Harden AuTests against timing races

[bneradt]

Two AuTests can fail depending on process scheduling and response
framing. The HTTP/2 chunked clients can reach their origins before the
listeners are ready, while the capped stale-response fetch can stop at
a body-block boundary and time out instead of exercising its memory
fallback.

This patch uses a deterministic raw origin that ignores readiness
probes, serves one real request, and exits. It also sizes and documents
the stale-response header at the cap-plus-sentinel read boundary so
memory-limit rejection is independent of body segmentation.

[Copilot]

Pull request overview

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Hardens flaky AuTest gold tests by reducing timing and framing sensitivity in chunked HTTP/2 origin tests and the stale-response memory-cap test.

Changes:

• Makes netcat-based origins persistent (nc -lk) to avoid one-shot listener races.
• Adds port-readiness gating via When.PortOpen(...) before starting HTTP/2 clients.
• Pads the stale-response header to align with a cap-plus-sentinel read boundary for deterministic memory rejection.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
tests/gold_tests/pluginTest/stale_response/stale_response_max_memory.replay.yaml	Adds a padding header to stabilize the memory-cap rejection boundary.
tests/gold_tests/chunked_encoding/server2.sh	Switches netcat listener to keep-open mode for the origin script.
tests/gold_tests/chunked_encoding/server3.sh	Switches netcat listener to keep-open mode for the origin script.
tests/gold_tests/chunked_encoding/delay-server.sh	Switches netcat listener to keep-open mode for the delayed origin script.
tests/gold_tests/chunked_encoding/chunked_encoding_h2.test.py	Adds port-open readiness checks before starting HTTP/2 clients.

[Copilot]

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated no new comments.

Suppressed comments (3)

tests/gold_tests/chunked_encoding/chunked_encoding_h2_server.py:53

• request += data (and later body += data) repeatedly concatenates immutable bytes, which is O(n^2) as the request grows. Switching to a bytearray (or collecting chunks in a list and b"".join(...) once) avoids quadratic behavior and makes this helper more robust if a larger request is ever sent during tests.

    request = b""
    while b"\r\n\r\n" not in request:
        data = conn.recv(4096)
        if not data:
            return request
        request += data

tests/gold_tests/chunked_encoding/chunked_encoding_h2_server.py:52

• recv() can block indefinitely if a client connects and sends partial headers (or stalls) without closing the connection, which can hang the AuTest run. To make the test harness failure mode bounded, set a reasonable socket timeout on conn (and handle socket.timeout by closing/continuing) so the server can recover instead of blocking forever.

    while b"\r\n\r\n" not in request:
        data = conn.recv(4096)
        if not data:
            return request

tests/gold_tests/pluginTest/stale_response/stale_response_max_memory.replay.yaml:84

• The hardcoded “215 bytes”/“42 wire bytes” calculations are brittle if header serialization changes (e.g., ordering, quoting, or YAML emitter differences), which could reintroduce flakiness. Consider adding a short note indicating how/where the 256+1 boundary is enforced in the code under test and that X-Padding must be re-tuned if serialization changes; alternatively, set X-Padding longer than necessary (with a comment stating the target is “≥ cap+1”) to reduce sensitivity to minor formatting differences.

        # The preceding fields serialize to 215 bytes. X-Padding adds 42 wire
        # bytes (name, separator, 29-byte value, and CRLF), bringing the header
        # to the 256-byte limit plus the one-byte overflow sentinel. This makes
        # the memory rejection independent of body segmentation.
        - [ X-Padding, aaaaaaaaaaaaaaaaaaaaaaaaaaaaa ]

[cmcfarlen]

Cherry-picked to the 10.2.x branch as 2696089 for the 10.2.0 release.