Pango 1.50.14 triggers a false positive "source object is likely at address zero" with GCC 13 (build failure) in pango-language.c

Originally described here => https://bugs.gentoo.org/903259

With GCC 13 I have a build failure:

In function ‘find_best_lang_match_cached’,
    inlined from ‘find_best_lang_match_cached’ at ../pango-1.50.14/pango/pango-language.c:501:1,
    inlined from ‘pango_language_get_scripts’ at ../pango-1.50.14/pango/pango-language.c:661:21:
../pango-1.50.14/pango/pango-language.c:518:12: error: array subscript 0 is outside array bounds of ‘const void *[0:]’ [-Werror=array-bounds=]
  518 |     *cache = result;
      |     ~~~~~~~^~~~~~~~
In function ‘pango_language_get_scripts’:
cc1: note: source object is likely at address zero
cc1: some warnings being treated as errors

To circumvent it, I tried to enclose the function find_best_lang_match_cached() in pango-language.c with #pragmas like this:

#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Warray-bounds"

static gconstpointer
find_best_lang_match_cached (PangoLanguage *language,
                             gconstpointer *cache,
                             gconstpointer  records,
                             guint          num_records,
                             guint          record_size)
{
  gconstpointer result;

  if (G_LIKELY (cache && *cache != (gconstpointer) -1))
    return *cache;

  result = find_best_lang_match (language,
                                 records,
                                 num_records,
                                 record_size);

  if (cache)
    *cache = result;

  return result;
}

#pragma GCC diagnostic pop

Reading the code a bit, I do not see a reason to have a null-value for the pointer "result" but the compiler is not happy with the code :)