Previously, the Rez file would have entries like this:
read 'cmap' (128, "Roman") "roman";
That has been fixed to show this:
read 'cmap' (128, "Roman") "charmap_roman.dat";
This simplifies the conversion test, since we don't need to be careful
about which data we run the conversion test in. It will also simplify
the command-line conversion tool and its distribution. The classic Mac
OS version of this program will continue to embed conversion tables in
the resource fork.
Extract table generation to its own file, table.go, and refactor the
interface. This exposed an inconsistency in the way that line breaks
were handled: both CR and LF on the Mac side were mapped to LF on the
UTF-8 side, but when the conversion table was inverted, the reverse
mappings would conflict. Previously, there was no explicit handling for
it, and whichever Mac charecter had a higher byte value would take
precedence. Conflicts are now detected and return an error, so line
breaks must be mapped explicitly. The new code maps CR, LF, and CRLF to
CR when converting UTF-8 to Mac.
Dietrich Epp