Coverage increased (+0.03%) to 84.158% when pulling d2cec5c on itchyny:fix-multiple-string-multiplication into ccc79e5 on stedolan:master.

I just tried the reproducing example from #2115 against this branch and it doesn't crash (and prints the correct result). Sanitizers and valgrind are looking clean too.

Good find! I'll suggest a slightly more efficient solution:

jv res = jv_null();
int n = jv_number_value(num);
if (n > 0) {
  size_t alen = jv_string_length_bytes(jv_copy(str));
  res = jv_string_empty(alen * n);

  for (; n > 0; n--) {
    res = jv_string_append_buf(res, jv_string_value(str), alen);
  }
}
jv_free(str);
jv_free(num);
return res;

@wtlangford I update the patch, thanks. But considering the performance here seriously, we should reduce the number of jvp_utf8_is_valid checks.

@wtlangford I update the patch, thanks. But considering the performance here seriously, we should reduce the number of jvp_utf8_is_valid checks.

Agreed, we should- but for now, I think this is good. Thanks for tracking this down!

I noticed that 0.5 * "abc" is changed by this pull request. Should we keep the original behavior?

Yes, we definitely should