CI failing on `ergoCubSN000` and `iCubGenova09` with the latest `yarp-3.8` about robots-configuration

[traversaro]

This is more of an issue of embObjMultipleFTsensors, but indeed it will be mitigated by robotology/yarp#2967 .

It is related to the second fix from robotology/icub-main#871, and actually the reason why now it appears is that we fixed that bug. Before then, the value returned by embObjMultipleFTsensors::getNrOfTemperatureSensors was zero if no temperature measurement was received, and so no one called embObjMultipleFTsensors::getTemperatureSensorMeasure . robotology/yarp#2967 ensures that embObjMultipleFTsensors::getTemperatureSensorMeasure is not called if the status of the sensors is not MAS_OK, so it should ensure that this error does not appear.

However, after https://github.com/robotology/yarp/pull/2967/files you will see a different error in the terminal:

Failure in reading data from sensor %s, no data will be sent on the port.

So I am not sure if this is better from the point of view of CI.

[traversaro]

It is not the driver, it is the multipleanalogsensorsserver that will print errors if no valid sensor measure is available. I wonder how other network wrapper server deal with this, perhaps they do not complain and just do not publish data if it is not available?

[traversaro]

I think it could make sense if we agree on a dryrun or similar parameter to pass to every device so that every device is aware and suppress errors that do not make sense to print during a dryrun (such as errors because no sensor message arrived).

[traversaro]

@traversaro is on it.

My bad, apparently I forgot to push some changes, the fix is in robotology/yarp#2970 .

[pattacini]

Thanks for the clarification!

So I am not sure if this is better from the point of view of CI.

Does your last doubt mean that the driver per se in this configuration is not ready to be dry run at all?

[pattacini]

Yep, with the term driver I was referring to the whole stack.

This can be a major issue for the CI. In the worst case, I may try to disable these layers during the CI as done for others, but it'd be a pity after all.

We can discuss it.
Thanks

[pattacini]

Temporary disabled the failing tests in the CI via 93d60a4.

Failing tests can be still run by:

ctest -L [failing|all]

cc @traversaro

[pattacini]

ea68236 and 4d2081b streamline the handling of skipped tests.

Now, to run locally or on Gitpod skipped tests, one needs to do:

ctest -L skip

Tests to be skipped are specified in:

• https://github.com/robotology/robots-configuration/blob/master/tests/dry-run/skip_tests.txt

[pattacini]

Aligned T2T w/ @traversaro: all good then 👍🏻
The error message is not causing any harm to the CI.

The CI failed because of the yarprobotinterface crash, which will be solved via robotology/yarp#2967.

Note
Put in review.

Warning
Remember to reinstate the skipped tests before closing.

[pattacini]

The PR robotology/yarp#2967 has been merged although the dry-run still fails with the following log:

[ERROR] yarprobotinterface intercepted a segmentation fault caused by a faulty plugin:
[ERROR] /dependencies/install/lib/yarp/yarp_multipleanalogsensorsserver.so(+0x13f69) [0x7f42b401cf69]

@traversaro is on it.

---

The docker image has been regenerated:

• https://github.com/robotology/robots-configuration/pkgs/container/robots-configuration

[pattacini]

Thanks!