The test suite of Test2::Suite seems to expect larger rounding errors than there are if perl is configured with long double support. This makes tests fail on platforms with nvsize > 8. Tests for floating point comparisons should be more platform-independent.
$ perl -V:nvsize
nvsize='16';
$ perl -MTest2::Suite -le 'print Test2::Suite->VERSION'
0.000082
$ perl t/modules/Compare/Float.t
...
not ok 8 - rounding {
ok 1 - 60.48 == 60.48 +/- 1e-08 - inside tolerance
ok 2 - 60.48 == 60.48 +/- 1e-13 - inside tolerance
ok 3 - 60.48 == 60.48 +/- 1e-14 - inside tolerance
ok 4 - 125 - 64.52 == 60.48 +/- 1e-08 - inside tolerance
ok 5 - 125 - 64.52 == 60.48 +/- 1e-13 - inside tolerance
not ok 6 - 125 - 64.52 != 60.48 +/- 1e-14 - outside tolerance
# Failed test '125 - 64.52 != 60.48 +/- 1e-14 - outside tolerance'
# at t/modules/Compare/Float.t line 87.
1..6
}
# Failed test 'rounding'
# at t/modules/Compare/Float.t line 88.
The test suite of Test2::Suite seems to expect larger rounding errors than there are if perl is configured with long double support. This makes tests fail on platforms with nvsize > 8. Tests for floating point comparisons should be more platform-independent.