If this PR is merged, I plan to also require unsetenv() to build and always make os.unsetenv() available, once https://bugs.python.org/issue39413 will be fixed.

🤖 New build scheduled with the buildbot fleet by @vstinner for commit d499e9a98444fb9dad78ca64f6ee2f42954db415 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

The PR built successfully on all tested buildbots: Linux (Ubuntu, Fedora, RHEL7, RHEL8, Gentoo, Arch Linux, ...), Windows (7, 8.1, 10, ...), macOS, FreeBSD.

I checked the 3 buildbot failures: they are unrelated.

Hum, I think I will wait until my PR #18163 is merged, so I can also require os.unsetenv() to be available in this PR.

I merged my PR #18163 so os.unsetenv() is now also available on Windows. I updated my PR to also require unsetenv() to build Python.

setenv() and unsetenv() should be available on all platforms supported by Python. If it's not the case, someone can maintain a downstream patch which is a revert of this PR.

Or we can revisit the idea of unsetting variables set by Python at exit, to workaround putenv() issue: https://bugs.python.org/issue39395

Or we can revisit the idea of unsetting variables set by Python at exit, to workaround putenv() issue: https://bugs.python.org/issue39395

I'm talking about something this PR #18078 that I just closed. But only if someone asks to support a platform which doesn't provide setenv() and unsetenv().