a minor x86 bug reported

[David G. Hough at validgh]

Get the story at http://www.x86.org

Not in the same class as the infamous division bug, more of the sort of
common chip flaw listed in errata sheets.    It's a failure to report
the correct exception on conversion from floating-point format to
integer format.

More interesting is the assertion that the bug might have been masked
by nonstandard Fortran usage in one of Kahan's test programs written in
response to earlier x86 problems in this area.    This assertion seems
doubtful to me - chip people usually translate test programs to C before
they use them.