Show print output while pytest runs

For a normal serial run, use python -m pytest -s. The -s option is shorthand for --capture=no, which disables pytest's standard output/error capture.

python -m pytest -s
python -m pytest tests/test_api.py -s -v

Pytest normally captures stdout and stderr so successful tests do not flood the result summary. Captured output is typically shown with a failure report. A print statement that appears to do nothing may simply be captured.

Choose a capture mode

python -m pytest --capture=tee-sys
python -m pytest --show-capture=no
python -m pytest -q -s

tee-sys both captures and passes through writes to Python's sys.stdout and sys.stderr. It is not equivalent to teeing every low-level file-descriptor or subprocess write. The default fd capture covers a broader class of stream writes on supported systems.

--show-capture=no hides captured sections in reports; it does not disable capture. -q reduces pytest's own progress verbosity and is independent of application printing.

Disabling pytest capture cannot remove buffering in the application, subprocess, terminal, or CI log transport. Use print(..., flush=True) when an immediate Python-level flush is needed, while checking the rest of the pipeline too.

Configure logging separately

python -m pytest --log-cli-level=INFO

This enables live logging at the selected level while stdout/stderr can remain captured. Pytest's logging plugin and normal stream capture are separate mechanisms. -s does not automatically turn every logger's INFO or DEBUG records into visible output; logger levels, handlers, filters, and pytest logging settings still apply.

For assertions about printed text, the capsys fixture is often better than watching the terminal:

def test_message(capsys):
    print("Operation complete")
    captured = capsys.readouterr()
    assert captured.out == "Operation complete\n"

Capture fixtures such as capsys take precedence over the global capture setting, so the test above can still capture its output when the suite runs with -s.

FAQ

Do I need -s for pdb?

Pytest temporarily suspends capture when entering its integrated debugger, including normal breakpoint()/--pdb workflows. A serial run is easiest to inspect interactively.

Does -s stream prints from xdist workers?

No. Xdist does not forward worker stdout/stderr this way. Run the failing selection without xdist, or with -n 0, for ordinary interactive output. Captured failure reports and deliberately configured per-worker files are alternatives for distributed diagnosis.

Can I see logs while keeping print output captured?

Yes. Use live logging, such as --log-cli-level=DEBUG, without disabling stream capture. Keep output levels appropriate for the test run and avoid logging credentials merely to diagnose a missing print.

References: pytest output capture, pytest logging, pytest-xdist limitations.