fix: the initial camera controls not used when starting camera

[christian-nils]

Fixes issues including the one described in #37 where the sink camera is not aware of the targeted framerate causing problems with the autoexposure algorithm.

[christianrauch]

Thanks a lot for this PR. Is the issue related to #30?

Can you give an example of what did not work for you before this PR and how that changed with this PR?

[christian-nils]

I don't think this will solve #30 (the bug report is also a bit confusing). I guess some of the listed stream formats are just not supported by the camera causing the issue reported.

For this PR, I can explain a bit more why it is relevant. In my case, I have two IMX477 sensors, one configured as a source and the other one as a sink using the following dtoverlay configuration:

dtoverlay=imx477,cam0,rotation=0,always-on,sync-sink
dtoverlay=imx477,rotation=0,always-on,sync-source

This means that the sink sensor will only capture frames when it receives a trigger on its XVS line.

If you don't add the expected framerate for the sink before starting the camera, you'll have issue with the AEC/AGC algorithm. After the 3rd frame of the startup phase (to initialize the algorithms I guess? those frames are dropped after https://github.com/raspberrypi/libcamera/blob/eb00c13d7c9f937732305d47af5b8ccf895e700f/src/libcamera/pipeline/rpi/pisp/data/example.yaml#L22), you will get an updated metadata with Exposure=0.

From the report:

[2:08:07.563255269] [673] DEBUG IPARPI ipa_base.cpp:1274 Metadata - Exposure: 985.83us Frame length: 2197 Line length: 15.17us Gain: 1
[2:08:07.563389046] [673] DEBUG IPARPI cam_helper.cpp:251 Metadata updated - Exposure: 0.00us Frame length: 0 Line length: 0.00us Gain: 1 Temperature: 0

This will happen for every following frames.

When using the fix, you set the controls before the camera as explained here (https://libcamera.org/api-html/classlibcamera_1_1Camera.html#a5c033a72ac66fbdd2c2bb9f934389dcf). Which leads to fixing the error reported above:

[2:19:04.777075042] [862] DEBUG IPARPI ipa_base.cpp:1274 Metadata - Exposure: 985.83us Frame length: 2175 Line length: 15.17us Gain: 1
[2:19:04.777131041] [862] DEBUG IPARPI cam_helper.cpp:251 Metadata updated - Exposure: 985.83us Frame length: 2175 Line length: 15.17us Gain: 1 Temperature: 27

So, in short, we get the following before the fix (obstructing the first frames to get a high exposure which bugs the successive capture):
bad.webm

we get the following after applying the patch
good.webm

Please let me know if that addresses your questions.