cp: Do not dereference symlink even when dangling - fix issue #3364

[anastygnome]

Fixes an issue with cp not copying symlinks in spite of the -P (no dereference option)
fixes #3364

> ln -s no-such-file dangle
> LANG=C cp dangle d2
cp: cannot stat 'dangle': No such file or directory
[1]> LANG=C cp -P dangle d2
> rm d2
>  ./uu_cp dangle d2
cp: cannot stat 'dangle': No such file or directory
[1]> ./uu_cp -P dangle d2
> rm d2

[tertsdiepraam]

I think your test cases are backwards from your example in the PR. I.e. you do

cp (-P) link file # in the example
cp (-P) file link # in the test

It's probably best to test for both and check what the expected behaviour should be.

[tertsdiepraam]

Did you remove the tests from the PR? I thought they were really close to being correct and it would be great to have them!

[tertsdiepraam]

Don't worry about the failing Windows tests, I'm trying to fix those here: #3460

[anastygnome]

@tertsdiepraam I had forgot to add the tests. There are now back in the PR.

I also managed to get rid when removing the symlink, which should make the whole thing faster !

[anastygnome]

@tertsdiepraam @sylvestre
please resume the workflow

[sylvestre]

please rename the commit for something more explicit

[anastygnome]

will do

[anastygnome]

@sylvestre done

[anastygnome]

sorry for the hassle @sylvestre my PC is driving me nuts.

Just caught a nasty bug. fixed

[sylvestre]

yeah, please push when it is ready ;)

[anastygnome]

@sylvestre I'm done

[anastygnome]

Can we enable feature is symlink or do you want me to revert this part of the commit @sylvestre

[sylvestre]

For now, we haven't discussed about bumping MinRustV.

The error for others:
error[E0658]: use of unstable library feature 'is_symlink'

[anastygnome]

I've brought back the old method using metadata for now and added a comment, maybe you can relaunch the tests.

[anastygnome]

rebased to pass tests from main @sylvestre

[anastygnome]

@sylvestre failing test is unrelated, good to be merged

[sylvestre]

why did you push over my changes ?

[anastygnome]

@sylvestre thank the colleague who put force by default in my pull and push parameters.
fixed and brought back yours.

The touch command you removed in the test actually encovered a bug (if you try to run the test twice in a row you get an error 17 as the first dangling link is not removed prior to recreation.
I replace an .exists with a .is_file in cp.rs to fix it .

I also squash the add space commit to reduce noise

[anastygnome]

failing test has been corrected

[sylvestre]

thanks :)