0.42.0 - two things can no longer share a pin
The ATX buttons and the round LCD are both wired by hand, and both pick their GPIOs in the console. Nothing used to notice when they named the same pin - or when one of them named a pin the board's own hardware holds, like the capture bus or the Ethernet PHY. Neither shows up as an error. The panel simply stays dark, or the network stops.
A write that would do that is now refused, naming the pin and what already has it. The pins are weighed as a set and after the request is imagined applied, so moving the display off a pin the buttons want still goes through in one go, and a clash a device already has stored does not block unrelated changes. The defaults moved too: the buttons used to start unassigned with the wiring page suggesting the same three pins the LCD starts on, so a device wearing both and set up by following that page ended up with three pins claimed twice.
The capture path no longer calls the TC358743 by name. A driver registers a detect function, and the path asks for whatever answers on the bus, getting back a name to show and a table of operations to drive it with. Adding a bridge is adding a component. Nothing about the picture changes - there is still one driver and it is the same one - and the interface is a guess until a second bridge exists to shape it. The idea comes from Espressif's esp_cam_sensor.
Then two silences. A machine that boots faster than this firmware looks at the input, finds no monitor, and configures no output; single-board computers in particular probe once and never look again. The device now watches for that exact case - the source is powered and attached, but nothing is arriving - and pretends to be unplugged and plugged back in, at ten seconds, then twenty, then forty, and then it stops. A target that is simply switched off is left in peace. The console says which silence it is looking at, instead of offering the same list of things to check either way.
The other silence is the keyboard. A restart resets this side of the USB cable while the target's power never drops, so the target goes on holding a connection this side has forgotten and every keystroke goes nowhere. The firmware already came back as a new device once at start-up; now it tries again while nothing has enumerated it, and clicking the USB dot in the status bar offers "Re-plug USB", which does the same on demand. Off the bus for 100 ms and back, without restarting either machine.