I do not currently understand the failures of test_python_executable_option (assert 0 == 2) on various piplatest on GHA. tox -r -e "$(echo py{38,39,310,311,312,py3}-pip{previous,latest,main}-coverage | tr ' ' ',')" -- --no-cov -n=0 -v -k test_python_executable_option passes locally.

The test for this seems difficult to fit into the current testing framework—the test requires access to multiple (differing) interpreters—so I was hoping someone with more familiarity could help with that.

Basically, we might just as well use the same interpreter and a corresponding marker?

Basically, we might just as well use the same interpreter and a corresponding marker?

I don't think I understand, where are you suggesting putting a marker?

My understanding is that for the bug to occur, InstallRequirement.markers.evaluate(environment) needs to produce a different result with the pip-sync interpreter's environment and with the --python-executable=... interpreter's environment. AFAIK this is only possible if the two environments differ, and AFAIK that is only possible if they are not the same interpreter.

Applied review suggestions and fixed a crash on Windows.