Johannes Rudolph

I still think this is a useful change (not least because it is somewhat of a simplification). Would be good to run some benchmarks. I'll see if I get to...

Thanks a lot, @maxim, I can reproduce the issue with your project. Here's another stack trace I gathered during the looping condition: ``` "server-akka.actor.default-dispatcher-21" #67 prio=5 os_prio=0 cpu=127637.79ms elapsed=149.42s tid=0x00007f6c2c001000...

> > Could it be same as #28092 ? > > Possibly - we expected that to be fixed by #29121 (released in 2.6.6) but perhaps we missed a case...

For me this seems to be a JDK bug where the SslEngine (after a previous failure) can get into a state where it produces no output but still reports `NEED_WRAP`...

I stepped through the JDK 11.0.10 code and found this: * engine.getHandshakeStatus() returns `NEED_WRAP` if `!outputRecord.isEmpty` (i.e. there's output ready to be sent, that's the case here) * on the...

It also happens with JDK 8. In JDK 8, throwing away the first 2 packets arriving at the client seems to reliably trigger the behavior.

Also filed at #30106. It doesn't seem to fail at my machine. I see if I can reproduce it somehow.

I found [this](https://github.com/AdoptOpenJDK/openjdk-jdk11u/commit/8d1b63a4db2c6348a97b3cf45bd4d2caa7cad6b5#diff-92fd40b83d43217042269d1a80fdc42abee7430318070b903568e44e379b1197) related bug fix that came in with 11.0.8. Jenkins runs 11.0.3. I could reproduce that the failure doesn't happen any more starting with JDK 11.0.8. What should...

> I wouldn't mind that we updated the version on Jenkins, but if the original problem is fixed in latest JDK 8 and 11 versions we should maybe just revert...

> I made workaround for this problem by implementing custom `SSLEngine` with handshake timeout. It spins in infinite loop up to a minute, and then throws `TimeoutException`. Not a good...