Render Betaflight VTX temperature element

[etiaro]

Betaflight does not allow to enter a custom vtx elements like INAV does, but it has VTX temperature widget that can be filled with reading on the VTX msposd. Tested on OpenIPC Thinker (Sigmastar).

Assisted-By: OpenAI GPT-5.6 Terra

[tipoman9]

I don't get the benefit of this PR. Can you explain or show a snapshot of the end result? VTx temp can easily be shown over the OSD if you have !TMW! text somewhere on the OSD, just set PilotName or CraftName to it and you will have a nice widget over the position of the !TMW! text. https://github.com/OpenIPC/msposd#custom-osd-elements

[henkwiedig]

He is refering to the betaflight msp_dp_sys command: https://betaflight.com/docs/development/API/DisplayPort#msp_dp_sys

// System elements rendered by VTX or Goggles
typedef enum {
    DISPLAYPORT_SYS_GOGGLE_VOLTAGE = 0,
    DISPLAYPORT_SYS_VTX_VOLTAGE = 1,
    DISPLAYPORT_SYS_BITRATE = 2,
    DISPLAYPORT_SYS_DELAY = 3,
    DISPLAYPORT_SYS_DISTANCE = 4,
    DISPLAYPORT_SYS_LQ = 5,
    DISPLAYPORT_SYS_GOGGLE_DVR = 6,
    DISPLAYPORT_SYS_VTX_DVR = 7,
    DISPLAYPORT_SYS_WARNINGS = 8,
    DISPLAYPORT_SYS_VTX_TEMP = 9,
    DISPLAYPORT_SYS_FAN_SPEED = 10,
    DISPLAYPORT_SYS_COUNT,
} displayPortSystemElement_e;

[tipoman9]

@etiaro
Seems you've discovered and fixed several bugs, one of them quite nasty (a resource leak):

1. event/hotkey leak. FlushDrawing() declared a local struct event *x11_event that shadowed the global, so the X11 hookup guard never latched — every frame re-registered the X11 event and re-grabbed hotkeys, leaking one event per frame indefinitely.

2. Fix double free of the libevent event_base on shutdown. handle_data() freed base and then called CloseMSP() → Close(), which freed base again — a double free/use-after-free on every normal shutdown. Reordered so Close() runs first and base is freed exactly once, afterward.

3. UDP mode: stop corrupting stdin's terminal settings. In UDP-only mode serial_fd stayed at its default value of 0 (stdin), so the UART termios setup (cfmakeraw, etc.) was applied to stdin, breaking the invoking terminal. Default changed to -1 and the termios block now only runs when a real UART fd is open.

Could we keep these bug fixes in a separate commit and avoid the unrelated indentation/style-only changes? That would make the PR much easier to review and the commit history easier to follow. Claude created it for me here

Could you please exclude msposd.c and render_gs.c from this commit, leaving only osd.c? Since osd.c contains all the code needed for the Betaflight widget, I think that would keep this change focused. I'll be happy to approve it once that's done.

Thank you for your effort!

[etiaro]

Sure, sorry, I planned to commit those fixes in separate PR and included accidentally, but I'll leave them here as you requested.

I introduced also the bitrate display and added a little more generic approach, for now I don't know about a good source for the remaining values (not that they could be filled on GS when GS-rendering).

Thanks for quick response.

[tipoman9]

but I'll leave them here as you requested.

Actually I meant the opposite ;-)
Could you please exclude msposd.c and render_gs.c from this commit, leaving only osd.c?

[etiaro]

Yeah, that's a misunderstanding of commit vs. PR. Anyway, it's done