I have a package on my universe that passes checks with the local CI using R for windows-latest: Merge pull request #124 from carpentries/update-syllabus-rendering · carpentries/sandpaper@17f02d5 · GitHub, but it fails on the r-universe build due to a failed test: sandpaper 0.0.0.9027 · r-universe/carpentries@962128f · GitHub
It’s a curious problem because I have not seen this particular failure anywhere else. The output from the failed test could only occur if a file located in a temporary directory was not updated.
Is there any way to debug this?
@jeroenooms any thoughts on this?
The main difference is the version of the dependencies. Your own CI gets checked against CRAN versions of all dependencies, and in R-universe, we take the packages from your universe if they exist.
So you could try to install the versions of the dependencies from R-universe and test it that way.
By the way, if you get a test error on R-universe, the package will still be deployed to your universe. And you won’t get angry emails with deadlines either 
1 Like
The thing is, we already do this by using the universe as our Alternate_repository field. You can see this happen in the logs of what is downloaded and the session info from the runs.
Is it possible to upload the test artifacts after a run fails? At the moment, the run shows the last few lines of test output, but I know that something had to fail above those last few lines to give me the behavior its showing.
There are no words to describe how much I appreciate this.
1 Like