On Fri, 30 Oct 2026 13:40:50 GMT, Timofei Fedotov <[email protected]> wrote:
>> Hi @Sovtouch ! Are you currently employed by a company which signed Oracle >> Contributor Agreement (OCA), or you are contributing as an individual? We >> may need to insert a company's copyright header in the former case. > > Hi @artur-oracle, I am contributing patch as an individual under my signed > individual OCA. @Sovtouch Consider fragments with mixed message types sharing a sequence number. All fragments have `epoch 0` and `message_seq = 0`, with record sequence numbers 0, 1, and 2, respectively: | Arrival | Handshake type | Message length | Offset | Fragment length | | ------- | --------------- | -------------- | ------ | --------------- | | 1 | `SERVER_HELLO` | 100 | 5 | 50 | | 2 | `HELLO_REQUEST` | 10 | 0 | 5 | | 3 | `HELLO_REQUEST` | 10 | 5 | 5 | The original version of your fix handled such scenario by having `previous.handshakeType != hsf.handshakeType` check, but in current version the above fragment sequence should result in `ArrayIndexOutOfBoundsException`. We would need a test coverage for such scenario as well. ------------- PR Comment: https://git.openjdk.org/jdk/pull/33633#issuecomment-6532395830
