Closed Crola1702 closed 1 year ago
So we can't actually do this revert without breaking other things (we'd also have to revert the similar PR in rclcpp, rmw_fastrtps, and rmw_connextdds). So I'm going to convert this to a draft for now and try to find out why this is failing.
What's happening is that this PR
RMW_EVENT_SUBSCRIPTION_INCOMPATIBLE_TYPE
event on the subscriberDDS_INCONSISTENT_TOPIC_STATUS
dds_set_status_mask
in gather_event_entities
(in rmw_cyclonedds_cpp
) because is_event_supported
(also rmw_cyclonedds_cpp
) says "sure"dds_set_status_mask
refuses the mask because this status is not supported on the reader and leaves the status mask whatever it is (everything enabled, probably)dds_wait
returns immediately because some status is set (subscription matched, or so)rclcpp::wait_for_message
tries to take data, finding nothing is available and returns false
test_wait_for_message.cpp:42
https://github.com/ros2/rmw_cyclonedds/pull/436 is relevant here, in that it follows the way Cyclone reports type incompatibilities. https://github.com/eclipse-cyclonedds/cyclonedds/issues/1523#issuecomment-1359605212 also goes into some detail.
I may well have reviewed something and declared it looked good because I overlooked this detail ... If so I apologise.
Hey @eboasson
I may well have reviewed something and declared it looked good because I overlooked this detail ... If so I apologise.
It is definitely not your fault, it's mine. I didn't test enough on Cyclone because I thought this was more-or-less a no-op. My mistake. Anyway, I definitely appreciate your analysis, and I'll see what I can do to make it happier. Thanks!
Given that we fixed this in #444, I'm going to close this out. Feel free to reopen if you think that was in error.
Reverts ros2/rmw_cyclonedds#431
As mentioned in https://github.com/ros2/rmw_cyclonedds/pull/431#issuecomment-1478510006. It seems this PR somehow made CycloneDDS fail on Rolling branch.
As CycloneDDS is at Tier 1 support level. I want to the PR until further investigation is done, and the issue is solved.
FYI: @clalancette @claraberendsen @Blast545