Potential memory leaks in message_filters

[SteveMacenski]

Hi,

Per ros-navigation/navigation2#1889, we're seeing memory leaks in Navigation2 in servers that make use of message filters that we believe message filters is the root cause (potentially could be TF2 ROS within message filters).

[clalancette]

It would be interesting to retry the test with Rolling. We've fixed a few memory leaks in tf2_ros there, but we haven't been able to backport all of them to Foxy (since they break ABI; see ros2/geometry2#330 for an in-progress PR).

[SteveMacenski]

Would the master branch be acceptable as well? We don't use rolling, we run master nightly builds.

[clalancette]

Would the master branch be acceptable as well? We don't use rolling, we run master nightly builds.

Yeah, that would also work. It's preferable, honestly, since sometimes there is a delay between when we merge things and when it makes it to the Rolling packages.

[SteveMacenski]

I just don't use rolling on anything I maintain because I refuse to deal with the onslaught of emails it generates. I actually had to drop maintaining image pipeline for that reason (some small things break, suddenly I have 3 * number of packages in repo emails in my inbox, daily). I go out of my way to minimize the number of notifications in my life and that was raising my blood pressure and anxiety for tiny changes.

[clalancette]

I believe this should be fixed in Foxy and on the master branches now (via ros2/geometry2#330). @SteveMacenski can you confirm?

[SteveMacenski]

Yes, we closed it in regards to that patch, thanks for bringing that home