mv hardlinks should be preserved when moving into a different directory

[sylvestre]

Testcase:

export C=/usr/bin/
export C="target/debug/coreutils "
dir=hlink
other_partition_tmpdir=/dev/shm/tmp769306

rm -rf $dir $other_partition_tmpdir
mkdir $dir $other_partition_tmpdir
> $dir/a
ln $dir/a $dir/b

${C}mv $dir "$other_partition_tmpdir"

ls -1i "$other_partition_tmpdir/$dir"

They should be equal. They aren't currently

[Skallwar]

I've try to reproduce this:

❯ mkdir test
❯ mkdir test2
❯ > test/a
❯ ln test/a test/b
❯ target/release-small/coreutils mv test test2
❯ ls -1i test2/test/
4219332 a
4219332 b

Does your test only work when moving the folder to another partition?

[tertsdiepraam]

That's very likely, mv can't do a simple rename across partitions, so it has some fallback procedure that's probably not working as expected here.

[Skallwar]

Ok thanks for your input, I will try to reproduce this with an USB key.

EDIT: I got the test working, I will try to fix this then

[Skallwar]

I've tracked down the issue to move_dir here. But this function comes from webdesus/fs_extra so I will have a look there

[Skallwar]

There is a similar issue for symlink and a PR was proposed without reaction from the author. I wonder if we should "just" implement it directly on our side...

[allaboutevemirolive]

I've try to reproduce this:

❯ mkdir test
❯ mkdir test2
❯ > test/a
❯ ln test/a test/b
❯ target/release-small/coreutils mv test test2
❯ ls -1i test2/test/
4219332 a
4219332 b

Does your test only work when moving the folder to another partition?

Why do I get an error when trying to move the test into test2, but it is OKAY when using GNU mv?

nemesis@nemesis:~/Documents/Github/test/bin_3$ /home/nemesis/Documents/Github/Focus/util/coreutils_master/target/release/coreutils mv test test2
mv: cannot move 'test' to a subdirectory of itself, 'test2/test'

[cakebaker]

@allaboutevemirolive it's a bug, I opened a ticket for it: #5481

[mfontanaar]

There is a similar issue for symlink and a PR was proposed without reaction from the author. I wonder if we should "just" implement it directly on our side...

I guess this is probably not a good first issue anymore then, as it involves a project management decision. Perhaps the label should be dropped or at least direction about how to proceed with this provided.

Also, fs_extra's maintainer seems to have little time for continuing the project. Could that become a problem as uutils keeps expanding?