svn commit: samba r3895 - in branches/SAMBA_3_0/source/smbd: .

[Jeremy Allison]

On Mon, Nov 22, 2004 at 04:58:31PM +0100, Volker.Lendecke at SerNet.DE wrote:
> 
> Seems to fix it. Although tridge's further tests show that this only glosses
> over the real problem. He has done a test using samba4 as a proxy towards W2k3,
> forcing an immediate update of the last write time stamp, and the problem did
> not happen. IIRC tridge's current theory is the lack of nanosecond resolution
> in time stamps. But this remains to be seen once samba4 has oplock
> functionality.

Well I'm not sure - what I committed is actually a change in functionality
that we weren't doing right before. It turns out that SetFileTime() is
sticky once set, and all subsequent writes don't change the write time
until the file is closed.

It may be nothing to do with the nanosecond resolution, and more
to do with the fact that a write after the SetFileTime would change
the timstamp (until my patch).

Just wanted to point that out :-).

Jeremy.