Write times fix for S3

[Jeremy Allison]

On Tue, Sep 09, 2008 at 05:50:06PM -0700, Jeremy Allison wrote:
> On Tue, Sep 09, 2008 at 07:31:47PM +0200, Stefan (metze) Metzmacher wrote:
> > Jeremy Allison schrieb:
> > > On Tue, Sep 09, 2008 at 08:35:27AM +0200, Stefan (metze) Metzmacher wrote:
> > >> Jeremy Allison schrieb:
> > >>> On Tue, Sep 09, 2008 at 12:57:12PM +1000, Andrew Bartlett wrote:
> > >>>> On Fri, 2008-09-05 at 19:00 -0700, Jeremy Allison wrote:
> > >>>>> Ok, here's the fix for the write times breakage
> > >>>>> with the new tests in S4 smbtorture.
> > >>>> Are you intending to fix Samba4's file server similarly?
> > >>> Not right now, I'm still working on torture tests to
> > >>> discover it Metze's comments are correct (they probably
> > >>> are :-).
> > >> I think most of your changes were indeed correct
> > >> and the v3-devel branch from yesterday evening (my time)
> > >> passes all tests I added yesterday.
> > > 
> > > Well the thing I'm trying to understand is if doing
> > > a SET_END_OF_FILE, SET_ALLOCATION_SIZE or SMBwrite
> > > with zero length is a "sticky" write time set, or
> > > a "normal" write time set. That's the part of the
> > > data model I still need to understand (and will
> > > probe with more tests).
> > 
> > Your vision of the data model seems correct,
> > they update the normal write time. See
> > f81014db9d5afbf9e0b1c007bc56fc1d3a201309
> > that demonstrates it for the SMBwrite
> > with zero length.
> 
> That commit has a bug. You're using 
> 
> smbcli_write(), and you need to be using
> smbcli_smbwrite() - a smbcli_write call
> ignores a zero-length write and has no
> effect.

FYI: Metze, I have a fix for this, and one for Samba
if it turns out the data model is different,
so don't waste time on it until I get back
online on Wed (Pacific time). I've been working
at home today with no access to test VM's
(still working on that :-) so no ability
to test stuff properly. I'll be back in
@ Google and should fix this tomorrow.

Jeremy.