This comment has been made anonymously so we may never get a reply, but anyway ...

How did you trace this to perm_copy_file()?  Are you sure this is not about copying the selinux labels?  Can you show how this looks in ltrace?

Note that it's normal to get an error message if some permission information cannot be preserved here; only if no information is lost will there be no error.

> It looks like the code in perm_copy_file.c is not checking for
> ENOTSUP if acl_delete_def_file() fails, unlike the
> acl_set_file() call.

This is after acl_set_file() has succeeded, in which case we can safely assume that acl_delete_def_file() will be supported as well.

> (The error checks after the chmod() call after acl_set_file()
> fails are a bit off too - the ret value from chmod() isn't
> checked, and saved_errno should be checked.)

Thanks, this should indeed be fixed.