hi，After my testing, the new modifications ran smoothly and the results met expectations. However, during the testing process, I discovered a new issue where I was unable to obtain $conid in the event_route [tcp: timeout] routing block. After analyzing, it was found that the find_conn method in the tcp_main. c file returned directly when it determined that the status of conn was BAD, and the relevant connection information was not successfully obtained. Given the relative proofreading of this issue, I will attempt to use the relevant logic from version 5.4.8 to rollback, mainly by removing the tcp_mait_closed_event call from the tcp_main. c file and adding this function call to the relevant code in tcp_read. c. After modification and testing, all issues disappeared, and the desired parameters could be obtained and executed normally within each close event. After observation, the relevant code should be executed in the child process, which does not have any impact on the main process. I think it is more reasonable. All the related functions of the new version have been placed in the main process for global management. The advantage is that the management is convenient and there will be no omissions. However, the disadvantage is that the main process does too much, and all the objects in the child processes need to have copies in the main process to be called normally. The main process is too cumbersome, and I don't think it's a good design. I haven't thought of a better solution for this while, The occupancy time can only be reverted back to the old version mode, which is currently more practical.

—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.