utf8-check.c

   1 /*
   2  *      Check that the input is a proper UTF-8 file
   3  *
   4  *      Written in 2017 by Martin Mares <mj@ucw.cz>
   5  *      and placed into public domain.
   6  */
   7
   8 #include <stdio.h>
   9 #include <stdlib.h>
  10
  11 typedef unsigned int uint;
  12
  13 static void reject(const char *msg, uint arg)
  14 {
  15         printf("Error: ");
  16         printf(msg, arg);
  17         putchar('\n');
  18         exit(1);
  19 }
  20
  21 int main(void)
  22 {
  23         int c;
  24         while ((c = getchar()) >= 0) {
  25                 if (c < 0x20) {
  26                         if (c != '\r' && c != '\n' && c != '\t' && c != 0x0c)
  27                                 reject("ASCII control character %02x (only CR, LF, HT, FF allowed)", c);
  28                 } else if (c < 0x80) {
  29                         if (c == 0x7f)
  30                                 reject("ASCII DEL not allowed", c);
  31                 } else if (c < 0xc0) {
  32                         reject("Unexpected continuation byte %02x", c);
  33                 } else if (c < 0xf8) {
  34                         uint bytes = 1 + (c >= 0xe0) + (c >= 0xf0);
  35                         uint x = c & (0x3f >> bytes);
  36                         for (uint i=0; i<bytes; i++) {
  37                                 c = getchar();
  38                                 if (c < 0x80 || c >= 0xc0)
  39                                         reject("Incomplete multi-byte sequence at byte %02x", c);
  40                                 x = (x << 6) | (c & 0x3f);
  41                         }
  42
  43                         static const uint min_code[] = { 0, 0x80, 0x800, 0x10000 };
  44                         if (x < min_code[bytes])
  45                                 reject("Non-minimalistic encoding of %06x", x);
  46                         if (x > 0x10ffff)
  47                                 reject("Codepoint too high: %06x", x);
  48
  49                         if (x >= 0x0080 && x <= 0x009f)
  50                                 reject("C1 control character %04x", x);
  51                         if (x >= 0xd800 && x <= 0xdfff)
  52                                 reject("Surrogate code-point %04x", x);
  53                         if ((x & 0xffff) >= 0xfffe)
  54                                 reject("Non-character %06x", x);
  55                         if (x >= 0xe000 && x <= 0xf8ff ||
  56                             x >= 0xf0000 && x <= 0xffffd ||
  57                             x >= 0x100000 && x <= 0x10fffd)
  58                                 reject("Private-use character %06x", x);
  59                 } else {
  60                         reject("Invalid byte %02x", c);
  61                 }
  62         }
  63
  64         return 0;
  65 }