https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109098
--- Comment #3 from David Malcolm <dmalcolm at gcc dot gnu.org> --- (In reply to Andrew Pinski from comment #1) > I would have assumed you need -finput-charset= for the non-utf8 ones really > if your LANG/LANGUAGE is not set to C/UTF8 really. Yeah, but when complaining about encoding issues, the error message we emit should at least be properly encoded :/ It's a major pain for my integration testing where two(?) bad bytes in one source file lead to an unparseable .sarif file (out of thousands). When quoting source in the .sarif output, we should ensure that the final JSON output is all valid UTF-8, perhaps falling back to not quoting source for cases where e.g. - the source file isn't validly encoded, or - the -finput-charset= is wrong, or - the -finput-charset= is missing or - where the source file (erroneously) uses a mixture of different encodings in different parts of itself Probably should also check we do something sane for trojan source attacks
