https://bz.apache.org/bugzilla/show_bug.cgi?id=63191
--- Comment #6 from Mark Thomas <ma...@apache.org> --- There are some error conditions that (as per the spec) mean that the javax.websocket.RemoteEndpoint.Async.sendText callback is never called. So yes, it is perfectly possible for the bug to be elsewhere. Attempts to send async messages concurrently is one of the error conditions that will trigger that. It bears repeating that the Tomcat committers don't know the CometD code. If the blocked CometD promise leads to the Async.sendText() call that isn't clear from the information provided to date. A stack trace that shows exactly how that happens would be helpful. It appears that you are using the NIO Connector. Can you confirm that? It appears that you are using Tomcat 9.0.14. Can you test with 9.0.16? There were some improvements in I/O error handling that might impact this issue. Unlikely, but worth ruling out. Does message size have any impact on how easy this is to reproduce? I have a possible theory. If it is correct, larger messages should make the problem easier to reproduce. -- You are receiving this mail because: You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
