Encoding error

[riastradh]

The attached file in.txt does not round-trip under ledger print, which yields a file out.txt with invalid UTF-8 encoding.

Somehow, the input octet sequence e7 94 b0 (i.e., Unicode code point U+7530, CJK UNIFIED IDEOGRAPH-7530) is mapped to the output octet sequence e7 94, which is invalid UTF-8 because it is missing the b0 octet.

Reproduced in ledger 3.2.1-20200518 (which I'm still actively using in production because of #2318) and in ledger 3.3.2-20230330, under NetBSD 9.3.

I have also seen a variant of this where the output is sometimes encoded correctly and sometimes encoded incorrectly, varying randomly from invocation to invocation of the same ledger executable on the same input file.