- v0.013_008: Devel::Cover showed SKIP was too aggressive in t\09 and t\10, so never tested
negate in v0.013_007.
+ FIXED BUG: t\09-signbit: while verifying test coverage in v0.013_008, discovered
that signbit test was never testing negate, because the SKIP was too aggressive;
once I de-aggressified the SKIP, finding more failures.
+ it looks like NEG_IND is not properly sign-changed to POS_IND, either... need to
investigate further on multiple systems
+ Determined that it's pointless to keep trying to make CORE::abs() pass the tests.
Rename Data::IEEE754::Tools::absolute() to ...::abs(), and stop testing the
CORE::abs(). Just make a note that exporting abs or :signbit will override the
builtin behavior, and explain how to still access CORE::abs().
+ remove isCoreAbsWrongForNegNaN(), since apparently Perl itself is the culprit, not
certain implementations. (The auto-stringification of NaN was sometimes not
distinguishing the sign bits.)
+ t\10-copysign: switch to the to_hex_floatingpoint() for comparing abs($x) to abs($z),
to avoid the same display bug that masked the CORE::abs() bug for 5.8.5 and 5.24.0.
- v0.013_007: CPAN Testers for a slew of errors for v0.013_006, so fix those
FIXED = <
https://rt.cpan.org/Ticket/Display.html?id=117163>
+ BUG DESCRIPTION: in some configs, the 09-signbit and 10-copysign values have wrong sign
+ added isCoreAbsWrongForNegNaN() to indicate systems for which CORE::abs() has the wrong
sign.
+ t\09-signbit: Use that to skip the NaN tests relying on CORE::abs().
+ t\10-copysign: Use absolute() instead of abs() for verifying the magnitude copied correctly
from argument x.
