On 04/05/2015 19:49, Ognjen Blagojevic wrote: > On 3.5.2015 19:36, Rémy Maucherat wrote: >> 2015-05-03 18:28 GMT+02:00 Felix Schumacher < >> felix.schumac...@internetallee.de>: >> >>> In my test case position() will be either 0 or 341 (complete request). >>> They seem to happen at about the same rate.
Felix, Thanks for tracking this down. >>> r1672626 fixed a loop issue, but oversimplified. So I am going back >>> to the >> previous code, with an added fix to resolve the loop, and hopefully it >> will >> be fine now. >> >> Since there's something wrong in both cases, it's not really a regression >> and I doubt it is worth doing a new build for this. I agree 8.0.21 and 8.0.22 are both broken to some extent for NIO2 + TLS. > I repeated my smoke/stress tests against latest 8.0.x/trunk (r1677647), > and all connectors pass them, with and without TLS. > > I still get NIO2 warnings and APR 20005 errors in the log, as I reported > earlier, but everything else seems Ok. They seem to be unrelated to the > NIO2+TLS failure I reported with 8.0.22. > > Same warnings and errors exists with 8.0.21. > > I wouldn't say that 8.0.22 is not a regression compared to 8.0.21, as > 8.0.21 passes my tests, 8.0.x/trunk also passes them, but 8.0.22 does > not pass them. They are both broken, it is just more obvious with 8.0.22. I'll make my mind up finally tomorrow but I am leaning towards releasing 8.0.22 making it clear that there is a known issue with NIO2 + TLS and then looking at 8.0.23 almost immediately (I want to look at BZ 57802 first). Mark --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
