https://bz.apache.org/bugzilla/show_bug.cgi?id=60750
--- Comment #21 from Christopher Schultz <ch...@christopherschultz.net> --- (In reply to Konstantin Kolinko from comment #20) > the correct behaviour after an OOM like that is to abort the connection, > without any terminating chunk. Okay, then Tomcat 8 is non-compliant. > OOM = the processing failed, the client should note that the failure > occurred, e.g. that the document that the person downloads is truncated. > > Sending the terminating 0-length chunk gives an illusion that everything is > OK, while it certainly is not. > > > See r1602443 and CLOSE_NOW action that I mentioned in Comment 9. > > Discussion that lead to that change is in thread > titled "Error handling" on dev@, started 2014-05-29. > http://tomcat.markmail.org/thread/znillhttbmvsl5e5 If the desired behavior is to close the connection without a terminating 0-byte chunk (which is reasonable behavior IMO), then I think Tomcat should be consistent across versions. But I have now published a simple reproducible test case that can be used across Tomcat versions to verify the behavior claimed in this bug, so that should make it easier to test should a decision be made either way. -- 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
