gh-71189: Support all-but-last mode in os.path.realpath()

[serhiy-storchaka]

• Issue: Support different modes in os.path.realpath() #71189

[serhiy-storchaka]

The original patch was created in 2016, but it has been rewritten from scratch several times as the realpath() code has changed several times over the years. The Windows implementation was created only now.

[zooba]

The change seems fine to me. I have no opinion on whether it's worth adding the functionality (calling realpath(dirname(p)) seems like a sufficient approach in any case that's useful - if I might need the real filename then I probably need to know whether it's real or not, so this option doesn't help me).

[serhiy-storchaka]

This is not so simple. The last component may be a link which refers to a path in which only the last component does not exist. For example:

$ ln -s dir/sub/file link
$ realpath linkrealpath: link: No such file or directory
$ mkdir dir
$ realpath linkrealpath: link: No such file or directory
$ mkdir dir/sub
$ realpath link/tmp/dir/sub/file

This is also the default mode of realpath, as it is the most useful.

[zooba]

Oh okay, so it's only for the case where the last segment is a link but the last segment of the target doesn't exist?

In your example above, what would realpath dir/sub/file show? Does it resolve dir/sub and then assume /file? Or does it fail because file doesn't exist?

[serhiy-storchaka]

It shows dir/sub/file if dir/sub exists and is a directory. It fails if dir or dir/sub doe not exist or are not directories.

This is yet one difference from resolving the parent -- it must also be a directory.

[zooba]

Makes sense. Still can't think of what I'd use it for, but being able to easily match a POSIX tool's defaults seems reasonable enough.

[bedevere-bot]

⚠️⚠️⚠️ Buildbot failure ⚠️⚠️⚠️

Hi! The buildbot x86-64 MacOS Intel NoGIL 3.x (tier-1) has failed when building commit 9d3b53c.

What do you need to do:

1. Don't panic.
2. Check the buildbot page in the devguide if you don't know what the buildbots are or how they work.
3. Go to the page of the buildbot that failed (https://buildbot.python.org/#/builders/1258/builds/5560) and take a look at the build logs.
4. Check if the failure is related to this commit (9d3b53c) or if it is a false positive.
5. If the failure is related to this commit, please, reflect that on the issue and make a new Pull Request with a fix.

You can take a look at the buildbot page here:

https://buildbot.python.org/#/builders/1258/builds/5560

Failed tests:

• test_subprocess

Failed subtests:

• test_leak_fast_process_del_killed - test.test_subprocess.POSIXProcessTestCase.test_leak_fast_process_del_killed

Summary of the results of the build (if available):

==

Click to see traceback logs

Traceback (most recent call last):
File "/Users/ec2-user/buildbot/buildarea/3.x.itamaro-macos-intel-aws.nogil/build/Lib/test/test_subprocess.py", line 3195, in test_leak_fast_process_del_killedself.assertRaises(OSError, os.waitpid, pid, 0)
~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^AssertionError: OSError not raised by waitpid