Identify (and potentially address) Windows-only issues in CRAN packages

[hturner]

This task has been proposed by the CRAN Team, cc @kurthornik. This work could be distributed among a large group of contributors.

Task

1. Identify CRAN packages that fail on Windows (but not other OS)
2. Find out what the issues are
3. Generate a list of the issues and prepare bug reports if feasible

Hints

• tools::CRAN_check_results() can be used to find the status (OK < NOTE < WARNING < ERROR < FAILURE) of checks on different platforms
• tools::CRAN_check_details() can be used to get the output of non-OK checks
• Submit PRs for CRAN packages with "Lost braces" NOTE on "Rd files" check #110 gives some code to find the URL for the source code repository, if provided in the package DESCRIPTION, where an issue or PR could be opened.

[JayaniLakshika]

R Dev Day Australia 2025

• accSDA-v1.1.3: PR opened: Fix the issues gumeo/accSDA#7
• actilifecounts-v1.1.1: PR opened: Fix description bug cran/actilifecounts#1
• AnchorRegression-v0.1.3: PR opened: Fixed the note simzim96/AnchorRegression#5
• airship-v1.4.3: PR opened: Fix Depends R (>= 4.1.0) el-meyer/airship#163
• adobeanaliticsr-v0.5.0: PR opened: modify R version dependency benrwoodard/adobeanalyticsr#190

We are using this code to find packages:

# Finding R packages with problems 
# tools::CRAN_check_results() can be used to find the status (OK < NOTE < WARNING < ERROR < FAILURE) of checks on different platforms
# tools::CRAN_check_details() can be used to get the output of non-OK checks

library(tools)
library(tidyverse)

pkg <- CRAN_package_db()

# package with github in description 
pkg.git <- pkg |> 
  dplyr::filter(grepl("github", Description, ignore.case = TRUE)) %>%
  dplyr::select(Package, Version, Description)


# packages with status not OK
pkg.err <- CRAN_check_results() |>  filter(Status != 'OK')
pkg.reduced <- inner_join(pkg.err, pkg.git, by = c("Package", "Version")) 

# pick one package to work in
pkg.reduced |> select(Package, Version, Status) |> head()

# get details of problems for that package
problems <- CRAN_check_details() 
problems |> filter( Package == 'activAnalyzer' )|> View()

# what about packages with no public github repo
pkg.nogit <- pkg |> 
  dplyr::filter(!grepl("github", Description, ignore.case = TRUE)) |> 
  dplyr::select(Package, Version, Description) |> 
  inner_join(pkg.err, by = c("Package", "Version"))

[JayaniLakshika]

If the R package doesn't have a GitHub repo.

Patch: #131

[JayaniLakshika]

Fix bugs with the CRAN mirror:

• ABACUS: updated Authors@R and removed LazyData field. Send the patch to the maintainer (dr.m.nath@gmail.com).

[hturner]

Update for R Dev Day @ New Zealand:

Identify CRAN packages that fail on Windows (but not other OS)

library(tools)
check_results <- CRAN_check_results()

## packages that fail on Windows
check_windows <- subset(check_results, 
                        grepl("windows", Flavor) & Status != "OK")

## packages that fail on non-Windows   
check_other <- subset(check_results, 
                      !grepl("windows", Flavor) & Status != "OK")

## packages that fail on Windows but not other OS
fail <- setdiff(check_windows$Package, check_other$Package)
length(fail)
#> [1] 190

Find out why they fail

check_details <- CRAN_check_details()

fail_details <- subset(check_details,
                       grepl("windows", Flavor) & Status != "OK" &
                           Package %in% fail)

unique(fail_details$Check)
#>  [1] "whether package can be installed"                            
#>  [2] "C++ specification"                                           
#>  [3] "R code for possible problems"                                
#>  [4] "tests"                                                       
#>  [5] "compiled code"                                               
#>  [6] "re-building of vignette outputs"                             
#>  [7] "dependencies in R code"                                      
#>  [8] "package subdirectories"                                      
#>  [9] "package dependencies"                                        
#> [10] "examples"                                                    
#> [11] "data for non-ASCII characters"                               
#> [12] "installed package size"                                      
#> [13] "DESCRIPTION meta-information"                                
#> [14] "whether the namespace can be loaded with stated dependencies"
#> [15] "LazyData"

Focus on packages with GitHub repo that fail a specific check

Contributors could focus on a specific category depending on their knowledge and expertise. E. g. issues with "package dependencies" may be simpler to solve than issues with "compiled code".

## merge package info with check results
pdb <- CRAN_package_db()
fail_details <- merge(fail_details, pdb)

## subset to specific category of failures that have a repo on GitHu
focus <- subset(fail_details, 
                Check == "package dependencies" &
                    grepl("^http?s://github.com.*issues", BugReports))
dim(focus)
#>   [1] 26 74
focus[1, c("Package", "BugReports", "Check", "Status", "Output")]
#>   Package                            BugReports                Check Status
#> 5     apa https://github.com/dgromer/apa/issues package dependencies  ERROR
#>                                                                                                                     Output
#> 5 Package required but not available: 'MBESS'\n\nSee section 'The DESCRIPTION file' in the 'Writing R Extensions'\nmanual.