get_all_params error (std::future_error: Promise already satisfied)

[elbarto1980]

Hi, I am running MAVSDK 1.4.0 on Ubuntu 20.04.
I am trying to get all PX4 params using the get_all_params() function and I noticed that the prom.set_value(all_params); line of code causes a segmentation fault. It happens every time, if the program does not terminate immediately after receiving the parameters list.
The error I get is:

terminate called after throwing an instance of 'std::future_error'
what(): std::future_error: Promise already satisfied

I think this issue might be somehow related to another issue ( #1527) that @julianoes fixed on September 2021.

Hopefully, these few following lines of code should be useful to reproduce the error:

#include <mavsdk/mavsdk.h>
#include <mavsdk/plugins/param/param.h>
#include <thread>
#include <iostream>

using namespace mavsdk;

int main() {
    Mavsdk mavsdk;

    mavsdk.add_udp_connection();
    std::this_thread::sleep_for(std::chrono::seconds(2));

    auto system = mavsdk.systems().at(0);
    auto param = Param{system};
    auto all_params = param.get_all_params();

    for (auto param : all_params.int_params) {
        std::cout << param.name << std::endl;
    }

    int i = 60;
    do {
        std::this_thread::sleep_for(std::chrono::seconds(1));
        i--;
        std::cout << i << std::endl;
    } while (i>0);
}

Thank you!

[julianoes]

Thanks for the issue @elbarto1980, and sorry that I broke that during the recent refactoring it's broken. I will look into it.

[elbarto1980]

Thank you @julianoes for providing a fix so quickly!
Could you please leave this issue open just for a couple more days? My code (the actual code, not the example above) is still crashing at the same point with the same error and I would like to investigate the error more thoroughly. Thanks

A few hours later.....

Sorry for the multiple edits but I think I found what makes my code crash... I send a set_takeoff_altitude() command after the get_all_params() command, and it makes it crash as it crashed before.

The following example should reproduce what happens:

#include <mavsdk/mavsdk.h>
#include <mavsdk/plugins/action/action.h>
#include <mavsdk/plugins/param/param.h>
#include <thread>
#include <iostream>

using namespace mavsdk;

int main() {
    Mavsdk mavsdk;

    mavsdk.add_udp_connection();
    std::this_thread::sleep_for(std::chrono::seconds(2));

    auto system = mavsdk.systems().at(0);

    auto param = Param{system};
    auto all_params = param.get_all_params();

    for (auto param : all_params.int_params) {
        std::cout << param.name << std::endl;
    }

    std::this_thread::sleep_for(std::chrono::seconds(2));

    auto action = Action{system};
    auto setTakeoffAltitudeResult = action.set_takeoff_altitude(15.0f);
    if (setTakeoffAltitudeResult != Action::Result::Success) {
        std::cerr << "Set takeoff altitude failed: " << setTakeoffAltitudeResult << std::endl;
    } else {
        auto takeoffResult = action.takeoff();
        if (takeoffResult != Action::Result::Success) {
            std::cout << "Takeoff failed: " << takeoffResult << std::endl;
        }
    }

    int i = 60;
    do {
        std::this_thread::sleep_for(std::chrono::seconds(1));
        i--;
        std::cout << i << std::endl;
    } while (i>0);
}

Thank you

[julianoes]

Oh, a new twist on this? 🤔 Let's see.

[julianoes]

There was indeed another bug, thanks for the nice example to reproduce it.

See: 5ce3a5c

[julianoes]

Thanks again for your help @elbarto1980. The minimal examples to reproduce this were very helpful, and allowed me to quickly pinpoint what was wrong!

[elbarto1980]

Glad to be of help! Thank you @julianoes for fixing it... it's working flawlessly now.

[julianoes]

Nice, yes I backported the fixes to v1.4.2, until v2.0.0 ships.