deploy failure dev shiny

[TNRiley]

Inline comment causing deployment failure

[TNRiley]

fixed the inline comment, but have further issues regarding CRAN mirrors. Haven't been able to solve.

Deploy error

Error in contrib.url(repos, type) :
trying to use CRAN without setting a mirror
Calls: install.packages -> startsWith -> contrib.url
Execution halted
Error: Process completed with exit code 1.

[LukasWallrich]

Sounds like we have "install.packages" somewhere in the code - that should not anywhere

[LukasWallrich]

Ignore that comment - clearly the issue is in the action, not our code

[LukasWallrich]

Currently, we are setting up R in two ways -

    uses: r-lib/actions/setup-r@v2

AND

    uses: eddelbuettel/github-actions/r-ci-setup@master

I believe we still use the second version even after your change (which allows for much faster package install) - but should not set up both. I will try to track down how we can change the version in the r-ci version, and see if that fixes the CRAN issue (that version uses an inbuilt CRAN, so we should not change it)

[LukasWallrich]

I asked the r2u author whether we can specify the R version: eddelbuettel/r2u#74 - that approach is faster and more robust, so I would not want to rebuild the action for now.

R 4.2.2 was released 8 days ago, and shinyapps.io usually wants to support new versions within a couple of days - so the problem will go away temporarily very soon anyway, but we should obviously know how to fix this when it recurrs.

[LukasWallrich]

It cannot be done with r2u, and I would not want to fix the R version in our automated deploy script as that would mean updating that regularly. Instead, I added a manual deploy script that we can use when shinyapps.io and the latest R diverge. To run that, go to Actions and select document-and-deploy-manual, click on Run workflow, select whether you want to run it for main or dev and specify the R version (the current default works this time round). It takes twice as long as the usual deploy, but gets the job done.

[LukasWallrich]

@TNRiley hope that makes sense - let me know if you run into further issues / have better ideas.

[LukasWallrich]

I realized that the performance catches up when the action is run more than once (due to caching) and that it is better to have only one version - so I switched to this, and we now always deploy to shinyapps.io with the previous R version. That should be robust and easier to maintain than two actions. @TNRiley so no manual deploy is needed.

[TNRiley]

I realized that the performance catches up when the action is run more than once (due to caching) and that it is better to have only one version - so I switched to this, and we now always deploy to shinyapps.io with the previous R version. That should be robust and easier to maintain than two actions. @TNRiley so no manual deploy is needed.

sounds good. much of this deployment stuff is over my head as I've not yet played around with it. Appreciate you working on this and resolving it. I see the latest now has the fixes I had made.