Re: "cp --no-preserve=mode" doesn't work as-expected

[Jim Meyering]

Matt Griswold <address@hidden> wrote:
> * Jim Meyering <address@hidden> [080425 08:55]:
>> Patches welcome.
>
> Apologies if I'm over-simplifying, I'm not familiar with the coreutils
> code.  This does seem to cause --no-preserve=mode to work as expected
> for both directories and files (although it keeps the source's x bit),
> and does pass all tests from make check.
>
> Patch against git trunk:
>
> --- a/src/copy.c        2008-04-25 12:21:57.597167502 -0500
> +++ b/src/copy.c        2008-04-25 12:35:28.799419214 -0500
> @@ -1624,7 +1624,11 @@
>       special mode bits may change after the directory is created),
>       omit some permissions at first, so unauthorized users cannot nip
>       in before the file is ready.  */
> -  dst_mode_bits = (x->set_mode ? x->mode : src_mode) & CHMOD_MODE_BITS;
> +  dst_mode_bits = (x->set_mode ? x->mode :
> +       (x->preserve_mode ? src_mode :
> +        src_mode | (S_IRUSR | S_IRGRP | S_IROTH)
> +        | (S_IWUSR | S_IWGRP | S_IWOTH)
> +       )) & CHMOD_MODE_BITS;
>    omitted_permissions =
>      (dst_mode_bits
>       & (x->preserve_ownership ? S_IRWXG | S_IRWXO
> @@ -1827,7 +1831,7 @@
>          used as the 3rd argument in the open call.  Historical
>          practice passed all the source mode bits to 'open', but the
> extra
>          bits were ignored, so it should be the same either way.  */
> -      if (! copy_reg (src_name, dst_name, x, src_mode & S_IRWXUGO,
> +      if (! copy_reg (src_name, dst_name, x, dst_mode_bits & S_IRWXUGO,
>                       omitted_permissions, &new_dst, &src_sb))
>         goto un_backup;
>      }

Thanks for the patch.
It's good to see someone else working on it.

However, this doesn't work how I would expect:

    $ ./cp -a --no-preserve=mode c c2; ls -glo c c2
    -rwx------ 1 85048 Apr 25 23:14 c*
    -rwxrw-rw- 1 85048 Apr 25 23:14 c2*

As you can see, the existing tests can be improved on this front.