shutil.copytree incorrectly does not copy symlink contents if called
with symlink=False and ignore_dangling_symlinks=True.

The wrong behaviour can be reproduced like this:

$ tree
.
└── a
    ├── a.txt
    └── b
        └── a.txt -> ../a.txt

$ python3 -c "import shutil;shutil.copytree('a/b', 'c', symlinks=False, ignore_dangling_symlinks=True)"

As a result directoy c will be created but it will remain empty.
Expected result is a file c/a.txt with the contents of a/b/a.txt.

This still seems to be an issue even though the underlying PR was closed. Even though ignore_dangling_symlinks is passed to recursive calls of copy_tree now, the function will still incorrectly classify symlinks as dangling even if they aren't because it attempts to do a os.readlink relative to the current working directory, not the copied directory

The issue is still present in Python 3.12.5 and must be fixed since copied data with enabled option symlinks=False and ignore_dangling_symlinks=True misses all resolved symbolic links file contents.

Fix in shutil _copytree could look like:

linkPath = linkto if os.path.isabs(linkto) else os.path.join(os.path.dirname(srcname), linkto)
# ignore dangling symlink if the flag is on
if not os.path.exists(linkPath) and ignore_dangling_symlinks:

@hugovk since you closed my duplicate: may I ask when this fix will finally go into Python 3.12 or 3.13? It is annoying to always fix the one line for each new python release to get the correct behavior. Thanks upfront!

Hello! The linked PR was closed before merge:

Seems that this has been solved in #22937 and it is also a duplicate of #82704. Therefore I close this PR.

If it's still a problem, someone needs to submit a PR first, then we can merge that into main/3.14, and then we can backport it to 3.12 and 3.13, and it would go out in the subsequent releases. So there's currently no timeline on a fix.

Are you interested in creating a PR to fix it, with tests?

Hi @hugovk , thanks for the reply - since I am not familiar with the working style I would prefer the experts doing it.
However, since the bug is still present in Python 3.14 at least one ticket should be available - will you take care here that this small fix is not lost again?
Thanks and best regards!

I'm not planning on working on the fix, but this is the ticket for it.

Hi all,
any plan when to fix this issue?
I consider it as quite critical since the resulting data is not as intended.