+ int state, be;
+ unsigned int accum;
+ size_t i;
+ unsigned char cu;
+ const char *cp;
+ const long one = 1L;
+
+ cp = b->buf + (int)b->offs;
+ state = 0;
+ accum = 0U;
+ be = 0;
+
+ /* Quick test for big-endian value. */
+
+ if ( ! (*((char *)(&one))))
+ be = 1;
+
+ for (i = b->offs; i < b->sz; i++) {
+ cu = (unsigned char)*cp++;
+ if (state) {
+ if ( ! (cu & 128) || (cu & 64)) {
+ /* Bad sequence header. */
+ return(0);
+ }
+
+ /* Accept only legitimate bit patterns. */
+
+ if (cu > 191 || cu < 128) {
+ /* Bad in-sequence bits. */
+ return(0);
+ }
+
+ accum |= (cu & 63) << --state * 6;
+
+ /*
+ * Accum is held in little-endian order as
+ * stipulated by the UTF-8 sequence coding. We
+ * need to convert to a native big-endian if our
+ * architecture requires it.
+ */
+
+ if (0 == state && be)
+ accum = (accum >> 24) |
+ ((accum << 8) & 0x00FF0000) |
+ ((accum >> 8) & 0x0000FF00) |
+ (accum << 24);
+
+ if (0 == state) {
+ accum < 128U ? putchar(accum) :
+ printf("\\[u%.4X]", accum);
+ accum = 0U;
+ }
+ } else if (cu & (1 << 7)) {
+ /*
+ * Entering a UTF-8 state: if we encounter a
+ * UTF-8 bitmask, calculate the expected UTF-8
+ * state from it.
+ */
+ for (state = 0; state < 7; state++)
+ if ( ! (cu & (1 << (7 - state))))
+ break;
+
+ /* Accept only legitimate bit patterns. */
+
+ switch (state) {
+ case (4):
+ if (cu <= 244 && cu >= 240) {
+ accum = (cu & 7) << 18;
+ break;
+ }
+ /* Bad 4-sequence start bits. */
+ return(0);
+ case (3):
+ if (cu <= 239 && cu >= 224) {
+ accum = (cu & 15) << 12;
+ break;
+ }
+ /* Bad 3-sequence start bits. */
+ return(0);
+ case (2):
+ if (cu <= 223 && cu >= 194) {
+ accum = (cu & 31) << 6;
+ break;
+ }
+ /* Bad 2-sequence start bits. */
+ return(0);
+ default:
+ /* Bad sequence bit mask. */
+ return(0);
+ }
+ state--;
+ } else
+ putchar(cu);
+ }
+
+ if (0 != state) {
+ /* Bad trailing bits. */
+ return(0);
+ }