Closed christophfroehlich closed 10 months ago
:eyes: @devwrite @mamut-m
This comes from https://github.com/ros2/rmw_dds_common/pull/68 changing the format of an internal topic that maintains the node graph, so CLI tools like ros2 node list
can get the distributed graph information.
PR https://github.com/ros2/rmw_fastrtps/pull/665 would ignore messages for which an exception is thrown when deserializing
Thanks @MiguelCompany for the quick reply, your PR solves my issue.
Fixed with #737
I already asked at RSE without any response, so I try my luck here. Please point me to a different repository if you think this is related to a different package.
Bug report
We experienced that humble executables crash if there are any iron or rolling nodes are running in the same network: it returns
std::bad_alloc
without any further warning.The problem is that the publisher/subsribers aren't working, but the nodes crashes immediately.
Any hints would be highly appreciated!
Required Info:
Steps to reproduce issue
We created a demo using the official docker containers from humble and iron and the minimal_publisher/subscriber: iron_humble_pub_sub_issue.tar.gz
If you run
./run_humble_subscriber.sh
in one terminal and./run_iron_publisher.sh
in another (in that order), you will usually see the following error at the iron humble subscriber terminal:sometimes it does not appear, usually restarting the publisher causes the problem.
Expected behavior
No errors.
Actual behavior
Subscriber or topic list crashes with
std::bad_alloc
A stack trace from the debugger gives