Boyer Moore k-mismatches algorithm fails - c++

Question

I've done a program for string comparison with one mismatch at a programming website. It gives me wrong answer. I've working on it extensively but, I couldn't find testcases where my code fails. Can somebody provide me test cases where my code fails. I've done the comparison using Boyer Moore Horspool k-mismatches algorithm as it's the fastest searching algorithm
The code is as such
int BMSearch_k(string text, string pattern, int tlen, int mlen,int pos)
{
int i, j=0,ready[256],skip2[256][mlen-1],neq;
for(i=0; i<256; ++i) ready[i] = mlen;
for(int a=0; a<256;a++) {
for(i = mlen;i>mlen-k;i--)
skip2[i][a] = mlen;
}
for(i = mlen-2;i>=1;i--) {
for(j=ready[pattern[i]]-1;j>=max(i,mlen-k);j--)
skip2[j][pattern[i]] = j-i;
ready[pattern[i]] = max(i,mlen-k);
}
j = mlen-1+pos;
//cout<<"\n--jafffa--\n"<<pos<<"+"<<mlen<<"="<<j<<endl;
while(j<tlen+k) {
//cout<<"\t--"<<j<<endl;
int h = j;
i=mlen-1;
int neq=0,shift = mlen-k;
while(i>=0&&neq<=k) {
//cout<<"\t--"<<i<<endl;
if(i>=mlen-k)
shift = min(shift,skip2[i][text[h]]);
if(text[h]!= pattern[i])
neq++;
i--;
h--;
}
if(neq<=k)
return j-1;
j += shift;
}
return -1;
}

Answer 1

You aren't initialising your arrays correctly,
int i, j=0,ready[256],skip2[256][mlen-1],neq;
for(i=0; i<256; ++i) ready[i] = mlen;
for(int a=0; a<256;a++) {
for(i = mlen;i>mlen-k;i--)
skip2[i][a] = mlen;
}
On the one hand, you declare skip2 as a 256×(mlen-1) array, on the other hand, you fill it as a (mlen+1)×256 array.
In the next loop,
for(i = mlen-2;i>=1;i--) {
for(j=ready[pattern[i]]-1;j>=max(i,mlen-k);j--)
skip2[j][pattern[i]] = j-i;
ready[pattern[i]] = max(i,mlen-k);
}
you use ready[pattern[i]] before it has been set. I don't know if those mistakes are what's causing the failing testcase, but it's easily imaginable that they do.

Answer 2

If Daniel's suggestions do not solve the problem, here are a couple more things that look odd:
return j-1; // I would expect "return j;" here
This seems odd as if you have k=0,mlen=1, then the highest value that j can take is tlen+k-1, and so the highest return value is tlen-2. In other words matching a pattern 'a' against a string 'a' will not return a match at position 0.
Another oddity is the loop:
for(i = mlen-2;i>=1;i--) // I would expect "for(i = mlen-2;i>=0;i--)" here
it seems odd that in the preprocessing you will never access the first character in your pattern (i.e. pattern[0] is not read).