expect_comparison_linter() gets bad metadata for faulty expect_true() usage

[MichaelChirico]

Came across this:

expect_true(is.count(n_draws), n_draws > 1)

This throws expect_comparison_linter(), but gives a bad message:

lint(
  text = "expect_true(is.count(n_draws), n_draws > 1)",
  linters = expect_comparison_linter()
)
<text>:1:1: warning: [expect_comparison_linter] NA(x, y) is better than expect_true(x ( y).
expect_true(is.count(n_draws), n_draws > 1)
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Maybe we shouldn't invest in doing anything here since expect_true()'s been used incorrectly, but certainly it's not a helpful lint message.

[AshesITR]

Can the bug be triggered by a valid usage like

expect_true(info = "my info first", a > b)

?

If so, I think putting a more robust XPath in place to find the comparator is a good idea.

[MichaelChirico]

expect_true(info = "my info first", a > b)

this doesn't lint, it's unusual enough that I would wait for a request first before supporting that.