Avoid R package warning in loadSupport() if autoload is disabled

[krlmlr]

in loadSupport(), if _disable_autoload.R exists.

[wch]

@krlmlr Can you provide instructions on how to reproduce the issue?

[krlmlr]

Create a Shiny app with a package structure (with DESCRIPTION and NAMESPACE files) that contains a file R/_disable_autoload.R and uses e.g. pkgload::load_all() to load the files in R/ . The warning still occurs.

Do we need a test here?

[CLAassistant]

All committers have signed the CLA.

[tanho63]

Minimal code for reproducing bogus warning here:

withr::with_dir(tempdir(check = TRUE),{
  dir.create("R", showWarnings = FALSE)
  # loadSupport documentation suggests adding a blank R/_disable_autoload.R file
  # will disable automatic loading of the R/ files
  file.create("R/_disable_autoload.R")
  # test that code was not loaded by adding a print statement
  writeLines(
    "print('R/ code was sourced')",
    "R/code.R"
  )
  # loadSupport checks for a NAMESPACE file and/or a DESCRIPTION file to determine
  # if the directory contains a package, add NAMESPACE to trigger the check
  file.create("NAMESPACE")
  # add minimal app.R code
  writeLines(
    text = "
    shiny::shinyApp(
      ui = shiny::fluidPage(), 
      server = function(input, output, session) {}
    )
    ", 
    con = "app.R"
  )
  shiny::runApp()
  # still triggers warning, despite not actually running the code in question
})

[tanho63]

@krlmlr maybe simpler to move this early exit on disabled to L356 instead of extending the if clause?

[gadenbuie]

I'm working on a change like that... will push back to this PR in a minute.

[gadenbuie]

In the end, I think the warning was happening too early. By moving it down we can instead warn just before evaluating the code in the R/ support directory.

[krlmlr]

Nice, thanks for picking this up!