Don't assume from="UTF-8" in encoding test

[MichaelChirico]

For #7217 (possibly closing it?). As described by Ivan below.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.01%. Comparing base (dcaa87c) to head (8fe3782).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #7852   +/-   ##
=======================================
  Coverage   99.01%   99.01%           
=======================================
  Files          88       88           
  Lines       17286    17286           
=======================================
  Hits        17116    17116           
  Misses        170      170           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[aitap]

Here, the test is being overly pessimistic. If the native encoding is capable of representing all the needed characters, it should be possible for the test to pass. y just needs the correctly converted names, not from UTF-8, but from the native encoding: 52271bb

This looks extremely counter-intuitive. setnames(y , iconv(c("\u00f6", "\u00fc", "\u00e4"), from = "UTF-8", to = "latin1")) on its own would be right (string literals with Unicode escapes do get marked as UTF-8 as intended), but here we're doing eval(parse(text = 'iconv(...)')). In order to parse the source code, R first converts the Unicode source code into the native encoding. After parse(), we're evaluating iconv(c("ö", "ü", "ä"), to = 'latin1'), which must use from = '' (default), not from = "UTF-8".

[MichaelChirico]

makes sense... maybe utf8_check() isn't the perfect name for the helper? I'm still wondering why this is only an issue for ancient windows

[aitap]

True, the check is more like representable_in_locale_encoding().

Almost everywhere else except on old R or old Windows the locale encoding is UTF-8, so iconv(from = "UTF-8") instead of from = "" does not break anything. You can also break this test on Linux using a non-Unicode locale that represents the necessary umlauts (say, de_DE.ISO-8859-1).