Pierre De Rop

The previous commit avoids a buffer leak in the _WebsocketServerOperations_ class. So, In previous 4.x, in _WebsocketServerOperations_ class the _DefaultFullHttpRequest_ passed to the _WebSocketServerHandshaker.hanshake_ method was using unpooled byte buffer...

Pushed a fix in 618e866 previous commit, which avoids the following memory leaks caused by the NettyOutboundTest: (this PR will have to be rebased once the other #2484 is merged)....

in c1ead0f commit, fixed the following memory leak reproduced by the TcpServerTests.retryStrategiesWhenServerFails (sendGroups ...): ``` 16:22:52.136 [Cleaner-0] ERROR io.netty5.buffer.LoggingLeakCallback - LEAK: Object "buffer (20 bytes)" was not property closed before...

yes, will you do it or may I do ?

rebased on netty5, in order to get #2484 . now, if I'm correct we should only get some memory leaks for reactor-netty5-http module and the reactor-netty5-core module should pass (hopefully)....

@violetagg , this is a great improvement and optimization ! +1

the Netty PR (https://github.com/netty/netty/pull/12828), if accepted and merged, will normally resolve the following leak for the _HttpServerTests.nonContentStatusCodes_ test: ``` 22:42:36.850 [Cleaner-0] ERROR io.netty5.buffer.LoggingLeakCallback - LEAK: Object "buffer (0 bytes)" was...

The previous commit 9e29d50 fixes a memory leak reproduced by the HttpServerTests.testIssue1978H2WithDelay test. It was caused by the Http2StreamBridgeServerHandler.channelRead method, which was using old _ReferenceCountUtil.release(msg)_ instead of doing _Resource.dispose(msg)_

@simonbasle , Usually, we do not create separate PR: we first merge the PR into the **1.0.x** branch, then we merge it into **main** (for version 1.1.0), then after, we...

but I will let Violeta confirm.