R-CMD-check not roxygenizing documentation?

[dcnorris]

Although local devtools::check() works fine, and the documentation builds fine locally too, I find that GitHub action R-CMD-check fails with documentation for all package exports failing to build:

── R CMD check results ──────────────────────────────── precautionary 0.2.4 ────
Error: Error: R CMD check found WARNINGs
Execution halted
Duration: 4m 33.1s

❯ checking for missing documentation entries ... WARNING
  Undocumented code objects:
    ‘Boin’ ‘Ccd’ ‘Cpe’ ‘Crm’ ‘cohorts_of_n’ ‘crmh’ ‘crmh_xo’ ‘crmht’
    ‘crmht2’ ‘exact’ ‘extend’ ‘hyper_mtdi_lognormal’ ‘mtdi_lognormal’
    ‘safety_kable’ ‘stop_for_excess_toxicity_empiric’ ‘test_draw_samples’
    ‘viola_dtp’
  Undocumented data sets:
    ‘viola_dtp’
  Undocumented S4 methods:
    generic 'plot' and siglist 'mtdi_distribution,ANY'
    generic 'plot' and siglist 'mtdi_generator,ANY'
    generic 'simulate_trials' and siglist
      'exact,missing,mtdi_distribution,ANY'
    generic 'simulate_trials' and siglist
      'exact,numeric,hyper_mtdi_distribution,missing'
    generic 'simulate_trials' and siglist
      'exact,numeric,mtdi_distribution,missing'
    generic 'simulate_trials' and siglist
      'selector_factory,numeric,hyper_mtdi_distribution,ANY'
    generic 'simulate_trials' and siglist
      'selector_factory,numeric,mtdi_distribution,ANY'

Have I omitted some step, or does this represent an error in the R-CMD-check github action?

[dcnorris]

As someone who typically follows avoids keeping generated code files under version control, I had not been checking in my man/*Rd files. But I find that pushing these files immediately solves the 'problem'.

Now my question becomes, does the 'Happy Git' book's advice about generated analytical outputs apply now also to generated documentation for an R package on GitHub that takes advantage of actions and pkgdown?