[MSVC] pow (double (x), double (int (y))) and pow (double (x), int (y)) returns
sometimes the different results on 64-bit MSVC
------------------------------------------------------------------------------------------------------------------------------
Key: STDCXX-899
URL: https://issues.apache.org/jira/browse/STDCXX-899
Project: C++ Standard Library
Issue Type: Bug
Components: External
Environment: 64-bit MSVC
Reporter: Farid Zaripov
Priority: Minor
The 26.c.math test fails on 64-bit MSVC with the following assertions:
{noformat}
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (-9.0, -8) = 2.32306e-08
# CLAUSE: lib.c.math
# LINE: 226
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (-9.0L, -8) = 2.32306e-08
# CLAUSE: lib.c.math
# LINE: 240
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (-9.0, -3) = -0.00137174, std::pow (-9.0, -3.0) = -0.00137174
# CLAUSE: lib.c.math
# LINE: 226
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (-9.0L, -3) = -0.00137174, std::pow (-9.0L, -3.0L) =
-0.00137174
# CLAUSE: lib.c.math
# LINE: 240
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (-6.0, -9) = -9.9229e-08
# CLAUSE: lib.c.math
# LINE: 226
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (-6.0L, -9) = -9.9229e-08
# CLAUSE: lib.c.math
# LINE: 240
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (-6.0, -6) = 2.14335e-05
# CLAUSE: lib.c.math
# LINE: 226
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (-6.0L, -6) = 2.14335e-05
# CLAUSE: lib.c.math
# LINE: 240
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (-3.0, -9) = -5.08053e-05
# CLAUSE: lib.c.math
# LINE: 226
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (-3.0L, -9) = -5.08053e-05
# CLAUSE: lib.c.math
# LINE: 240
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (-3.0, -6) = 0.00137174, std::pow (-3.0, -6.0) = 0.00137174
# CLAUSE: lib.c.math
# LINE: 226
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (-3.0L, -6) = 0.00137174, std::pow (-3.0L, -6.0L) = 0.00137174
# CLAUSE: lib.c.math
# LINE: 240
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (3.0, -9) = 5.08053e-05
# CLAUSE: lib.c.math
# LINE: 226
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (3.0L, -9) = 5.08053e-05
# CLAUSE: lib.c.math
# LINE: 240
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (3.0, -6) = 0.00137174, std::pow (3.0, -6.0) = 0.00137174
# CLAUSE: lib.c.math
# LINE: 226
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (3.0L, -6) = 0.00137174, std::pow (3.0L, -6.0L) = 0.00137174
# CLAUSE: lib.c.math
# LINE: 240
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (6.0, -9) = 9.9229e-08
# CLAUSE: lib.c.math
# LINE: 226
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (6.0L, -9) = 9.9229e-08
# CLAUSE: lib.c.math
# LINE: 240
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (6.0, -6) = 2.14335e-05
# CLAUSE: lib.c.math
# LINE: 226
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (6.0L, -6) = 2.14335e-05
# CLAUSE: lib.c.math
# LINE: 240
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (9.0, -8) = 2.32306e-08
# CLAUSE: lib.c.math
# LINE: 226
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (9.0L, -8) = 2.32306e-08
# CLAUSE: lib.c.math
# LINE: 240
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (9.0, -3) = 0.00137174, std::pow (9.0, -3.0) = 0.00137174
# CLAUSE: lib.c.math
# LINE: 226
# ASSERTION (S7) (4 lines):
# TEXT: std::pow (9.0L, -3) = 0.00137174, std::pow (9.0L, -3.0L) = 0.00137174
# CLAUSE: lib.c.math
# LINE: 240
{noformat}
The printed resulting values are the same, but actually they differs in one/two
least significant digits.
Since the problem in MSVC CRT, I suppose that the only way to get rid of these
asserts is convert them to rw_warn() when _RWSTD_STRICT_ANSI is not #defined.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
