search

pharmaR / riskassessment

Collaborative Deployment: https://app.pharmar.org/riskassessment/ Risk Assessment Demo App: https://rinpharma.shinyapps.io/riskassessment
https://pharmar.github.io/riskassessment/
Other
101 stars 28 forks source link

Checks failing on macOS #549

Closed Jeff-Thompson12 closed 2 months ago

Jeff-Thompson12 commented 1 year ago

Has anyone else noticed the errors for macOS lately?

Snag_af88293
imran3004 commented 1 year ago

@Jeff-Thompson12 , what are the steps to reproduce this please?

Jeff-Thompson12 commented 1 year ago

@imran3004 It is on the Github Actions for the Next Release PR.

Jeff-Thompson12 commented 1 year ago

I think I am slowing figuring out what is occurring here. It appears that providing the repository in renv.lock does not dictate how the packages are being installed in renv::restore(). When doing the restore, {renv} utilizes the HASH.

You can further notice that we are using versions of packages that are older than the RSPM snapshot date. It probably makes sense to reset the project based on that date.

Jeff-Thompson12 commented 1 year ago

Some of the HASHs in the renv.lock file are pointing to the microsoft.mran portal which is scheduled to close soon. I believe this to mean that users will not be able to initialize their project with our lock file unless we update the packages to point to the RSPM url.

Jeff-Thompson12 commented 1 year ago

Note the retirement of MRAN here

Jeff-Thompson12 commented 1 year ago

The issue boils down to this synopsis:

The Microsoft R Application Network (MRAN) is being retired on July 1, 2023. When running renv::restore(), the package and version from renv.lock is searched for on MRAN. This is done because MRAN was storing the binaries for different operating systems. By disabling MRAN, renv::restore() will use other means (like the repo specified in the lock file), but will try to install these packages from source. This is causing the GHAs to fail during the setup process. The only way to fix this is to figure out how to configure the virtual machine so that the packages can be installed from source. This only gives us a benefit if we anticipate both that these operating systems will be used for deployment of the application and that there is some functional difference between OS that require us to account for that different behavior. The second is unlikely because we are not compiling any code for use in the application.

It is probably beneficial for us to disable using MRAN, since an older version of {renv} is being utilized in the project. But we should probably consider removing the R-CMD checks for Windows and Mac.

I should also note that MRAN is only used for Windows and Mac. That is why we have not seen issues with Ubuntu.

jthompson-arcus commented 2 months ago

Stale