Don't re-await main_task in run_app finally when it's already done

[AndrewKarelin]

What

In web.run_app's finally block, only call main_task.cancel() and re-run loop.run_until_complete(main_task) when the task is still running.

 finally:
     try:
-        main_task.cancel()
-        with suppress(asyncio.CancelledError):
-            loop.run_until_complete(main_task)
+        if not main_task.done():
+            main_task.cancel()
+            with suppress(asyncio.CancelledError):
+                loop.run_until_complete(main_task)
     finally:
         _cancel_tasks(asyncio.all_tasks(loop), loop)
         loop.run_until_complete(loop.shutdown_asyncgens())
         loop.close()

Why

The cancel/re-await branch is intended for the GracefulExit / KeyboardInterrupt path, where the first run_until_complete returned without draining main_task. On the startup-failure path, main_task is already done() with the exception, and re-running it strips most of the traceback before it reaches the caller.

Future.result() on an already-finished task does raise self._exception.with_traceback(self._exception_tb). The with_traceback call resets exc.__traceback__ to the earlier-saved _exception_tb, which by this point reflects only the outermost frames. The deep frames the user actually needs (the cleanup_ctx/on_startup body, the user code that raised) are gone.

Skipping the second run_until_complete when the task is already done preserves the full traceback in the startup-failure case and is a no-op in the graceful-shutdown case (the task is still running there).

Versions affected

• Last good: 3.10.5
• First bad: 3.10.6 (regression)
• Still bad: 3.13.5 and current master

Reproducer

import traceback, sys
from aiohttp import web

async def bad_ctx(app):
    raise RuntimeError("boom from ctx startup")
    yield

app = web.Application()
app.cleanup_ctx.append(bad_ctx)
try:
    web.run_app(app, host='127.0.0.1', port=0, print=None)
except Exception:
    traceback.print_exception(*sys.exc_info())

Before this PR (3.10.6..3.13.5): 3 frames, no bad_ctx shown. Traceback stops at loop.run_until_complete.

After this PR: 9 frames, including:

File ".../aiohttp/web_app.py", line 600, in _on_startup
    await it.__anext__()
File "repro.py", line 5, in bad_ctx
    raise RuntimeError("boom from ctx startup")
RuntimeError: boom from ctx startup

The same chain you get today via asyncio.run(runner.setup()) — the bug is strictly in web.run_app's finally.

Test

Added test_run_app_preserves_startup_traceback in tests/test_run_app.py: registers a failing cleanup_ctx, asserts that the user function's frame appears in the traceback walked from the propagated exception.

Compatibility

• No public API change.
• GracefulExit/KeyboardInterrupt path: identical behavior (task is not done at that point, so the inner block runs as before).
• Startup-failure path: traceback now contains the actual cause, instead of stopping at loop.run_until_complete.
• Normal shutdown path: main_task is done, second run_until_complete is skipped — same observable behavior, slightly cheaper.

CHANGES file will be added once the PR number is assigned.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.92%. Comparing base (7d534ab) to head (85aba50).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@           Coverage Diff           @@
##           master   #12493   +/-   ##
=======================================
  Coverage   98.92%   98.92%           
=======================================
  Files         134      134           
  Lines       46750    46761   +11     
  Branches     2429     2430    +1     
=======================================
+ Hits        46248    46259   +11     
  Misses        373      373           
  Partials      129      129           

Flag	Coverage Δ
CI-GHA	98.98% <100.00%> (+<0.01%)	⬆️
OS-Linux	98.72% <100.00%> (-0.01%)	⬇️
OS-Windows	96.98% <100.00%> (+<0.01%)	⬆️
OS-macOS	97.89% <100.00%> (+<0.01%)	⬆️
Py-3.10.11	97.38% <100.00%> (-0.01%)	⬇️
Py-3.10.20	97.86% <100.00%> (+<0.01%)	⬆️
Py-3.11.15	98.12% <100.00%> (+<0.01%)	⬆️
Py-3.11.9	97.65% <100.00%> (+<0.01%)	⬆️
Py-3.12.10	97.73% <100.00%> (+<0.01%)	⬆️
Py-3.12.13	98.21% <100.00%> (+<0.01%)	⬆️
Py-3.13.13	98.44% <100.00%> (-0.01%)	⬇️
Py-3.14.4	98.50% <100.00%> (+<0.01%)	⬆️
Py-3.14.4t	97.52% <100.00%> (-0.01%)	⬇️
Py-pypy3.11.15-7.3.21	97.35% <92.85%> (-0.01%)	⬇️
VM-macos	97.89% <100.00%> (+<0.01%)	⬆️
VM-ubuntu	98.72% <100.00%> (-0.01%)	⬇️
VM-windows	96.98% <100.00%> (+<0.01%)	⬆️
cython-coverage	38.06% <0.00%> (-0.02%)	⬇️

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

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

[codspeed-hq]

Merging this PR will not alter performance

✅ 67 untouched benchmarks
⏩ 4 skipped benchmarks¹

---

_{Comparing AndrewKarelin:fix-run_app-traceback (85aba50) with master (7d534ab)}

Footnotes

1. 4 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. ↩