* MockLink - Support REQUEST_MESSAGE:AUTOPILOT_VERSION, REQUEST_MESSAGE:PROTOCOL_VERSION
* Fix bugs in Vehicle::requestMessage
* Switch to all INT mission items
* Create unit tests to exercise Initial Connect state mache state failure conditions
Before this change it is not allowed to set up the flight modes unless
all sensor calibrations are completed.
I don't think that makes sense because it is possible to calibrate RC
without having the sensors calibrated, and it's the logical next step
right after that.
It is also a problem if HITL (hardware-in-the-loop) is used where the
sensors are not marked as calibrated but you still need to be able to
configure the flight modes. This is what prompted me to suggest this
change.
* Fix more racey instances of `if (!weakPtr.expired()){...weakPtr.lock()...}`
* Fix `_multiLinkSingleVehicleTest` thinking that link 1 will always be the primary link
* Remove the `LinkInterface::Channel` object
* Remove the reserved channel 0 and make channel 0 reservable
* Ensure that any user of the `reserveMavlinkChannel()` function will get a compiler error to note that a return of 0 no longer means error
* Make the `APMFirmwarePlugin` use a static channel member mutexed also by a static member
* `_handleIncomingMavlinkBytes` needs to use a different channel because it processes messages parallel to `MAVLinkProtocol::receiveBytes`
* `_handleIncomingMavlinkBytes`'s auxiliary channel needs its own mutex guard
* causes all calls out to mavlink library from `MockLink` to use channel 0,
which causes problems with at least `VehicleLinkManagerTest::_multiLinkSingleVehicleTest()`
and possibly other tests
The main problem with the previous implementation is that the QGCCacheWorker
does not have an event loop (not calling exec).
Since network function that are not async (such as waitConnection) needs an eventloop,
the Fuctor will run in another thread.
To fix this behaviour, the implementation was moved from to an async approach with signals.
Signed-off-by: Patrick José Pereira <patrickelectric@gmail.com>
For increased type safety, some enums have been changed to
scoped enums in Qt 6.2, see
https://codereview.qt-project.org/c/qt/qtconnectivity/+/336678
This patch adapts qgroundcontrol to this change.
Since C++11, enums inject their symbols in both their own
and their parent namespace, so this patch can be merged right
now.
Signed-off-by: Andreas Buhr <andreas.buhr@qt.io>
* Fix racey access of `QGCCacheWorker::_taskQueue`
* `QGCCacheWorker::_valid` needs to be atomic
* Use `QScopedPointer<QSqlDatabase>` for safety
* Some refactoring for readability & debugging
* Name and layout change for UDP/TCP settings
- To be more coherent with terminology used by MAVLink proxy/router softwares, the connection fields' labels was changed.
- A slight redesign of UDP server/host connection settings to allow for explaining notes and easier overview.
* Inherit widths from parent for less static layout
* Removed non-functional and unused reference
Don Gagne
DonLakeFlyer
Don Gagne
Andrew C. Smith
numan.uyar
Julian Oes
Aleksey Kontsevich
Patrick José Pereira
Thies Lennart Alff
Keith Bennett
PX4BuildBot
Andreas Buhr
Christoffer Björnram
Remek Zajac