diff options
| author | Matthew Brett <matthew.brett@gmail.com> | 2017-02-21 14:40:56 -0800 |
|---|---|---|
| committer | Matthew Brett <matthew.brett@gmail.com> | 2017-02-21 14:40:56 -0800 |
| commit | 9df4a528b57d0d95d84132394178e65ba2683159 (patch) | |
| tree | 3b1d7726e4483bc6086074ade5b4d92e204dfb2a /numpy/doc/subclassing.py | |
| parent | 2aabeafb97bea4e1bfa29d946fbf31e1104e7ae0 (diff) | |
| download | numpy-9df4a528b57d0d95d84132394178e65ba2683159.tar.gz | |
BUG: fix denormal linspace test for longdouble
Linspace test was testing for case where simply dividing the range by
the number of samples generates 0, as the case for very small denormals.
The implementation generating the denormal only worked to exercise that
code for float64. Make a small denormal that does exercise that case,
as least for longdouble, as well as float64.
Closes gh-8565.
Diffstat (limited to 'numpy/doc/subclassing.py')
0 files changed, 0 insertions, 0 deletions