Hello,

GNU Libidn uses code from GLIB to implement Unicode NFKC normalization. 
During interoperability testing, it was discovered that the NFKC operation
may have a problem.  I have reproduced the problem in GLIB standalone as
well.  The problem is that Hangul composition is not performed.  See:

http://www.unicode.org/reports/tr15/tr15-22.html#Hangul

for the details how to implement this.  Unicode Inc. actually has test
vectors for Unicode normalization, I think it may be useful to test GLIB
against it.  I suspect this would have catched this problem.  See:

http://www.unicode.org/Public/UNIDATA/NormalizationTest.txt

A simple test case is the following code snippet.

If I can be of any assistance in tracking down or analyzing this problem,
let me know, I'm currently working on a fix, but thought I should let you
know ASAP.

#include <stdio.h>
#include <glib.h>

/* Unicode NFKC of algorithmic Hangul composition, by Simon Josefsson */
int main ()
{
  const gchar *in = "\xef\xbf\xbd\xc3\xad\x64\x6e";
  gchar *out;
  size_t i;

  printf("in: ");
  for (i = 0; in[i]; i++)
    printf("%02x ", in[i] & 0xFF);
  printf("\n");

  out = g_utf8_normalize (in, strlen(in), G_NORMALIZE_NFKC);

  /* Should (?) result in \x2e\xea\b0\x81. */

  printf("out: ");
  for (i = 0; out[i]; i++)
    printf("%02x ", out[i] & 0xFF);
  printf("\n");

  return 0;
}