A nightly coverage build would also have the benefit of actually being useful. Because we rarely run all of the coverage jobs on a single pull request, the report of the change in coverage from a diff is worthless.
Neal On Mon, Nov 4, 2019 at 9:16 AM Wes McKinney <[email protected]> wrote: > > I'm okay with disabling it. It will speed up many of our CI builds. > > If we care about code coverage, I would suggest running nightly > coverage builds rather than every commit > > On Mon, Nov 4, 2019 at 11:09 AM Antoine Pitrou <[email protected]> wrote: > > > > > > Hello, > > > > Code coverage computation on Travis-CI builds probably takes a > > non-trivial amount of time and it's not obvious anyone is using the > > results concretely. Should we disable it? > > > > (note we may also disable it selectively on C++, Python, or Rust, > > depending on the answers) > > > > Regards > > > > Antoine.
