[Bug math/17097] New: ldbl-128 powl bad sign in overflow cases

["jsm28 at gcc dot gnu.org" <sourceware-bugzilla at sourceware dot org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=17097

            Bug ID: 17097
           Summary: ldbl-128 powl bad sign in overflow cases
           Product: glibc
           Version: 2.19
            Status: NEW
          Severity: normal
          Priority: P2
         Component: math
          Assignee: unassigned at sourceware dot org
          Reporter: jsm28 at gcc dot gnu.org

As noted in https://sourceware.org/ml/libc-alpha/2014-06/msg00834.html ldbl-128
powl produces results with the wrong sign for some overflow cases, e.g.:

Failure: Test: pow_downward (-0x2p+0, 0x1.ffffffffffffffffffffffffffffp+112)
Result:
 is:          1.18973149535723176508e+4932   0x1.ffffffffffffffffffffp+16383
 should be:  -inf  -inf

This applies for non-default rounding modes, and in the default rounding mode
if -lieee is used.

-- 
You are receiving this mail because:
You are on the CC list for the bug.