bcgov/bcmaps

Checking package with R_FUTURE_PLAN=multisession reveals problems

Closed this issue · 4 comments

HenrikBengtsson commented

UPDATE 2020-09-21: I've posted R CMD check output for another package. Please discard this issue.

Hi, I run revdep checks where I force packages to use 'multisession' futures instead of 'sequential' by default. This will help assert that the package work in parallel processing. When I did this for your package, the following error popped up:

$ export R_FUTURE_PLAN=multisession
$ R CMD check https://cran.r-project.org/src/contrib/bcmaps_0.18.1.tar.gz
...
ateucher commented

Thanks @HenrikBengtsson,

Somehow, this check output isn't from bcmaps - it looks like it's from the steps package (here is the kangaroo_model_example.Rmd vignette in that package: https://github.com/steps-dev/steps/blob/master/vignettes/kangaroo_model_example.Rmd).

However I am re-checking the package with R_FUTURE_PLAN=multisession and am going to rework the internals to follow your advice and rely on the user to set their future settings rather than the package doing it for them.

HenrikBengtsson commented

Oh... sorry about that. My bad. I must have cut'n'pasted from the wrong terminal window or log file (I had lots of the open at the same time). I went back and looked at the R_FUTURE_PLAN=multisession checks on bcmaps and it turns out it timed out after running for 60 minutes. So, that could reveal another type of bug. I leave it open but please feel free to close.

  • 👍1
ateucher commented

Yup, mine timed out too, thanks for confirming! I'll close this but I am working on it and we'll get a new version up soon. Thanks for the heads up.

HenrikBengtsson commented

Thanks. It's on my to-do list to write a vignette on 'Best Practices For Package Developers using Futures' (futureverse/future#375) where this check will be included. I just never get around to write that doc - oh well.

  • 👍1