feat(UI): Show antenna RSSI/SNR as 'XX YY' instead of single value for multiple antennas setups

[koriaf]

this feature must be enabled by creating a file as shown in the code, default behavior - no changes

doc_2026-08-04_08-27-40.mp4

@PetruSoroaga FYI

[wkumik]

Thanks for this — per-antenna visibility is genuinely useful on multi-antenna ground setups, and gating it behind showantennavalues keeps it zero-risk for everyone else. Two things I'd like resolved before this goes in, one of them a cross-version compatibility break.

1. The struct you extended is transmitted over the radio, not just shared memory

shared_mem_radio_stats_radio_interface_rx_signal_all (shared_mem_radio.h:54) is embedded as signalInfo in both shared_mem_radio_stats_radio_interface (:70) and shared_mem_radio_stats_radio_interface_compact (:115). The vehicle sends both to the ground station (r_vehicle/periodic_loop.cpp:449/469/493), and the GS validates them by exact size:

// r_central/process_router_messages.cpp:1398
if ( pPH->total_length == (sizeof(t_packet_header) + 2*sizeof(u8) + sizeof(shared_mem_radio_stats_radio_interface)) )

Two int[MAX_RADIO_ANTENNAS] adds 32 bytes, so a GS built with this patch talking to a vehicle without it (or vice versa) fails that comparison and silently drops every vehicle Rx stats packet — the vehicle radio stats OSD pages just go empty. Anyone running a mixed pair while updating one side at a time hits it.

Since the OSD only reads g_SM_RadioStats.radio_interfaces[i] — the ground station's own cards — these values never need to cross the link. Moving them to a new top-level field in shared_mem_radio_stats (which is shm-only and never transmitted), e.g. int iAntennaDBM[MAX_RADIO_INTERFACES][MAX_RADIO_ANTENNAS], gives you the same feature with the wire format untouched.

2. SNR is computed without checking the noise value is valid

In osd_links.cpp:

aAntennaSNR[iAntennaValidCount] = iAntDBM - g_SM_RadioStats.radio_interfaces[i].signalInfo.iAntennaDBMNoise[k];

Only the dBm side is range-checked. radiolink.c:1104 guards the noise side before doing the same subtraction:

if ( iAntennaDBMNoise[i] < 500 )
   int iSNR = iAntennaDBM[i] - iAntennaDBMNoise[i];

That guard matters on Realtek 8812 cards: the driver never reports the radiotap antenna-noise field, so iAntennaDBMNoise stays at the 1000 sentinel (this is why the normal OSD shows SNR: --- on those cards). The subtraction then yields -65 - 1000 = -1065, and _osd_format_antenna_values() runs it through abs(), so the OSD displays 1065 1070 instead of SNR: ---. Same < 500 check here, falling back to the existing single-value path when noise is unavailable, should cover it.

Minor

• Per-antenna values are never aged out — once an antenna has reported, its last dBm persists indefinitely in runtimeInterfaceInfoRx, so a dead or unplugged antenna keeps displaying a stale reading. The aggregate path recomputes per packet, so this is new behaviour.
• _osd_show_antenna_values_enabled() caches on first call, so creating or removing the marker file needs a ruby_central restart. Worth a line in the PR description.
• In SNR mode the multi-antenna string drops the SNR: prefix, so 65 70 renders identically to the dBm case.

Everything else checks out — buffers are comfortably sized (szLine1/2 are 64 bytes against a ~20 char worst case), the default arguments are fine since _osd_show_radio_bars_info is file-local with no header declaration, and the aAntenna* arrays are never read when iAntennaValidCount <= 1.

Happy to push the two fixes myself if you'd rather not — just say the word.