Changes between Initial Version and Version 17 of Ticket #26360
- Timestamp:
- 10/27/18 07:44:19 (2 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
Ticket #26360
- Property Cc fbissey added
-
Property
Commit
changed from
00d3ccd2ee230d9252b9cdc3183adf9c7f9f45ad
to30cc778d46579bd0c7537ed33e8d7a4f40fd5c31
-
Property
Status
changed from
new
toneeds_review
-
Property
Summary
changed from
Upgrade arb to 2.15.0
toUpgrade arb to 2.15.1
-
Ticket #26360 – Description
initial v17 1 Arb 2.15. 0is out and the upgrade breaks a lot (~200) doctests with changes that are not actually failures. Nearly all the changes are minor differences in the accuracy of the radius. As previously discussed in #25966, I think we should fix this once and for all. The tests are brittle and break with more or less every arb upgrade. All those "failures" distract from the actual failures. It makes arb upgrades painful and is a burden on distributions.1 Arb 2.15.1 is out and the upgrade breaks a lot (~200) doctests with changes that are not actually failures. Nearly all the changes are minor differences in the accuracy of the radius. As previously discussed in #25966, I think we should fix this once and for all. The tests are brittle and break with more or less every arb upgrade. All those "failures" distract from the actual failures. It makes arb upgrades painful and is a burden on distributions. 2 2 3 3 I suspect the solution here (lots of `...`) will be controversial. I don't like it very much myself. But I think it is better than the status quo. What we really want are different tolerances for the "mid" and the "radius". I don't know if that is possible in the doctesting framework without explicitly testing for `.mid()` and `.radius()` in different tests each time. We may even want to test the `.accuracy()` instead of the radius. 4 4 5 Upstream tarball: https://github.com/fredrik-johansson/arb/archive/2.15. 0.tar.gz5 Upstream tarball: https://github.com/fredrik-johansson/arb/archive/2.15.1.tar.gz