View Full Version : Take some old bugs out, put some fresh ones in

09-27-2007, 06:32 AM
"It was working, honest guv, but we've since improved it."

09-27-2007, 07:57 AM
In the linked MSDN blog post, commenters are falling over each other complaining about IEEE floating point inaccuracies. It's rather funny to watch.

09-28-2007, 12:28 PM
Only MS is this inventive. :rolleyes:

09-28-2007, 01:23 PM
It's only six numbers out of almost 10^19 that don't work. Good enough, I say. :D

09-28-2007, 01:24 PM
Least they didn't xray anyone to death.

09-28-2007, 03:48 PM
And the bug doesn't affect any formulas; it's just the display. Sure, it's a bug, but a fairly benign one. I just can't imagine how such a bug could pop up unless the display code is unnaturally convoluted.

09-28-2007, 04:09 PM
Which I wouldn't at all put past them. But I'm very puzzled, too.

09-28-2007, 07:10 PM
yeah that seems like a very odd bug to come up unless the person who programmed the code was drunk....I mean....how hard is it to say:

product = a * b;
output = product;

Rather...they want to say:

product= a * b;
if ( product== 65535 ) output = 100000;

09-28-2007, 08:42 PM
The bug has to do with conversion from binary to decimal for printing. If the binary result ends up being very, very close to 65535, but smaller, the conversion will fail. Maybe it has to do with lots of carried 1s?