In short we have a high frequency ROS publisher which sends data and a listener application which receives the first couple of thousand samples and then seems to end up in a dead lock.
This deadlock seems to cause also a seg-fault in roudi when the listener is terminated with a kill signal.
Furthermore, when the listener is in this deadlock roudi cannot be terminated any longer.
I suspect that we have here a couple of issues. A deadlock in the listener should never cause roudi to be no longer able to terminate, this we have to investigate further on iceoryx side. Also the segfault which is caused when the deadlocked listener is killed.
But I think that the problem maybe originates somewhere in dds therefore I created the ticket.
In iceoryx the following bug was reported: https://github.com/eclipse-iceoryx/iceoryx/issues/1740
In short we have a high frequency ROS publisher which sends data and a listener application which receives the first couple of thousand samples and then seems to end up in a dead lock. This deadlock seems to cause also a seg-fault in roudi when the listener is terminated with a kill signal.
Furthermore, when the listener is in this deadlock roudi cannot be terminated any longer.
This repository provides examplary code and instructions on how to produce the issue: https://github.com/ceccocats/iceoryx_bug
I suspect that we have here a couple of issues. A deadlock in the listener should never cause roudi to be no longer able to terminate, this we have to investigate further on iceoryx side. Also the segfault which is caused when the deadlocked listener is killed.
But I think that the problem maybe originates somewhere in dds therefore I created the ticket.