db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Andrew McIntyre <mcintyr...@gmail.com>
Subject Derby-1 and Mac OS X
Date Fri, 29 Apr 2005 09:23:54 GMT
Since it's April 29th where I am, I think it's safe for me to report  
that the latest version of Mac OS X - 10.4, aka "Tiger" - contains a  
fix for Derby-1. Here's the original Jira report:


The good news is that Derby now works without modification or setting  
system properties on Mac OS X 10.4, and more importantly, tests  
indicate that using "rws" and "rwd" modes for RandomAccessFiles results  
in data being written synchronously to disk as expected. Tests results  
using Jan Hlavaty's iotest program are attached at the end of this  
email. Kudos to Apple's JVM team for fixing the problem, and fixing it  

But, this does not mean that Derby-1 is resolved. If the longevity of  
Mac OS X 10.2 is any indication, Mac OS X 10.3 ("Panther") will be  
around for quite a while, and Derby-1 will continue to be a problem for  
developers and users of Derby there. In the original discussions from  
last September and October, several resolutions for the bug were  
proposed. Allow me to restate what appear to be the alternatives in  
dealing with this bug (and other JVM issues of its kind):

1) No fix in Derby. The problem should be documented, and Derby  
application developers should work around the issue in their specific  

2) Address the problem in Derby itself, in this case by catching the  
FileNotFoundException that was thrown. Suresh Thalamati posted a patch  
during the discussion of the original problem which worked around the  
issue by catching the FileFoundNotException thrown by Apple's JVM.

3) Have an exclusion/inclusion list to disable/enable features for JVMs  
known to have problems. An inclusion approach was proposed here:


and an exclusion approach was proposed here:


Over the course of the past six months, we've been implicitly forging  
ahead with approach #1 by simply not doing anything, and it seems to be  
working out, since we've not heard any further reports or discussion.

Is this the way we should continue? If so, then I think we should  
consider closing Derby-1 as Won't Fix and documenting the problem in  
the manuals as well as the FAQ.



Attaching test results for Mac OS X 10.3 (panther.txt) and Mac OS X  
10.4 (tiger.txt) from Jan Hlavaty's 'iotest' program:

View raw message