[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: Peculiar date problem with checkout of zero-length files

From: Mark Phippard <markphip_at_gmail.com>
Date: Fri, 9 Oct 2009 08:07:18 -0400

On Thu, Oct 8, 2009 at 6:33 PM, David Brodbeck <brodbd_at_u.washington.edu> wrote:
> On Oct 8, 2009, at 12:32 PM, Mark Phippard wrote:
>> On Thu, Oct 8, 2009 at 3:22 PM, David Brodbeck <brodbd_at_u.washington.edu
>> > wrote:
>>> I have a RedHat Linux server that mounts a filesystem via NFSv4 from
>>> an OpenSolaris system.  I'm seeing very peculiar behavior when I
>>> check
>>> out zero-length files using Subversion 1.5.5.  These files end up
>>> with
>>> random dates, often far in the past or future -- I've seen dates in
>>> 1917, 1955, 1958, 1987, and 2013.  Naturally this causes Subversion
>>> to
>>> report "bogus date" errors.  I've now reproduced the behavior on
>>> three
>>> different systems using two different Linux distributions (Ubuntu and
>>> RedHat) and both 32-bit and 64-bit kernels.  Some interesting
>>> details:
>>> - The bug DOES NOT occur with Subversion 1.4.2.
>>> - The bug DOES occur with Subversion 1.5.5 and 1.6.5.
>>> - The bug DOES NOT occur when checking out to an NFSv3 filesystem
>>> or a
>>> local ext3 filesystem; it only seems to occur with NFSv4.  (Possibly
>>> ACL related?)
>>> - The bug DOES NOT occur when both the client and server are
>>> OpenSolaris machines.
>>> Has anyone seen anything like this before?  Before I noticed that the
>>> behavior didn't occur in Subversion 1.4.2, I reported it to RedHat as
>>> a possible kernel bug (#527580), but now I'm wondering if later
>>> versions of Subversion are doing something peculiar when they create
>>> zero-length files.
>> It might be worth comparing the versions of APR used by the different
>> versions that are working and not working.  Subversion uses APR
>> functions for most (perhaps all) file related routines.
> Hi, Mark.
> That's an excellent thought.  I checked the copy of Subversion 1.5.5
> (which exhibits the bug) and the copy of Subversion 1.4.2 (which
> doesn't) that I used, and found both are linked against /usr/lib/
> libapr-1.so.0, so I don't think it's that.

Unless these are on the same physical system and linking to the exact
same library, then you should try to figure out the exact version.

Mark Phippard
To unsubscribe from this discussion, e-mail: [users-unsubscribe_at_subversion.tigris.org].
Received on 2009-10-09 14:08:24 CEST

This is an archived mail posted to the Subversion Users mailing list.