Re: Derby-1 and Mac OS X

[Andrew McIntyre <mc...@gmail.com>]

On Apr 29, 2005, at 6:11 AM, Brian McCallister wrote:

> I'd personally like to see a fix for it, but that is as I work mostly 
> on OS X =)
>
> I am of the opinion that detecting OS X and modifying behavior 
> somewhat is okay as it *is* an exceptional case. It is like asking the 
> system for the pathname separator. It is a hack, but it is a hack 
> around a common (broken) JVM/Filesystem.
>
> In order to get around it, I apply the "OS X Fix" to all apps I use 
> Derby in, in case they are deployed on OS X. This is worse, but is so 
> much easier that it wins.

I'd like to see it fixed as well, especially since we already have a 
patch for the problem. JVM bugs like this are the rare exception, not 
the rule, and I personally don't see a reason not to include a fix that 
catches the exception and works around the problem. Suresh's patch has 
the benefit of fixing the problem on the one OS/JVM where it is a 
problem, with the new write sync functionality enabled everywhere else.

There was disagreement on how to fix the issue back then, but I wonder 
if there's as much resistance to the proposed fix now. I suppose anyone 
interested could call for a vote to have the patch committed, although 
the patch would need to be updated some. Suresh's original patch no 
longer applies without some minor modification.

andrew