I have no expectation of luck or desire to rely on it. At the point I suggested, to the OP, a way of knowing how much of a multi-statement string was accepted by sqlite3_prepare, the "strings" are simple char* but known to be referencing UTF-8 code sequences. Computing the accepted number of "characters" (or UTF-8 code points) has nothing to do with C# at that level, with this tiny proviso: Once control gets back to the C# domain, where we can presume the multi-statement string appears as a CLR String type, it is quite easy to lop off the accepted portion using the .Substring(int startIndex) method, where that index is *not* a count of UTF-16 words but *is* a zero-based "character position". That .Substring() method can be safely used without having to anticipate that the UTF-16 encoding used for CLR String objects will result in some other number of characters being lopped off or that a UTF-16 code point representation will be sliced into pieces. It is because zSql (and zTail) point to possibly multi-byte character representations that my suggested code did not do simple pointer arithmetic. Doing that might have worked, but only by luck. As we know, having something work by luck is often a form of bad luck.