deInit() on main tears down the parallel bus (behavior change vs 2.2.0)

[AsLY4]

Since 2.2.0, deInit() on main also calls bbepIODeInit, tearing down
the parallel bus. Firmware that uses an init/deinit-per-refresh pattern
(OpenDisplay Firmware does, to share pins and power down between
refreshes) works on 2.2.0 but hangs when built against main — the next
initPanel() finds the bus gone.

Observed on: M5Stack PaperS3 (BB_PANEL_M5PAPERS3), ESP32-S3.
Repro: build OpenDisplay Firmware against main instead of the 2.2.0 pin;
first refresh after a deinit/reinit cycle hangs.

Is the new deInit semantics intentional? If yes, could it be documented
(and/or a lighter "release frame only" API kept) so downstreams know
which teardown they get?

[bitbank2]

I will work on a fix. deinit should remove the parallel bus and everything connected to the GPIO pins, but the next init should restore everything.

[bitbank2]

I just tested this on the M5PAPERS3. I created a simple Arduino sketch which does 2 iterations of init/draw_something/deinit and it worked correctly. This is run against the main branch (which has additional commits beyond the 2.2.0 release). I don't have an easy way of testing the HA version. Are you saying that there is an issue with the 2.2.0 release or the latest code on the main branch? Either way, I cannot reproduce the problem that you claim exists.