PROBLEM:        (QAR 36520, QAR 48372)  (Patch ID: OSF375-350287)
********
Some valid programs compiled with ieee mode may receive a floating-point
exception even though they should run to completion.

The problem occurs because the floating-point exception handler is too strict
when attempting to enforce trap shadow rules (SRM 4.7.6.1).

The following C program fails when executed on a system with a DECchip 21064.

Compile the following program with the command:

cc -migrate -g -ieee_with_inexact t.c -o t

        f( int dum, ... ) {
        }

        main() {
            int ii = 50;
            f( 42,
                (float)ii / 100,
                (float)ii / 100,
                (float)ii / 100 );
        }

Results on a system with a 21046 chip before applying the patch:

quarry:ieee/48372 [4%] cc -migrate -g -ieee_with_inexact t.c -o t
quarry:ieee/48372 [5%] ./t
Floating exception(coredump)

Results after applying the patch:

sauron:ieee/48372 [2%] ./t
sauron:ieee/48372 [3%]