bug#17669: Fwd: Re: Solaris acl woes

[Ben Walton]

On Tue, Jun 3, 2014 at 7:27 PM, Pádraig Brady <address@hidden> wrote:
> On 06/03/2014 07:07 PM, Ben Walton wrote:
>>
>> On Jun 3, 2014 11:22 AM, "Pádraig Brady" <address@hidden 
>> <mailto:address@hidden>> wrote:
>>>
>>> On 06/03/2014 07:51 AM, Ben Walton wrote:
>>> > On Jun 2, 2014 6:46 PM, "Paul Eggert" <address@hidden 
>>> > <mailto:address@hidden>> wrote:
>>> >>
>>> >> [Forwarding this to Bug#17669 as bug-coreutils seems to have misfiled it
>>> > under 17664; closing 17664.]
>>> >>
>>> >> -------- Original Message --------
>>> >> Subject:        Re: Solaris acl woes
>>> >> Date:   Mon, 02 Jun 2014 06:56:03 -0700
>>> >> From:   Paul Eggert <address@hidden <mailto:address@hidden>>
>>> >> Organization:   UCLA Computer Science Department
>>> >> To:     Ben Walton <address@hidden <mailto:address@hidden>>, 
>>> >> address@hidden <mailto:address@hidden>,
>>> > address@hidden <mailto:address@hidden>
>>> >>
>>> >>
>>> >>
>>> >> Ben Walton wrote:
>>> >>
>>> >>> The lib/file-has-acl.c:acl_ace_nontrivial code that returns 1 is:
>>> >>
>>> >>
>>> >> Why is it returning 1, exactly?  What are the value of access_masks[0,
>>> >> 1] and how do they compare to the masks, and what bits are set that
>>> >> shouldn't be if we want the ACLs to be trivial?
>>> >
>>> > I didn't get back to this yesterday but will tonight.
>>> >
>>> > What do you think about the fact that the Solaris tools seem to exhibit 
>>> > the
>>> > same behavior?
>>>
>>> I'd probably adjust the tests to first:
>>>
>>> getfacl test.acl | setfacl -f - test.acl || skip_ "system is unable to copy 
>>> ACLs"
>>>
>>
>> Not a bad idea, but those tools have different names on different systems 
>> and possibly different calling conventions.
>>
>> If this is a preferred approach, at the very least, a presence check for the 
>> binary needs to wrap the precondition.
>
> We already have that in require_acl_
> Though yes it's very awkward that there is no standard here.
> This is how one copies ACLs on the systems I've just checked:
>
> solaris: getfacl file1 | setfacl -f - file2
> linux:   getfacl file1 | setfacl --set-file=- file2
> freebsd: getfacl file1 | setfacl -b -n -M - file2
>
> So not ideal at all.

Ok, let me see what I can bang together. I vaguely recall seeing other
code that was doing something along these lines so I might be able to
find a useful reference to start from.

>
> Which 6 tests did this affect BTW?

FAIL: tests/cp/backup-dir
  cp -a
FAIL: tests/cp/cp-parents
  cp -a
FAIL: tests/cp/parent-perm-race
  cp --preserve=mode
FAIL: tests/cp/preserve-link
  cp -a
FAIL: tests/cp/reflink-perm
  cp --preserve
FAIL: tests/cp/src-base-dot
  cp -a

Thanks
-Ben
-- 
---------------------------------------------------------------------------------------------------------------------------
Take the risk of thinking for yourself.  Much more happiness,
truth, beauty and wisdom will come to you that way.

-Christopher Hitchens
---------------------------------------------------------------------------------------------------------------------------