On Thu, 12 Nov 2026 12:21:50 GMT, Daniel Jeliński <[email protected]> wrote:
>> Roland Mesde has updated the pull request incrementally with one additional
>> commit since the last revision:
>>
>> Remove unused prepareToRelease, slicedToDataFrame and frameProcessed()
>> from FramesDecoder
>
> src/java.net.http/share/classes/jdk/internal/net/http/Http2Connection.java
> line 947:
>
>> 945: Log.logError(malformedFrame.getMessage());
>> 946: int errorStream = malformedFrame.getErrorStream();
>> 947: if (errorStream == 0) {
>
> use the new isConnectionError getter here?
@djelinski Done.
> src/java.net.http/share/classes/jdk/internal/net/http/Http2Connection.java
> line 957:
>
>> 955: debug.log(() -> "Reset stream " + errorStream +
>> ": "
>> 956: + malformedFrame.getMessage());
>> 957: stream.onProtocolError(new ProtocolException(
>
> why did you replace `resetStream` with the `isTerminalState` /
> `onProtocolError` combo?
@djelinski resetStream() alone would cancel the request with the server but
never report it to the caller, who would wait forever. onProtocolError()
cancels it and fails the caller's request with a ProtocolException. We also
don't cancel requests that have already finished or never started as indicated
in RFC 9113 (5.1, 6.4).
> src/java.net.http/share/classes/jdk/internal/net/http/frame/FramesDecoder.java
> line 553:
>
>> 551: private Http2Frame parseAltSvcFrame(int frameLength, int
>> frameStreamid, int frameFlags) {
>> 552: if (frameLength < 2) {
>> 553: skipBytes(frameLength);
>
> no need to skipBytes if you're returning a connection-closing malformedFrame.
@djelinski Done.
> src/java.net.http/share/classes/jdk/internal/net/http/frame/FramesDecoder.java
> line 559:
>
>> 557: var len = getShort();
>> 558: if (len > frameLength - 2) {
>> 559: skipBytes(frameLength - 2);
>
> same here
@djelinski Done.
> src/java.net.http/share/classes/jdk/internal/net/http/frame/MalformedFrame.java
> line 73:
>
>> 71: }
>> 72:
>> 73: public int getErrorStream() {
>
> could you drop the errorStream field and getter and use the streamId
> inherited field instead? Keeping two fields (one unused) for the same
> information (stream id) was the root cause of this bug, after all.
@djelinski Done.
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/33824#discussion_r4553071603
PR Review Comment: https://git.openjdk.org/jdk/pull/33824#discussion_r4553070546
PR Review Comment: https://git.openjdk.org/jdk/pull/33824#discussion_r4553075662
PR Review Comment: https://git.openjdk.org/jdk/pull/33824#discussion_r4553074484
PR Review Comment: https://git.openjdk.org/jdk/pull/33824#discussion_r4553072651