8335181: Incorrect handling of HTTP/2 GOAWAY frames in HttpClient

[Jaikiran Pai]

Hello Ryan,

The backporting of fixes into update releases is managed as a separate 
project, details of which are available here 
https://openjdk.org/projects/jdk-updates/. It has its own mailing list 
too https://mail.openjdk.org/mailman/listinfo/jdk-updates-dev.

-Jaikiran

On 10/10/24 12:18 am, Ryan Flegel wrote:
> Hi there,
>
> I'm wondering if someone would be able to review
> https://github.com/openjdk/jdk21u-dev/pull/1020 for me. There were a 
> couple small merge conflicts.
>
> I did try to cc this list from the PR, but that didn't seem to work 
> for me.
>
> ----
>
> I am backporting this fix because I am being affected by 
> [JDK-8335181](https://bugs.openjdk.org/browse/JDK-8335181). I had to 
> downgrade to HTTP/1.1 in order to use `HttpClient`.
>
> I needed to resolve two minor conflicts in `MultiExchange.java` due to 
> [JDK-8191494](https://bugs.openjdk.org/browse/JDK-8191494) being 
> applied slightly differently in different projects. In both cases, I 
> kept the change closer to upstream.
>
> The conflicts were:
> * Explicit vs implicit null field initialization
> * Different debug log message
>
> ### Tests All tests in the GitHub actions have passed.
>
> I have run the `jdk/java/net` tests locally and there were several 
> failures, including 4 within `java/net/httpclient`. However, I have 
> confirmed that all the failed tests also fail before applying the 
> patch, so they seem to be unrelated.
>
> Here is a full list of the failures:
>
> * java/net/DatagramSocket/B6411513.java: 
> java.net.DatagramSocket.receive: packet isn't received
> * java/net/DatagramSocket/SendReceiveMaxSize.java: This test verifies 
> that on macOS, the send buffer size is configured by default so that 
> none of our implementations of the UDP protocol will fail with a 
> "packet too large" exception when trying to send a packet of the 
> maximum possible size allowed by the protocol.
> * java/net/httpclient/ConnectTimeoutNoProxyAsync.java: Tests for 
> connection related timeouts
> * java/net/httpclient/ConnectTimeoutNoProxySync.java: Tests for 
> connection related timeouts
> * java/net/httpclient/ConnectTimeoutWithProxyAsync.java: Tests for 
> connection related timeouts
> * java/net/httpclient/ConnectTimeoutWithProxySync.java: Tests for 
> connection related timeouts
>
> UPDATE: I've run the rest of the tier2 tests and there were two 
> additional JDK failures that appear to be unrelated:
>
> * com/sun/jndi/dns/ConfigTests/Timeout.java: Tests that we can set the 
> initial UDP timeout interval and the number of retries.
> * com/sun/jndi/ldap/LdapPoolTimeoutTest.java: Multi-threaded client 
> timeout tests for ldap pool
>
> -- 
> Ryan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/net-dev/attachments/20241010/19168b40/attachment.htm>