keep websocket compression state across interleaved control frames

[arshsmith1]

What do these changes do?

When a permessage-deflate message is split across several frames, a peer may interleave a control frame (ping/pong/close) between the fragments (RFC 6455 §5.4). The reader uses _frame_fin to spot the start of a new message and (re)latch the per-message compression flag, but control frames always carry FIN, so an interleaved one makes the following continuation look like a fresh message and clears that flag. The continuation is then handled as uncompressed: raw deflate bytes are delivered for a binary message, or a UTF-8 error closes the connection for a text one.

The fix lets only data frames advance that tracker, so control frames stay transparent to the message framing and the compression state carries across them. The added test feeds a compressed first fragment, a ping, then the continuation and checks the binary message still decompresses.

Are there changes in behavior for the user?

No public API change. A compressed message that previously came back corrupted (or dropped the connection) when a control frame landed mid-message now decodes correctly.

Is it a substantial burden for the maintainers to support this?

No, it is a few lines in the reader plus one regression test.

Related issue number

N/A

Checklist

• I think the code is well written
• Unit tests for the changes exist
• Documentation reflects the changes
• If you provide code modification, please add yourself to CONTRIBUTORS.txt
• Add a new news fragment into the CHANGES/ folder

[Dreamsorcerer]

Is this ready? PR is still in draft.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.97%. Comparing base (4306796) to head (3bd36f6).
⚠️ Report is 39 commits behind head on master.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #12988      +/-   ##
==========================================
+ Coverage   98.96%   98.97%   +0.01%     
==========================================
  Files         131      131              
  Lines       48087    48747     +660     
  Branches     2496     2541      +45     
==========================================
+ Hits        47587    48247     +660     
  Misses        376      376              
  Partials      124      124              

Flag	Coverage Δ
Autobahn	22.12% <5.55%> (-0.11%)	⬇️
CI-GHA	98.90% <100.00%> (-0.01%)	⬇️
OS-Linux	98.67% <100.00%> (+<0.01%)	⬆️
OS-Windows	97.02% <100.00%> (-0.03%)	⬇️
OS-macOS	97.93% <100.00%> (-0.02%)	⬇️
Py-3.10	98.11% <100.00%> (-0.03%)	⬇️
Py-3.11	98.37% <100.00%> (-0.04%)	⬇️
Py-3.12	98.46% <100.00%> (-0.05%)	⬇️
Py-3.13	98.44% <100.00%> (-0.03%)	⬇️
Py-3.14	98.46% <100.00%> (-0.03%)	⬇️
Py-3.14t	97.55% <100.00%> (-0.03%)	⬇️
Py-pypy-3.11	97.43% <100.00%> (-0.02%)	⬇️
VM-macos	97.93% <100.00%> (-0.02%)	⬇️
VM-ubuntu	98.67% <100.00%> (+<0.01%)	⬆️
VM-windows	97.02% <100.00%> (-0.03%)	⬇️
cython-coverage	37.89% <100.00%> (-0.13%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

[codspeed-hq]

Merging this PR will not alter performance

✅ 83 untouched benchmarks
⏩ 83 skipped benchmarks¹

---

_{Comparing arshsmith1:ws-compress-interleaved-control (3bd36f6) with master (706ecdc)}

Footnotes

1. 83 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[arshsmith1]

Yeah, it's ready, just marked it out of draft. The fix, regression test, changelog and CONTRIBUTORS entry are all in and CI is green. Sorry for leaving it in draft.

[Dreamsorcerer]

Why are we not using the public .feed_data() here?

[arshsmith1]

Good point, no reason not to. Switched all three feeds to the public feed_data() and dropped a duplicate comment line that slipped in when I applied the suggestions. Suite still green (67 passed).