Run prepare hooks for parser errors

[koriyoshi2041]

What

Parser-error responses now flow through the application system-route path when the request factory can attach app match info. This lets middleware handle the 400 response and lets on_response_prepare callbacks adjust headers such as Server.

Fixes #6463.

Tests

• python3 -m compileall -q aiohttp/web_app.py aiohttp/web_protocol.py aiohttp/web_runner.py tests/test_web_functional.py
• git diff --check
• PYTHONPATH=$PWD uv run --no-project --with pytest --with pytest-aiohttp --with pytest-mock --with pytest-timeout --with trustme --with cryptography --with proxy.py --with gunicorn --with uvloop --with dirty-equals --with coverage pytest tests/test_web_functional.py::test_signal_on_error_handler tests/test_web_functional.py::test_signal_on_parser_error_handler tests/test_web_functional.py::test_bad_method_for_c_http_parser_not_hangs -q

The C-parser-specific test is skipped locally because the C HTTP parser is unavailable in this checkout.

Risk

The change keeps the fallback protocol error handler for custom request factories that do not attach application match info.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.95%. Comparing base (69344c6) to head (c8d844a).
⚠️ Report is 73 commits behind head on master.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #12890      +/-   ##
==========================================
+ Coverage   98.94%   98.95%   +0.01%     
==========================================
  Files         131      131              
  Lines       47099    47861     +762     
  Branches     2435     2483      +48     
==========================================
+ Hits        46600    47359     +759     
- Misses        376      377       +1     
- Partials      123      125       +2     

Flag	Coverage Δ
Autobahn	22.28% <26.53%> (-0.12%)	⬇️
CI-GHA	98.92% <100.00%> (+0.01%)	⬆️
OS-Linux	98.68% <100.00%> (+0.01%)	⬆️
OS-Windows	97.04% <100.00%> (+<0.01%)	⬆️
OS-macOS	97.96% <100.00%> (+0.02%)	⬆️
Py-3.10	98.17% <100.00%> (+0.02%)	⬆️
Py-3.11	98.42% <100.00%> (+0.01%)	⬆️
Py-3.12	98.51% <100.00%> (+0.01%)	⬆️
Py-3.13	98.49% <100.00%> (+0.01%)	⬆️
Py-3.14	98.50% <100.00%> (+0.01%)	⬆️
Py-3.14t	97.59% <100.00%> (+0.03%)	⬆️
Py-pypy-3.11	97.44% <100.00%> (+0.02%)	⬆️
VM-macos	97.96% <100.00%> (+0.02%)	⬆️
VM-ubuntu	98.68% <100.00%> (+0.01%)	⬆️
VM-windows	97.04% <100.00%> (+<0.01%)	⬆️
cython-coverage	38.11% <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

✅ 79 untouched benchmarks
⏩ 79 skipped benchmarks¹

---

_{Comparing koriyoshi2041:fix-parser-error-prepare-hooks (c8d844a) with master (ff38c23)}

Footnotes

1. 79 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. ↩

[koriyoshi2041]

CI update: the code/test/doc checks are green now. The only failing check I see is Backport label added, which is failing because the PR has no label starting with backport.

The contributing docs say an aiohttp committer adds the backport label when a bugfix should be backported, so I am leaving that label choice to maintainers rather than adding backport:skip myself.

[Dreamsorcerer]

Doesn't this end up doing about the same thing as the old code? Seems like it's not going to solve the features requested..

[koriyoshi2041]

I think the difference is where this lets the app request handler back in for the parser-error request.

For the default AppRunner path, web_runner._make_request() sees message is ERROR and attaches request._match_info = MatchInfoError(HTTPBadRequest()). With this change, RequestHandler.start() builds that request first; when _match_info exists, it uses the app-level request handler instead of the protocol _make_error_handler fallback.

Then Application._handle() takes the existing system-route match info, builds middleware for SystemRoute, and returns the HTTPBadRequest response through the normal app path. finish_response() still calls resp.prepare(request), so on_response_prepare fires. The fallback error handler remains for custom request factories that do not attach match info.

I rechecked the targeted coverage locally:

PYTHONPATH=$PWD uv run --no-project --with pytest --with pytest-aiohttp --with pytest-mock --with pytest-timeout --with trustme --with cryptography --with proxy.py --with gunicorn --with uvloop --with dirty-equals --with coverage pytest tests/test_web_functional.py::test_signal_on_error_handler tests/test_web_functional.py::test_signal_on_parser_error_handler tests/test_web_functional.py::test_bad_method_for_c_http_parser_not_hangs -q

Result: 2 passed, 1 skipped; the skipped one is the local checkout missing the C HTTP parser.

[Dreamsorcerer]

Thanks for taking a look at this. Unfortunately, this task is deceptively complex and there's a number of details that need be addressed. I've created a new PR to resolve this issue instead: #12925.