2.8RC1 ELRS Lua BLE Joystick stops working when long press exit

[RipperGOT]

Is there an existing issue for this problem?

• I have searched the existing issues

What part of EdgeTX is the focus of this bug?

Transmitter firmware

Current Behavior

In 2.7 when you activate the BLE joystick you can do a long press to exit lua and it stays active. In 2.8 if you long press to exit, it stops working.
Sometimes I get this error: Script syntax error ?:0: attempt to index field '?' (to nil value)
I also see the Lua a little lazy

Expected Behavior

You could long press RTN to exit the LUA and keep it active.

Steps To Reproduce

1 Enter in ELRS LUA-BLE Joystick and activate
2 Long press RTN to exit to model screen
3 BLE stop working

Version

Other (Please specify below)

Transmitter

Radiomaster TX16S / TX16SMK2

Anything else?

[RipperGOT]

In RC3 the ELRS lua is still lazy and the BLE joystick crashes from time to time. I also can't get out of the BLE without it disconnecting.

[pfeerick]

By lazy, do you mean slow? As I've noticed that also, and it's not LVGL specific as B&W also seems sluggish at times.

[pfeerick]

PK from ELRS just commented that we must be sending the EVT_VIRTUAL_EXIT event on exit (which I think we are), whereas 2.7 just killed the script. So somehow that is probably triggering the stopping of wifi/ble that would happen on a short RTN press.

[RipperGOT]

Exactly, it's slow. It looks like lag.

[raphaelcoeffic]

Exactly, it's slow. It looks like lag.

This is caused by the fact that we switched to LVGL, and the LUA script does a full redraw on each frame. It is faster on radios that do not require the screen buffer to be inverted in software (T18, NV14, EL18). The screen inversion (180° rotation) in software costs us around 20ms on each frame. That does add up to the rest, which makes it look slower.

[raphaelcoeffic]

PK from ELRS just commented that we must be sending the EVT_VIRTUAL_EXIT event on exit (which I think we are), whereas 2.7 just killed the script. So somehow that is probably triggering the stopping of wifi/ble that would happen on a short RTN press.

This is probably caused by the fact the event still gets passed to the script while we actually kill it. I can check if there isn't some weirdness here.

[gagarinlg]

Long term goal: expose the UI API to Lua in a way, that LVGL handles partialnredraws of Lua scripts