[Bioc-devel] Re-encoding failures on OS X (that always puzzled me)

[Henrik Bengtsson]

Hi,

not sure if this is a Bioconductor specific problem or not (because I
think I've seen it on CRAN as well), but at least it appears on the
BioC farm:

For aroma.light 1.27.0 (devel) I get the following R CMD check
warnings on OSX (only):

* checking whether package 'aroma.light' can be installed ... WARNING
Found the following significant warnings:
  Warning: unable to re-encode '901.CalibrationAndNormalization.R'
line(s) 28,67,162
  Warning: unable to re-encode 'normalizeCurveFit.matrix.R' line(s) 112
See '/Users/biocbuild/bbs-2.11-bioc/meat/aroma.light.Rcheck/00install.out'
for details.

See http://bioconductor.org/checkResults/devel/bioc-LATEST/aroma.light/perceval-checksrc.html

The above two files contains non-ASCII characters, e.g. ö (o with umlaut) as in:

#   \enc{Hössjer}{Hossjer} (2006).

(If anyone wonders, these non-ASCII characters are not part of actual
code, but part of in-source-code Rd documentation hidden within R
comments.)

To tell R that these special characters exists, I've added the
following to the aroma.light DESCRIPTION file:

Encoding: latin1

This seems to work on both Linux and Windows, but not on OSX.

Before I get into trial and error on this, does anyone why/a solution
to this?  Am I doing something wrong, or is it just that "latin1" is
not supported on (this particular?) OSX machine?  The 'Writing R
Extensions' docs says that "Only encoding names latin1, latin2 and
UTF-8 are known to be portable."  Should I use "UTF-8" instead?

Thanks,

Henrik