[mainline/4.0?] Fix 32-bit ieee tests on Linux/x86-64

[Mike Stump]

On May 26, 2005, at 3:30 PM, Mark Mitchell wrote:
> I'm not unwilling to be more restrictive if your position is in  
> fact the developer consensus.  But, with equal respect, my original  
> opinion remains unchanged as of yet.

A slightly different rant:

I'd like to see 0 unexpected failures on the primary platforms.  We  
can maintain this by requiring people to mark expected to fail  
testcases as expected to fail.  I'd like to see the regression  
checkers enforce this by flagging new tests that show up a unexpected  
fails.