Debugging Unittest Hangs Checklist

Debugging Unittest Hangs Checklist | Generated by AI

Home 2025.10

Hangs in unittest are usually from deadlocks (threads/async), stuck I/O, subprocess/multiprocessing issues, or a test waiting on a network call. Here’s a tight checklist to see where it’s stuck and unstick it—works on Python 3.9–3.11.

1) Run unittest verbosely (and correctly)

# If xx.py calls unittest.main()
python xx.py -v

# Or use discovery (recommended)
python -m unittest -v
python -m unittest discover -v -s tests -p "test_*.py"

Note: python -v is interpreter import-verbose, not test-verbose. Use -m unittest -v for test names and progress.

2) Turn on faulthandler + dev mode (stack dumps on hang, stricter warnings)

# One-off
python -X faulthandler -X dev -u -m unittest -v
# Or via env
export PYTHONFAULTHANDLER=1
python -X dev -u -m unittest -v

-X faulthandler lets Python print thread stack traces on fatal signals/timeouts.
-X dev makes warnings/errors louder.
-u unbuffers stdout/stderr so you see prints in real time.

3) Force a traceback when it seems stuck

Option A — from another terminal (Linux/macOS):

kill -SIGUSR1 <pid>  # with faulthandler enabled, dumps all thread stacks

Option B — add to your test bootstrap (top of xx.py):

import faulthandler, signal, sys
faulthandler.enable()
# Dump stacks on SIGUSR1:
faulthandler.register(signal.SIGUSR1, all_threads=True)
# Also auto-dump if it hangs > 120s:
faulthandler.dump_traceback_later(120, repeat=True)

4) Trace execution step-by-step (heavy but decisive)

python -m trace --trace xx.py
# or
python -m trace --trace -m unittest discover -v

You’ll see every line executed; stop when the output “freezes”—that’s your hang site.

5) Use the debugger right away

python -m pdb xx.py         # if xx.py calls unittest.main()
# Break on a suspected line:
# (Pdb) b mymodule.py:123
# (Pdb) c

For discovery runs, add import pdb; pdb.set_trace() at the suspected spot.

6) Common causes & quick fixes

Multiprocessing on macOS/Windows: always guard test entry.

if __name__ == "__main__":
    import unittest
    unittest.main()

If you spawn processes in tests on macOS:

import multiprocessing as mp
if __name__ == "__main__":
    mp.set_start_method("fork")  # sometimes avoids hangs vs default "spawn"

(Only do this if you know your code is fork-safe.)

Thread deadlocks: check for join() without timeouts, locks not released, or background threads waiting on main thread. Replace:
```
t.join()
```
with:
```
t.join(timeout=5); assert not t.is_alive()
```
Async tests: ensure you’re using IsolatedAsyncioTestCase (3.8+) or asyncio.run(). Double event loops or forgotten await often hang.
Network calls: add short timeouts everywhere:
```
requests.get(url, timeout=5)
```
stdin waits: remove input() or guard it in tests; CI/stdin can block forever.
Mocking gone wrong: a mocked method that never resolves a Future/promise or a stubbed server not started.

7) Extra visibility

See allocations if leak/spin is suspected:

python -X tracemalloc=25 -m unittest -v

Show import noise (rarely needed for hangs):
```
python -v -m unittest -v
```

8) If it only hangs on one Python version

Try both:

pyenv local 3.11.9
python -X faulthandler -X dev -m unittest -v

pyenv local 3.9.19
python -X faulthandler -X dev -m unittest -v

Differences usually surface around multiprocessing, asyncio, or C-extension behavior.

If you paste the traceback from faulthandler (or the last lines from trace --trace), I’ll point to the exact culprit and suggest a targeted fix.

Back

openai/gpt-5

Donate