fix: release buffered data accounting when an incoming HTTP/2 stream is shut down

[pjfanning]

Motivation

IncomingStreamBuffer.shutdown() fails the entity outlet and drops whatever is still buffered, but unlike the other discard paths (onRstStreamFrame, onDownstreamFinish, fixed in #1259) it never subtracted those bytes from totalBufferedData.

One of its callers leaves the connection running: when a peer sends more data than the stream-level window allows, onDataFrame resets that single stream with FLOW_CONTROL_ERROR and the connection carries on. The bytes still buffered for the stream, plus the payload of the frame that is dropped without ever being buffered, keep counting as buffered for the lifetime of the connection.

The connection-level flow controller only emits a WINDOW_UPDATE while outstanding + buffered stays below half of incoming-connection-level-buffer-size, so the leaked accounting permanently reduces the replenishment of the connection window — and once it reaches half the buffer size, stops it entirely, at which point every stream on that connection stalls.

Modification

Release the buffer in IncomingStreamBuffer.shutdown() through the existing discardBuffer(), and subtract the payload of the frame that trips the stream-level window check, since that frame is never buffered either.

The other two shutdown() callers are GOAWAY paths where the connection is going away anyway, so the extra release is a no-op for them.

Result

Resetting a stream for a flow-control violation releases the connection-level window its data reserved, so the connection keeps being replenished and no longer stalls.

Tests

• sbt "http2-tests/testOnly org.apache.pekko.http.impl.engine.http2.Http2ServerSpec" — Not run - environment failure: http2-tests / update cannot resolve io.github.summerwind:h2spec_darwin_amd64:2.6.0 (h2spec_darwin_amd64.tar.gz not found under https://github.com/summerwind/h2spec/releases/download/v2.6.0/), which blocks the whole module locally. Needs the CI run — that is why this is a draft.
• New test release connection-level flow control accounting when a stream-level window is exceeded in Http2ServerSpec: a peer overruns the stream-level window with incoming-connection-level-buffer-size sized so the discarded frame is more than half of it, then the test asserts the connection window is handed back in full. Without the fix no WINDOW_UPDATE is emitted at all, so the expectation times out.
• The first CI run failed on the new test only, with requirement failed: incoming-connection-level-buffer-size must be > 0: the buffer size was a val in the anonymous setup class, so the superclass constructor read settings before it was initialised and saw 0. The constants now live outside that class, and the final check drains window updates with pollForWindowUpdates instead of expecting exactly one frame.
• scalafmt --mode diff-ref=upstream/main — pass.
• MiMa not run: internal impl.engine.http2 change with no public API or binary shape impact.

References

Refs #1259 — same accounting leak on the remaining discard path.