Hi Ingvar, On 14-Mar-25 9:52 AM, Ingvar Hagelund wrote: > to., 13.03.2025 kl. 19.43 +0100, skrev Hans de Goede: >> Here is v8 of the patch to upstream the OV02C10 sensor driver >> originally writen by Intel which Heimir has been working on >> upstreaming. >> > > Many thanks to Heimir and Hans for this excellent work. This makes my > workday easier. > >> (...) >> >> 1. I don't have hardware to test. I hope that others can test this >> soon, >> if things don't work the idea is that people can apply my cleanups >> 1 by 1 and then we will know which change has broken things. > > Seems to work fine on my Dell XPS 13 9340. I have not found any > glitches so far. Tested with on fedora 41 with qcam, cheese, obs, and > firefox - tested with websites gum, jitsi, and webcamtests.com. All > these work fine. webcamtests.com was even able to select highest > resolution/zoom. Note that chromium does *not* work yet, at least not > in Fedora. Great thank you for testing. I was afraid that I would have broken something it is good to hear that I did not break anything. Sakari, do you want to take a look at the incremental patches in this v8 to get an idea of what I changed after your last review and maybe give feedback on specific changes, or shall I post a squashed v9 which might be easier for you to review ? > I now use this as my daily camera, without any problems. > > I only miss more controls available, for example adjusting colors and > white balance. Some basic automatic adjustment seem to work, like when > changing rooms or lightning, but when for example you sit by a window > with sun on half your face, the image distorts, like getting over- or > underexposed. Also, the colors seems a bit washed out compared to my > usb cam (and to real life colors) Yes we need to work on improving the image quality and things like the autoexposure algorithm, note this is all work which needs to be done on the libcamera / softisp side not on the kernel side. > Switching to and fro between v7 and v8 without rebooting gave unstable > and strange results, but I presume that is less important, or even > expected. This is expected the sensor and CSI receiver get linked together once both have loaded, rmmod-ing the sensor driver after this linking is done is not really supported. Regards, Hans
