empty string with non-empty trailing context consumes code units

[skvadrik]

Given the following code:

/*!re2c
    "" / "a" {}
*/

re2c generates:

{
        YYCTYPE yych;
        YYCTXMARKER = YYCURSOR + 1;
        if (YYLIMIT <= YYCURSOR) YYFILL(1);
        yych = *YYCURSOR;
        switch (yych) {
        case 'a':       goto yy3;
        default:        goto yy2;
        }
yy2:
yy3:
        ++YYCURSOR;
        YYCURSOR = YYCTXMARKER;
        {}
}

On strings starting with "a" this lexer will consume single code unit ('a') and advance YYCURSOR to code unit next to 'a'. This is clearly not what was intended.

Lookahead is a useful feature and this should be fixed.

[skvadrik]

Fixed in this commit.