[jcifs] timezone problems

[Julian Reschke]

Christopher R. Hertel wrote:

> On Tue, Mar 09, 2004 at 03:01:08PM +0100, Julian Reschke wrote:
> 
>>Julian Reschke wrote:
> 
> :
> 
>>Running this script on W2K and Linux yields the following results:
>>
>>
>>Host: Linux, Files on Linux (Samba): java.io.File and jcifs.smb.SmbFile 
>>yield consistente results
> 
> 
> Good.
> 
> 
>>Host: W2K, File on W2K (Windows Networking): java.io.File returns 
>>consistent dates, but jcifs.smb.SmbFile breaks the timestampe for the 
>>"summer" file by substracting one hour
> 
> 
> Either that or there's a problem with the JVM on Windows.  More...

That was just the observation of the behaviour. As a matter of fact, 
removing the TZ-mappings from JCifs fixes the inconsistencies on the 
Windows platform.

>>Host: W2K, File on Linux (Samba): java.io.File returns inconsistent 
>>dates (file in summer is 60 minutes ahead), but jcifs.smb.SmbFile dates 
>>are consistent
> 
> 
> Your theory below is good, but I have another.  'Course, it's just a 
> theory...
> 
> In the above, you say that java.io.File returns an inconsistent date when 
> it accesses a file on a Samba server (I assume the Samba share is mounted 
> locally on the W2K system so that java.io.File can access the file).

Yes.

> If that's the case, then jCIFS is out of the loop in that test.  Your 
> theory that Samba is adjusting by 1 hour could still hold, but then we'd 
> see that problem in other situations.  For example, you'd see it when you 
> get a long listing in Windows Explorer.

As a matter of fact, I do:

Linux:

/home/public > date -r fsummer -u
Don Jun 22 00:00:00 UTC 2000
/home/public > date -r fwinter -u
Sam Dez 25 00:00:00 UTC 1999

W2K, Cygwin, Samba mount of /home/public:

//krusty/public> date -r fsummer -u
Thu Jun 22 01:00:00 UTC 2000
//krusty/public> date -r fwinter -u
Sat Dec 25 00:00:00 UTC 1999


(note the +1 hour difference for the file where lastmod falls into DST).


> I think I need to know more about this particular test, but it seems to me 
> that it's just as likely that the W2K JVM is doing the adjusting here.

With the example above, there's no Java VM involved at all.

> I also need to know which version of Samba you're running, and whether 
> other clients see the hour shift you're reporting.

Version 2.2.0 (on Suse)

And yes, all clients see the same timestamps.

> ...

Regards, Julian

-- 
<green/>bytes GmbH -- http://www.greenbytes.de -- tel:+492512807760