gh-80145: Exclude '.' directory when adding directories to a tar archive

[websurfer5]

As a result of the current tarfile behavior, shutil.make_archive (xxx, tar, root_dir) results in a './' entry in the archive, which is wrong.

• Issue: shutil.make_archive (xxx, tar, root_dir) is adding './' entry to archive which is wrong #80145

[taleinat]

I don't think the change to shutil warrants a mention in "What's New". The similar fix for zip archives did not add such a mention. The "What's New" line about the change to tarfile is fine IMO, though.

Also, since this changes the tarfile module (rather than shutil), this also should be tested in the tar module's tests.

[bedevere-bot]

A Python core developer has requested some changes be made to your pull request before we can consider merging it. If you could please address their requests along with any other requests in other reviews from core developers that would be appreciated.

Once you have made the requested changes, please leave a comment on this pull request containing the phrase I have made the requested changes; please review again. I will then notify any core developers who have left a review that you're ready for them to take another look at this pull request.

[taleinat]

Looks good to me.

This should also be reviewed by @gustaebel, the expert on the tarfile module.

[terryjreedy]

I changed the title to reflect what the patch does, as opposed the the symptom it prevents. This raises the question of whether it is always wrong for tarfile to add './', so that this is truly a bug fix, or only wrong when tarfile is called from shutil? In other words, is the bug possibly in the arguments passed by shutil, or tarfile not interpreting them fully? Tal Einat's approval of this PR was conditioned on an answer from a 'tarfile expert'. Lars is no longer reviewing. @serhiy-storchaka Can you decide this?

Otherwise, since the tests still pass after merging main, I think this good to go. The OP changed test_tarfile as Tal requested. I removed the 3.11 backport label and updated a comment from 'bpo' to 'gh'.

[gvanrossum]

Well, it would leave out the permissions ("mode") and other metadata for the current directory, which might be used when extracting (possibly using a different tar implementation) into a specific directory that doesn't exist yet but which that utility might create, setting the mode from the tarchive? So it seems to me this might not be 100% backwards compatible.

[nineteendo]

Isn't that the same as for zip archives? #72674

[serhiy-storchaka]

I am not sure that we should change tarfile.add(), currently it is consistent with the tar command which adds the '.' entry if it is explicitly specified. Also, while this PR removes the '.' entry, it leaves the './' prefix in other entries.

I created an alternate PR #118152 which only changes shutil.make_archive(). You can restore the old behavior if explicitly specify base_dir='.'.

[gvanrossum]

Agreed, tar itself does this too. What's special about shutil.make_archive() that it should be different? I'd prefer to be consistent and just reject the original issue.

[serhiy-storchaka]

It is special because we have no way to create a tar file without the '.' entry and the ./ prefix. With the tar utility we can at least call tar * instead of tar ..

[gvanrossum]

Surely we can do the equivalent of * using the glob module?

[serhiy-storchaka]

os.listdir() is simpler. We can do this inside of shutil.make_archive(), currently the user of shutil.make_archive() cannot affect this. This is what #118152 does -- changes the default behavior (to presumably more useful), while leaving the possibility to get the old result if the user really needs it.