[TarFile]: normalize also concatenated paths (#1072)

[grrtrr]

The result of normalize_path() may be a non-normalized path, containing e.g. ../, as a result of prepending self.directory.

Avoid the problem by using normpath() on the result. Resolves #1072.

[aiuto]

This makes one of the tar tests fail AssertionError: 'loremipsum.txt' != './loremipsum.txt'
So there is some subtle different behavior. We need to understand how that might change existing workflows.

What if you just normalized --directory?

[grrtrr]

This makes one of the tar tests fail AssertionError: 'loremipsum.txt' != './loremipsum.txt' So there is some subtle different behavior. We need to understand how that might change existing workflows.

What if you just normalized --directory?

I don't think the test failures are related. Normalizing --directory might work, doing this inside the function seemed to make the most sense, consistent with the earlier call to normpath().

[aiuto]

Can you try a merge from head and re-push so we can run with current tests?

[grrtrr]

Can you try a merge from head and re-push so we can run with current tests?

Yes, just refreshed.

[tonyaiuto]

I don't understand why you are still getting bazel7 in the tests.
So I patched it in and I see the same failure on my machine

AssertionError: 'loremipsum.txt' != './loremipsum.txt'
- loremipsum.txt
+ ./loremipsum.txt
? ++
 : Value `loremipsum.txt` for key `name` of file loremipsum.txt in archive /private/var/tmp/_bazel_tony/f3f83a225d68142b93b7d515eb69bd02/sandbox/darwin-sandbox/409/execroot/_main/bazel-out/darwin_arm64-fastbuild/bin/tests/tar/pkg_tar_test.runfiles/_main/tests/tar/test_tar_leading_dotslash.tar does not match expected value `./loremipsum.txt`

This is going to break people because the expect the leading ./ on their files.
How about if you just solve the one problem and make it
dest = (self.directory + dest).replace("/../", "/")

[grrtrr]

@aiuto - apologies, I had messed up the branch by using the 'merge' button.

Also researched (with the help of Claude) what was causing the problem - normpath() strips the leading './', which it should not:

input	os.path.normpath
'./loremipsum.txt'	'loremipsum.txt'
'./a/./b'	'a/b'
'./'	'.'
'../a'	'../a' (leading .. kept)
'./../a'	'../a'

Hence updated code to accommodate that. Tests now pass.