daattali
commented
4 years ago -
rocker fix: great, thanks for the explanation
-
pinning: I fully support pinning to exact version numbers in such cases. In fact, I've recently had a lot of convesations regading whether DESCRIPTION files should sometimes specify exact versions of dependencies and I think it's unfortunate that in the R world that can't really happen -- but that's entirely beside the point of what you were doing :)
-
do you know if there's a way to add a collaborator to travis? I couldn't find such a setting
crew102
This PR fixes the build on Travis. A few things:
b29778c is the commit that basically fixes things. In short, recent versions of ggplot2 started depending on the isoband package, which wasn't on CRAN until recently. rocker images (e.g., rocker/tidyverse) set the default repository to be a snapshop of CRAN (provided by MRAN) that is dated as of when the image was built. This meant that, when trying to build the dev version of ggplot2 during the tests, we were using a CRAN repo that didn't have all the deps needed to install the latest version of ggplot2.
Regarding 5ae1d07, I decided to pin our builds to a specific version of rocker/tidyverse b/c using a later version would result in slightly small changes to the expected figs (I believe due to a new version of Debian being used in the more up-to-date image), and I don't want to keep having to worry about updating/reviewing expected figs based on these types of changes in the future. In general, I think it makes sense to keep the environment we test in as stable as possible (i.e., pin to a specific version of rocker/tidyverse), as we're most interested in testing how changes in ggExtra code impacts expected figs, not how changes in code outside of ggExtra impacts figs.
ggplot2 versions 3.2.1 and 3.3.0 aren't under test yet (we go directly from 3.2.0 to dev version of ggplot2 in this PR), but I'll add those two versions sometime in the near future.
This build is going to fail on your Travis account, due to the issue mentioned here: https://github.com/daattali/ggExtra/pull/146#issuecomment-510713297. I created a fresh GITHUB_PAT and my build went from failing to passing, and I would expect yours to do the same...So maybe just create a fresh version of GITHUB_PAT and manually trigger the build again?