kern/91679: Re: setfacl file modification time

[craig at backfire.ca]

The following reply was made to PR kern/91679; it has been noted by GNATS.

From: craig at backfire.ca
To: bug-followup at FreeBSD.org
Cc: rwatson at FreeBSD.org
Subject: Re: kern/91679: Re: setfacl file modification time
Date: Fri, 17 Feb 2006 06:04:27 -0500

 > On Thu, 5 Jan 2006, Ceri Davies wrote:
 >
 > > On 5 Jan 2006, at 18:43, Ahnjoan Amous wrote:
 > >
 > >> In 5.2.1-RELEASE, setfacl updates the modification time of the file
 > >> when acls are changed.  I haven't been able to find any complaints
 > >> about this behavior, is this something folks on the list would expect
 > >> when using setfacl?  If so, does anyone know a work around?
 > >
 > > PR 76818 is open for this issue, but there is no progress logged at present.
 >
 > This likely occurs because EA writes modify the last modification timestamp.
 > I can imagine a couple of possible avenues to explore in fixing it, including:
 >
 > (1) Make system name space EA writes not update the modification time.
 >
 > (2) Add a flag to allow EA writes to be marked as not updating the
 >      modification time.
 >
 > It could be POSIX.1e has something to say about time stamp modifications for
 > setfacl; personally I'd expect it to match the behavior of chmod.  And I
 > believe chmod doesn't update the time stamp.
 >
 > Robert N M Watson
 
 I have not read the POSIX specification on this issue, nor do I have any
 authority on this matter, but I feel strongly that a file's mtime should not
 ever be changed by any filesystem utility, with the clear exception of
 /usr/bin/touch and perhaps others. If the POSIX spec requires that the mtime be
 changed during what is essentially a change to the file's metadata, I suggest
 that FreeBSD should diverge from the standard and document it accordingly.
 
 Cheers,
 
 -Craig Reyenga