Not all usage issues reported back to user

[AshesITR]

tst <- function() { a <- ...; a; b }
codetools::checkUsage(tst)
#> <anonymous>: ... may be used in an incorrect context (<text>:1)
#> <anonymous>: no visible binding for global variable 'b' (<text>:1)

^{Created on 2022-10-16 by the reprex package (v2.0.1)}

Lintr only reports one of the problems:

lintr::lint(text = "tst <- function() { a <- ...; a; b }", linters = lintr::object_usage_linter())
#> <text>:1:34: warning: [object_usage_linter] no visible binding for global variable 'b'
#> tst <- function() { a <- ...; a; b }
#>                                  ^

^{Created on 2022-10-15 by the reprex package (v2.0.1)}

---

Checking the code of R/codetools.R, all issues are collected using the format signalUsageIssue(message). The regex - if I read it correctly - implicitly assumes there always is an sQuote(...) in the message.

This is not the case for the following messages, scouring the source code:

w$signal("... may be used in an incorrect context", w) # l. 416
w$signal(paste(a, "may be used in an incorrect context:", pasteExpr(e)), w) # l. 436
w$signal(paste("wrong number of arguments to '.Internal':", pasteExpr(e)), w) # l. 676
w$signal(paste("bad argument to '.Internal':", pasteExpr(e[[2]])), w) # l. 682

Originally posted by @AshesITR in #1661 (review)

[AshesITR]

Here's a quick and dirty fix that will also make the if (any(missing)) conditional obsolete:

  res <- re_matches(
    vals,
    rex(
      function_name,
      capture(
        name = "message",
        zero_or_more(any, type = "lazy"),
        maybe(
          one_of(quote, "\u2018"),
          capture(name = "name", anything),
          one_of(quote, "\u2019"),
          anything
        )
      ),
      line_info
    )
  )