So that I can torture-test some of my code I've tried to build perls
with all the various combinations of integer and float sizes. And I've
come across a problem with 32-bit integers and quadruple-precsion floats
via libquadmath.
On a 32-bit Linux, perl builds just fine with the following Configure
options and passes all its tests:
-de
-de -Duselongdouble
-de -Duse64bitint
-de -Duse64bitint -Duselongdouble
-de -Duse64bitint -Dusequadmath
It builds, but has test failures with:
-de -Dusequadmath
I've tried with 5.32.0 and 5.34.0 and get exactly the same test failures
on both:
Failed 10 tests out of 2449, 99.59% okay.
../cpan/Math-BigInt-FastCalc/t/bigintfc.t
../cpan/Math-BigInt/t/bare_mbf.t
../cpan/Math-BigInt/t/bare_mbi.t
../cpan/Math-BigInt/t/bigfltpm.t
../cpan/Math-BigInt/t/bigintc.t
../cpan/Math-BigInt/t/bigintpm.t
../cpan/Math-BigInt/t/sub_mbf.t
../cpan/Math-BigInt/t/sub_mbi.t
../cpan/Math-BigInt/t/upgrade.t
../cpan/Math-BigInt/t/with_sub.t
That inclines me to think that the bug is *probably* in Math::BigInt,
but I suppose there's a possibility that that just happens to be the
only place in the tests that tickles a bug in perl itself.
Before I go digging into this, has anyone seen this before? Is anyone
familiar enough with BigInt to know what the problem is?
--
David Cantrell
with all the various combinations of integer and float sizes. And I've
come across a problem with 32-bit integers and quadruple-precsion floats
via libquadmath.
On a 32-bit Linux, perl builds just fine with the following Configure
options and passes all its tests:
-de
-de -Duselongdouble
-de -Duse64bitint
-de -Duse64bitint -Duselongdouble
-de -Duse64bitint -Dusequadmath
It builds, but has test failures with:
-de -Dusequadmath
I've tried with 5.32.0 and 5.34.0 and get exactly the same test failures
on both:
Failed 10 tests out of 2449, 99.59% okay.
../cpan/Math-BigInt-FastCalc/t/bigintfc.t
../cpan/Math-BigInt/t/bare_mbf.t
../cpan/Math-BigInt/t/bare_mbi.t
../cpan/Math-BigInt/t/bigfltpm.t
../cpan/Math-BigInt/t/bigintc.t
../cpan/Math-BigInt/t/bigintpm.t
../cpan/Math-BigInt/t/sub_mbf.t
../cpan/Math-BigInt/t/sub_mbi.t
../cpan/Math-BigInt/t/upgrade.t
../cpan/Math-BigInt/t/with_sub.t
That inclines me to think that the bug is *probably* in Math::BigInt,
but I suppose there's a possibility that that just happens to be the
only place in the tests that tickles a bug in perl itself.
Before I go digging into this, has anyone seen this before? Is anyone
familiar enough with BigInt to know what the problem is?
--
David Cantrell