Prevent readline from bugging out when the editline backend is used

[Viicos]

Fixes #12888.

Seems like having readline (or rlcompleter, which imports readline) imported at the root of the pdb module (as it is on 3.13) introduces the issue, but I'm not sure why. Perhaps the libedit backend tries to read something from stdin on startup? (Because running pytest -s works).

[The-Compiler]

Previous history:

• Prevent pyreadline from bugging out in --pdb mode with capture #1281
• Pytest run fails under Python 3.10.0-beta.2 on Windows #8733
• Pytest warning on readline #8847
• Drop readline workaround introduced in #1281 #8848

It would probably reintroduce the pyreadline breakage, but that seems fine to me, as that's just a broken and ancient setup at this point.

[Viicos]

It would probably reintroduce the pyreadline breakage, but that seems fine to me, as that's just a broken and ancient setup at this point.

Yeah I was thinking the same. We shouldn't expect pyreadline to be installed on any user environment :).

Should we investigate on why we get this failure with the editline backend and update the docstring?

[The-Compiler]

It might be worth having a minimal reproducer so this could be reported to CPython or libedit - other than that, if the workaround works, I think it's fine to use it as-is and close the issue with it. Seems harmless enough to me.

[nicoddemus]

Suggested change

	:mod:`pdb` module, as a side effect of importing :mod:`rlcompleter`.
	:mod:`pdb` module, as a side effect of importing :mod:`rlcompleter`.
	See #12888 for details.

[nicoddemus]

I'm also fine with the workaround as is if a test is not feasible.

Other than that we would need a changelog entry.