http://bugzilla.novell.com/show_bug.cgi?id=536495
User poeml@novell.com added comment
http://bugzilla.novell.com/show_bug.cgi?id=536495#c5
Peter Poeml changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|CLOSED |REOPENED
Resolution|FIXED |
--- Comment #5 from Peter Poeml 2009-09-08 08:33:24 MDT ---
It seems that using the test for inode number wasn't such a good idea after
all. There are filesystems where small changes like changing permissions result
into an inode change (VirtualBox shared folder), and flexibility is reduced if
this needs to be observed. This bites us in the OpenOffice.org setup. Also,
this hinders moving files around locally.
I'm going to switch the scheme to use mtime & size, as rsync does, which should
be entirely sufficient.
It turns out that we didn't use the check for mtime to its full potential yet -
the mtime wasn't actually set on the hash files, and Apache would only check
that the hash file is newer than the file in question.
The proposed change is to copy the mtime to the the hash file, and make Apache
check that it's identical. In addition, instead of the inode we'll use the file
size as additional test, appended to the hash file name.
To avoid rehashing 1.7 TB of files, I'll migrate the existing hash files to the
new scheme, and copy all mtimes at that occasion, so that I can adjust Apache
at the same time.
--
Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.
