TCPTransportInterface hangs on clean() [13648]

[hoanglongmoovita]

When calling delete delete_participant on DomainPariticpantFactory on publisher side, the program hangs, which traces back to TCPTransportInterface::clean() -> channel->clear() -> thread_.join().

Expected Behavior

The clean up function should exit.

Current Behavior

The clean up function hangs forever

Steps to Reproduce

1. Create publisher and subscriber profile with tcp transport ( 2 different participants ).
2. Connect and send message.
3. Clean up by deleting participants. The subscriber exits normally, the publisher hangs forever

System information

• Fast-RTPS version: commit 5b8b905
• OS: Ubuntu 20.04
• Network interfaces: eth0 (192.168.1.1), lo (127.0.0.1)

Additional context

I tried removing the line in TCPChannelResourceBasic::disconnect():106 and call change_status(eConnectionStatus::eDisconnected) at the beginning of the disconnect() function and the clean() function does not hang anymore.

Additional resources

• XML profiles file: (had to change file extension to log as github does not allow xml file attachment
  default.log

[EduPonz]

Hi @hoanglongmoovita ,

I think we have solved this issue in branch https://github.com/eProsima/Fast-DDS/tree/bugfix/tcp_client_block. There is still some work to be done there like some regression tests, but I'd be nice if you could give it a try.

[JLBuenoLopez]

I am going to close the issue due to inactivity. The fix mentioned in the comment above has been already merged and backported to the supported versions of Fast DDS. Feel free to reopen if the issue is still happening in one of the supported Fast DDS releases.