Skip byte counting at EOF in ExtendedBufferedReader.read

[rootvector2]

ExtendedBufferedReader.read() feeds the EOF sentinel (-1) to getEncodedCharLength when trackBytes is on, narrowing it to (char) -1 (U+FFFF) and pushing it through the charset encoder. A single-byte charset like ISO-8859-1 or US-ASCII cannot encode U+FFFF, so byte tracking throws UnmappableCharacterException at end of input on an otherwise valid file; with UTF-8/UTF-16 it silently adds a few phantom bytes to the count. Found while running getBytePosition with a non-UTF charset; the read(char[], off, len) overload already guards this with len > 0, so this mirrors that guard for the single-char path.

[garydgregory]

Good catch

[garydgregory]

Good catch @rootvector2 , merged 🚀