Thanks for the contribution, but it looks like some of the test dependencies haven't supported Python 3.14 yet. 👀

Indeed. I didn't identify this issue locally because apparently I ran the test suite on the only platform where everything builds fine (macOS). Let me list the issues I'm parsing from log here:

• cmarkgfm fails to build, at least on Linux
• test_run_with_windows_python_version fails (not sure where the bug is coming from)

the relevant PR upstream is already merged theacodes/cmarkgfm#77
what's lacking is a release

The remaining failure looks genuine to me. I even think it was probably already discovered once, but hidden away by updating the test itself instead of fixing the bug (#1647). Of course, this is just my conclusion from a very quick review, and I could be entirely mistaken.

Looks like issues are essentially caused by pycowsay not working on Windows with newer Python. Probably a good idea to fix the tests to use a different package instead.

Hum, that's really surprising for a package that has a pure-python wheel.
I'm inclined to go and dig a little deeper before we resort to ditching it entirely, since it's used everywhere in pipx's documentation

The log from CI is not particularly useful however

ERROR    pipx.util:util.py:347 Fatal error from pip prevented installation. Full pip output in file:
    C:\Users\runneradmin\AppData\Local\Temp\pytest-of-runneradmin\pytest-0\test_do_not_wait_for_input0\subdir\pipxhome\logs\cmd_2025-11-03_02.34.38_pip_errors.log

is there any way I could retrieve this file or see its content ?

fixed

Cannot replicate the pycowsay problem on GitHub Actions...

• GitHub Action CI cs01/pycowsay#5

I'm aware. This is already being discussed above. I don't plan to dig further myself at the moment, so if that's what drafting the PR signals, you might as well just close it.