Repository navigation
Add note about risk of round-off errors for Time. - #10135
Conversation
| .. |minimum_python_version| replace:: {0.__minimum_python_version__} | ||
| .. |minimum_numpy_version| replace:: {0.__minimum_numpy_version__} | ||
| .. |minimum_scipy_version| replace:: {0.__minimum_scipy_version__} | ||
| .. |minimum_scipy_version| replace:: nonsense |
There was a problem hiding this comment.
O, darn, that shouldn't have gone in - #10137 - I'll take it out! (and will use ci skip since this worked already)
b75304a to
5a5c293
Compare
|
Note to reviewers - test passed but had included something unrelated (#10135 (comment)), so re-ran skipping CI without that. |
|
@bsipocz - I see |
|
Yeah, that was just a home grown workaround for travis, but it doesn't work for circleCI. Use |
taldcroft
left a comment
There was a problem hiding this comment.
Good pending fix of the typo.
| errors can cause two times to be not strictly equal even if | ||
| mathematically they should be. For times in UTC in particular, this | ||
| can lead to surprising behaviour, because when one adds a | ||
| |TimeDelta|, which cannot be have a scale of UTC, the UTC time is |
5a5c293 to
8b997ec
Compare
|
OK, fixed. I cancelled most of the travis builds since this passed before; hoped to also cancel all but html circle-ci builds but seem unable to. |
Add note about risk of round-off errors for Time.
fixes #6970 - adding a note about round-off and the explicit example that adding 0 to times in UTC can be surprising.