Simone Bordet

> A single php-cgi.exe instance doesn’t facilitate a php loopback. Excuse my ignorance about Windows, but what does this mean exactly? I feel you are asking to implement php-fpm for...

> does the host and port fields in the HttpRequest have to be final? They are not anymore since #4787.

Sorry, but I don't understand what you're doing. Can you write a reproducible test case that we can run and that shows the problem? It's quite unlikely that connections end...

Sorry but the logging framework should deal with these issues, as `toString()` is entitled to lock and it must be possible to log with the same lock held. The problem...

@alekkol the same Jetty code with another logging library that is not Logback does not have any problem, because the other logging libraries do not have the Logback bug. It...

The lock in `toString()` is necessary for correctness, because it accesses internal state that is protected by that lock. We are not the only ones grabbing a lock in `toString()`,...

> That's why this issue was created. You have created it in the wrong project. It is not a Jetty issue. And we are not calling alien methods. We are...

The images you posted show 17 connections occupying a mere 7 MiB, and looks a completely normal state in case you have opened 17 connections to the server and sending...

Can you please take an `HttpClient` dump? You should cause the faulty state, and then trigger a call to `HttpClient.dump()` or `HttpClient.dumpStdErr()` and report the dump in this issue. It...

The `HttpClient` dump shows that the client is bombed with a large number of requests that it tries to send to backend destinations that are slow to process the requests....