Closed zkamvar closed 10 months ago
Currently testing in https://github.com/zkamvar/carpentries-actions-88
so far, I am unable to reproduce the behaviour. I used the following steps:
Unfortunately for my purposes, the workflow ran perfectly. I downgraded the cowsay package in the PR and am attempting it again now.
I have also forked R novice gapminder (https://github.com/zkamvar/r-novice-gapminder/) and am checking if it is a cache issue
Update: the second try with the MWE did not fail: https://github.com/zkamvar/carpentries-actions-88/actions/runs/6459830829/job/17536371072
But, the fork did fail: https://github.com/zkamvar/r-novice-gapminder/actions/runs/6459912227/job/17536613719
and this leaves me with 🤷🏼
The way to fix this was to use renv::snapshot(force = TRUE)
(in https://github.com/carpentries/vise/pull/7). I am still not sure why the preflight validation failed. It would say that packages that were available were not available and it was hella confusing.
In any case this should now work properly and if the pull request action fails, #89 will produce an artifact from this workflow that can replace the {renv} folder.
There is a situation where lesson containing packages that do not exist in the {sandpaper} suite fail to update due to a snapshot validation failure as shown in r-novice-gapminder, https://github.com/swcarpentry/r-novice-gapminder/actions/runs/6386889815/job/17334353257
This was not a problem for r-novice-inflammation, which does not use much more than knitr: https://github.com/swcarpentry/r-novice-inflammation/pull/595
I will test further and update here when I am able.