The camera work concerns the Latitude 7320 Detachable. The similarly named conventional laptop must not be assumed to contain the same camera hardware.
What was demonstrated
The August 9 patch report describes a front sensor advertised by ACPI as OVTI5678 but returning chip ID 0x005675. The author reports libcamerasrc capture at 2584 × 1944 and about 29.95 fps. That is the contributor’s result, not a PeopleAreGeek hardware test or a promise for all units.
The sensor produces an RGB-IR mosaic. Treating it as an ordinary Bayer color pattern does not make the infrared samples ordinary green samples; a white-balance adjustment alone cannot establish correct color reconstruction.
The September follow-up changes the wiring claim
In the September 1 development discussion, the contributor corrects the reset line to GPIO 5 from GPIO 3 after controlled tests on two machines. The maintainer still calls for an OV5678 driver and CommonRawSensorModel work. This remains development work, not evidence of complete support arriving in every distribution.
The Linux 7.3-rc2 board-data file does not contain the proposed 7320-specific entry. Front-camera progress also cannot be silently generalized to the rear camera.
Why a successful probe can mislead
The cover separates power and reset, sensor binding and media routing, then pixel interpretation. Getting an image proves that some path worked in that setup. It does not independently prove every rail name, reset assignment or format declaration is right.
The reset investigation illustrates a stronger test: vary the suspected control while keeping the rest fixed, and observe whether the predicted failure occurs. A line that appears in a working configuration may be irrelevant; a controlled low/high result provides more specific evidence. This is the diagnostic reasoning in the discussion, not an invitation to write undocumented camera registers on a daily-use system.
For a purchase or Linux migration, look for support in the exact distribution kernel and camera stack, plus working color, application capture and suspend/resume. A patch submission and a working raw stream are useful milestones. Neither, alone, answers that end-to-end question.
September 8: identify the Detachable model and separate capture, RGB-IR handling and upstream support; incorporate September reset-pin correction and outstanding driver work.