AP_GPS: fixed the check for disabling CONFIG_RATE_SOL config

[tridge]

we should stop looking for SOL as soon as we know we have PVT, instead of waiting for all others to be cleared

may help with
https://discuss.ardupilot.org/t/fc-tries-to-config-gps-constantly-when-trying-to-arm-in-loiter-mode/129930

[peterbarker]

I bisected the bug, this was the result:

a9dc7b440f7a5b9418733c7b88743767c38623d4 is the first bad commit
commit a9dc7b440f7a5b9418733c7b88743767c38623d4
Author: Andy Piper <github@andypiper.com>
Date:   Mon Apr 29 15:46:53 2024 +0100

    AP_GPS: support GPSx_GNSS_MODE for F9P
    
    support detecting F9P hardware variant
    fix bug in extension buffer management
    support NEO-F9P GNSS configuration
    allow multiple configuration values to be set in one go
    phase F9 configuration to account for GNSS reset

[andyp1per]

@peterbarker looks right but will need testing on F9P if you did not do that already

[peterbarker]

I can't help but wonder whether a tweak to the code affected by the patch which introduced the problems might be better. In particular:

@@ -334,6 +336,12 @@ AP_GPS_UBLOX::_request_next_config(void)
         }
         break;
     case STEP_POLL_GNSS:
+        if (supports_F9_config()) {
+            if (last_configured_gnss != params.gnss_mode) {
+                _unconfigured_messages |= CONFIG_F9;
+            }
+            break;
+        }
         if (!_send_message(CLASS_CFG, MSG_CFG_GNSS, nullptr, 0)) {
             _next_message--;
         }

Renaming "supports_f9_config" to "supports_valget_and_valset()` might stop this happening again?

[MallikarjunSE]

@peterbarker do you suggest testing on F10N too?

[tridge]

@andyp1per I have tested GPS1_GNSS_MODE with F9P and it still works fine

[peterbarker]

Unrelated to this PR, but it looks like we leak config_GNSS when the UBLOX backend is deleted.

[timtuxworth]

This has also have fixed a problem I was seeing with the QioTek M10S GPS.

[rmackay9]

this is included in 4.6.0-beta5