Hans van der Meer wrote:
Therefore I suspect it has something to do with the output routine where \s is taken as the next macro. A simpe experiment (\def\s{something}) shows that indeed a macro \s is executed. Is someone here messing up the catcode's of the letters?
Yes, the \startJAVA command :-) The problem is that context is loading the pdfr-ec resource file (Glyph -> Unicode mapping) while the prettytype catcodes are in effect, and it does not properly switch back to "non-verbatim" catcodes. I am not quite sure what the best way is to fix this, but I've redefined that loading macro in my preamble, like so, and it seems to work as a temporary workaround: \def\dododoincludepdffontresource#1% encoding {\bgroup \def\currentencoding{#1}% \doifvaluesomething\pdffontfileresource {\uncatcodeallcharacters % new \catcode`\[=12 % new* \catcode`\]=12 % new* \setnaturalcatcodes % new \startreadingfile \readsysfile{pdfr-\getvalue\pdffontfileresource} \donothing\donothing \stopreadingfile \letgvalue\pdffontfileresource\empty}% \egroup}