Small patch

[Darin Adler]

On Wednesday, July 30, 2003, at 10:39  PM, Bo Thorsen wrote:

> The problem is the max/min of long. I only know of one 64 bit 
> operating system that defined long as 32 bit (an MS one), everyone 
> else says it's 64 bit. But in this code, it's always treated like a 32 
> bit.

I think that's a misunderstanding of the code. There's nothing in the 
original code that's supposed to a "max and min of long". The "max" and 
"min" constants are the maximum and minimum values that are supposed to 
be stored as a simple number. They aren't supposed to actually be the 
max or min of any particular integer type.

The original code is supposed to work on systems where long is 64 bit 
and systems where long is 32 bit, and even some future system where 
long is some bigger size. Maybe there's a bug in it, but I don't see 
any evidence of that so far.

I recommend not landing the patch until you figure out if there's 
actually any bug here. You could make a test case that fails to 
demonstrate the original code is broken; that would sure convince me!

     -- Darin