[STDCXX-897] [MSVC] 18.numeric.special.float test fails

• Type: Bug
• Status: Closed
• Priority: Trivial
• Resolution: Fixed
• Affects Version/s: 4.2.1
• Fix Version/s: 4.2.2
• Component/s: Tests

Environment

MSVC

• Patch Info: Patch Available
• Severity: Incorrect Behavior
• Assignee: Farid Zaripov
• Reporter: Farid Zaripov
• Created: 29/Apr/08 17:23
• Updated: 12/May/08 15:05
• Resolved: 12/May/08 15:05

Description

The 18.numeric.special.float test fails with the following assertions:

# ASSERTION (S7) (4 lines):
# TEXT: _fpclass(numeric_limits<float>::signaling_NaN()) == 1 (_FPCLASS_SNAN), got 2 (_FPCLASS_QNAN)
# CLAUSE: numeric.special
# LINE: 318

# ASSERTION (S7) (4 lines):
# TEXT: numeric_limits<float>::has_denorm == -1, got 1
# CLAUSE: numeric.special
# LINE: 1360

# ASSERTION (S7) (4 lines):
# TEXT: numeric_limits<double>::has_denorm == -1, got 1
# CLAUSE: numeric.special
# LINE: 1360

# ASSERTION (S7) (4 lines):
# TEXT: numeric_limits<long double>::has_denorm == -1, got 1
# CLAUSE: numeric.special
# LINE: 1360

The first assertions is caused due to MSVC provides _fpclass() function for double type only and X87 and SSE floating point commands are incorrect converting float signaling NaN value to double type. So for example the float signaling NaN value with bitmask 0x7F800001, calculated in INFINITY config test, is converted to double value with bitmask 0x7ff8000020000000, which is a quiet NaN value.

The rest of the assertions are caused due to test expects, that numeric_limits<>::has_denorm == 1 on the AIX, __hpux and __osf_ platforms only.