Codecov Report

Merging #5062 into master will increase coverage by 1.6%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #5062      +/-   ##
==========================================
+ Coverage   87.61%   89.21%    +1.6%     
==========================================
  Files        1009      727     -282     
  Lines      144082   103189   -40893     
  Branches     1418        0    -1418     
==========================================
- Hits       126232    92057   -34175     
+ Misses      17488    11132    -6356     
+ Partials      362        0     -362

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1a3eab5...f4d89b3. Read the comment docs.

@wesm did you intend to close this PR rather than merge it? it appears you agreed with @kou that it should be merged once CI passes.

@kou I can confirm that this change eliminates the warnings when building RAPIDS libcudf.

@harrism It was merged actually, we're just using a dedicated merge script rather than Github's PR merge button. So false alert there :-)

@harrism right, we don't use the GitHub UI to merge PRs, our merge tool creates atomic patches from each pull request and commits those to master, see

e41ad0d

I'm actually a bit surprised a project with many developers like RAPIDS is able to tolerate a "dirty" commit history (https://github.com/rapidsai/cudf/commits/branch-0.9). How do you expect to be able to git-bisect?

I've git bisected myself and so have others without problems. log(N) is a very powerful tool.

Thanks for accepting my patch.

The number of commits is not an issue -- it's a question of having a mix of "good" (green CI build) and "bad" (failing CI build) in the commit history. In any case, that's a discussion for another time and place :)