On 07/06/2019 10:51, Rémy Maucherat wrote: > On Fri, Jun 7, 2019 at 11:00 AM Mark Thomas <ma...@apache.org > <mailto:ma...@apache.org>> wrote: > > On 06/06/2019 22:51, Mark Thomas wrote: > > <snip/> > > > I haven't quite tracked down the other deadlock yet. The timing gap is > > narrower so it is harder to capture. I have some ideas to work on > tomorrow. > > This is the stack trace: > > Essentially, something should call notify on the NioOperationState > instance but it never happens. I'm still trying to figure out the root > cause. > > > This emulates a blocking write, so either there's really a problem > notifying or it is that the client is no longer reading and the wait > will eventually timeout. The state update seems properly synced and the > notify is called when the operation is considered complete, this looks > ok to me.
The only potential issue I found didn't seem to have any impact although I've fixed it anyway. The client might stop reading due to header corruption as a result of the bug I re-introduced with the first deadlock fix. I have a alternative fix that I'll commit shortly and then re-test. Mark --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
