These are not portable: the pdf() and postscript() devices only support
characters in an 8-bit locale, usually Latin-1, and R CMD check uses
pdf(), on CRAN in a Western European UTF-8 locale where pdf() uses
Latin-1. In most cases you should be using plotmath ....
Checking with environment varianble _R_CHECK_MBCS_CONVERSIONFAILURE
set to a non-empty value will throw an error in current R-devel,
although you can also grep the .Rcheck directory for 'conversion
failure.*mbcsToSbcs'
Attempts to plot Greek letters on a pdf() device are seen in the checks
of packages
Plotmath can also be used for ≤ (visualize) and ≥ (visualize).
Some versions of R will transliterate to e.g. <= or >= rather than substitute '.'
but the use of plotmath will look better.
NB: the issues will not be seen on a platform whose libiconv does
transliteration -- known are macOS 14 and Alpine Linux. But for example
on macOS 14, the missing Greek symbols are silently substituted by ?,
which is not useful.
[The cairo_pdf() and cairo_ps() devices handle fonts differently and can
plot a wider (but platform-dependent) range of Unicode chars. It looks
like using plotmath will suffice, but if not you could arrange for your
examples/tests to use cairo_pdf() if dev.cur() shows the current device
is pdf().]
Please correct before 2023-11-14 to safely retain your package on CRAN.
From Ripley with other packages censored.