homepage

On "NetBSD 5 i386 3.x" buildbot, testTanh() of test_math fails because the sign of wrong.
configure scripts has a test to check if tanh(-0.0) keeps the sign or no, and on this buildbot the result is "no". pyconfig.h contains a TANH_PRESERVES_ZERO_SIGN define, but it is not used in Modules/mathmodule.c. Extract of configure comment:
# On FreeBSD 6.2, it appears that tanh(-0.) returns 0. instead of
# -0. on some architectures.
---
http://www.python.org/dev/buildbot/builders/NetBSD%205%20i386%203.x/builds/237/steps/configure/logs/stdio
...
checking whether tanh preserves the sign of zero... no
...
http://www.python.org/dev/buildbot/builders/NetBSD%205%20i386%203.x/builds/237/steps/test/logs/stdio
======================================================================
FAIL: testTanh (test.test_math.MathTests)
----------------------------------------------------------------------
Traceback (most recent call last):
 File "/home/buildbot/buildarea/3.x.smith-netbsd-i386/build/Lib/test/test_math.py", line 898, in testTanh
 math.copysign(1., -0.))
AssertionError: 1.0 != -1.0
----------------------------------------------------------------------

test_cmath does also fail:
http://www.python.org/dev/buildbot/builders/NetBSD%205%20i386%203.x/builds/237/steps/test/logs/stdio
======================================================================
FAIL: test_specific_values (test.test_cmath.CMathTests)
----------------------------------------------------------------------
Traceback (most recent call last):
 File "/home/buildbot/buildarea/3.x.smith-netbsd-i386/build/Lib/test/test_cmath.py", line 348, in test_specific_values
 msg=error_message)
 File "/home/buildbot/buildarea/3.x.smith-netbsd-i386/build/Lib/test/test_cmath.py", line 94, in rAssertAlmostEqual
 'got {!r}'.format(a, b))
AssertionError: tanh0002: tanh(complex(-0.0, 0.0))
Expected: complex(-0.0, 0.0)
Received: complex(0.0, 0.0)
Received value insufficiently close to expected value.
----------------------------------------------------------------------

I'm not sure what can reasonably be done about these failures. The Python test is failing because of a deficiency in the system math library.
Adding extra code to Python to handle this (making use of TANH_PRESERVES_ZERO_SIGN) is an option, but it seems like overkill to me.
I definitely don't want to remove these tests---they're valuable. If there's a way of marking the test as an expected failure on NetBSD 5 then that would be fine, I guess.
We should also report the problem upstream.

Here is a patch skipping math and cmath tests if TANH_PRESERVES_ZERO_SIGN is 0.

Commited to Python 3.2 (r86299) after a review of Mark Dickson on IRC. Thanks Mark ;-)

Nice fix! Thanks.

P.S. Greg, as the owner of the buildbot, do you feel like reporting this upstream? I get the impression that it's easier to do the reporting directly from the NetBSD machine in question...

> P.S. Greg, as the owner of the buildbot, do you feel like reporting this
> upstream? I get the impression that it's easier to do the reporting
> directly from the NetBSD machine in question...
I already reported the bug upstream: "Thank you for your problem report. You 
should receive confirmation of your problem report and an internal 
identification number by electronic mail in less than 24 hours."

NetBSD bug report: http://gnats.netbsd.org/cgi-bin/query-pr-single.pl?number=44057