11976 stop processing pipelined HTTP/1.1 requests that are received together

[glyph]

Scope and purpose

Fixes #11976

[adiroiban]

Not sure if this is ready for review.
I left some comments as I had a bit of time to look into this.

Thanks for the quick fix.

---

I think the macOS failure is flaky ... I did a manual retry

[glyph]

please review

[glyph]

sorry I forgot I need to summon the robot even for the first PR :)

[adiroiban]

The changes look good. Thanks for the quick fix.

I would prefer to have the new test added to t.w.test.test_http.PipeliningBodyTests , rather than test_web

This is the only major comment.

The others are minor.

Thanks again

[adiroiban]

why is this test added to test_web.py and not to test_http.py ?

Just asking.

[glyph]

test_http already has a test for this which doesn't cover the full stack top-to-bottom.

The boundaries are fuzzy, and this fix obviously has its foot in both worlds, but HTTP is for the HTTP protocol parsing, whereas web is for request handling. This is for the boundary where a parsed HTTP protocol message becomes an invocation of the request handling logic.

At first I felt like the problem might be at a higher level because there's already a test for this in test_http, before I realized that it was a TCP segment colocality issue. But, problems might be at a higher level in the future, so having the test live at this later, both in test_web and with server.Site being the layer of the public API getting tested, still makes sense to me.

[adiroiban]

thanks for the info

Maybe test_http.py needs a docstring that is more than Test HTTP support. same for test_web.py with Tests for various parts of L{twisted.web}.

I will keep this in mind and will try to update the docstring, next time I will touch these files

[adiroiban]

there was a flaky test failure on macOS . I manually triggered a retry.
I saw this test failing on macOS quite often.