`fmtscan` cannot pass the static analysis

[lumynou5]

fmtscan introduced in #254 cannot pass the static analysis.

tools/fmtscan.c:659:29: style: The scope of the variable 'nextch2' can be reduced. [variableScope]
        get_char_t nextch1, nextch2;
                            ^
tools/fmtscan.c:362:11: style: Variable 'ptr' can be declared as pointer to const [constVariablePointer]
    char *ptr, *dict;
          ^
tools/fmtscan.c:1071:55: style: Parameter 'p' can be declared as pointer to const. However it seems that 'parse_eof' is a callback function, if 'p' is declared with const you might also need to cast function pointer(s). [constParameterCallback]
static inline get_char_t parse_eof(parser_t *restrict p UNUSED,
                                                      ^
tools/fmtscan.c:1190:20: note: You might need to cast the function pointer here
    [PARSER_EOF] = parse_eof,
                   ^
tools/fmtscan.c:1071:55: note: Parameter 'p' can be declared as pointer to const
static inline get_char_t parse_eof(parser_t *restrict p UNUSED,
                                                      ^
tools/fmtscan.c:1072:54: style: Parameter 't' can be declared as pointer to const. However it seems that 'parse_eof' is a callback function, if 't' is declared with const you might also need to cast function pointer(s). [constParameterCallback]
                                   token_t *restrict t UNUSED,
                                                     ^
tools/fmtscan.c:1190:20: note: You might need to cast the function pointer here
    [PARSER_EOF] = parse_eof,
                   ^
tools/fmtscan.c:1072:54: note: Parameter 't' can be declared as pointer to const
                                   token_t *restrict t UNUSED,
                                                     ^

[lumynou5]

By the way, I suppose -Wformat-nonliteral -Werror=format flag for GCC would be enough without need to use another program for checking.

[jserv]

By the way, I suppose -Wformat-nonliteral -Werror=format flag for GCC would be enough without need to use another program for checking.

Not exactly. Since the Linux kernel does not support floats or doubles, only decimal, octal, and hexadecimal formats are handled. However, there is no way to restrict such requirements to GCC or Clang.