Memcpy causing segfault when copying struct arrays - c

Question

I'm trying to make a rehashing function that will work on very large hash tables (with more than 1 million entries) and my current method is very inefficient. Here are my structures
typedef struct {
int id, count, correct, valid;
char* word;
} Entry;
typedef struct {
Entry* arr;
int size, numValid;
} Hash;
Here is how it functions now (very slowly, not using memcpy):
void rehash(Hash* hash) {
Entry* tempArr;
int i;
tempArr = calloc(hash->size, sizeof(Entry));
for(i = 0; i < hash->size; i++) {
if(hash->arr[i].count) {
tempArr[i].count = hash->arr[i].count;
tempArr[i].correct = hash->arr[i].valid;
tempArr[i].word = malloc(strlen(hash->arr[i].word) + 1);
strcpy(tempArr[i].word,hash->arr[i].word);
tempLen++;
}
memcpy(&tempArr[i],&hash->arr[i],sizeof(Entry));
tempArr[i] = hash->arr[i];
}
removeAllEntries(hash);
resize(hash);
for(i = 0; i < (hash->size / 2); i++) {
if(tempArr[i].count > 0) {
addEntry(hash,tempArr[i].word,tempArr[i].count);
/*printf("Added %s with count %d\n",tempArr[i].word,tempArr[i].count);*/
free(tempArr[i].word);
}
}
free(tempArr);
}
I'd prefer to use memcpy, but I can't for the life of me get it to work correctly. Here's what I'm trying (this is the code that doesn't work, and what I'm looking for help with):
void rehash(Hash* hash) {
Entry* tempArr;
int i;
tempArr = calloc(hash->size, sizeof(Entry));
fprintf(stderr,"size: %d\n",hash->size * sizeof(Entry));
memcpy((tempArr),(hash->arr),hash->size * sizeof(Entry));
removeAllEntries(hash);
resize(hash);
for(i = 0; i < (hash->size / 2); i++) {
if(tempArr[i].count > 0) {
addEntry(hash,tempArr[i].word,tempArr[i].count);
/*printf("Added %s with count %d\n",tempArr[i].word,tempArr[i].count);*/
free(tempArr[i].word);
}
}
free(tempArr);
}
I'm sure it's an easy, one-line fix, but I just can't get myself to see it.
void addEntry(Hash* hash, char* tag, int count) {
int value = CHV(hash->size, tag), flag = 1, iter = 0;
int possIndex = findEntry(hash, tag);
/*fprintf(stderr,"AddEntry...\n");*/
if(possIndex >= 0) {
(hash->arr[possIndex].count)++;
return;
}
if((hash->size - hash->numValid) < ((double)hash->size / 10))
{
rehash(hash);
}
while(flag) {
iter++;
if(!(hash->arr[value].valid)) {
hash->arr[value].word = calloc(strlen(tag) + 1, sizeof(char));
strcpy(hash->arr[value].word, tag);
wordsAlloced++;
hash->arr[value].valid = 1;
hash->arr[value].correct = 1;
hash->arr[value].count = count;
flag = 0;
}
else {
value++;
if(value == hash->size) {
value = 0;
}
}
}
hash->numValid++;
}

Answer 1

I think that memcpy is no problem. problem is free all Entry by removeAllEntries(hash); then tempArr[i].word is the double free in free(tempArr[i].word); Because it's copied pointer. Also Using tempArr[i].word in addEntry(hash,tempArr[i].word,tempArr[i].count); is invalid. It's already free'd.
One solution proposes the use of realloc.
replace
void resize(Hash* hash) {
free(hash->arr);
hash->size *= 2;
hash->arr = calloc(hash->size, sizeof(Entry));
//TOTALALLOC += (hash->size * sizeof(Entry));
}
with
void resize(Hash* hash) {
Entry* tempArr;
if((tempArr = realloc(hash->arr, 2 * hash->size * sizeof(Entry)))==NULL){
fprintf(stderr,"failed realloc in resize.\n");
return ;
}
hash->size *= 2;
hash->arr = tempArr;
//TOTALALLOC += (hash->size * sizeof(Entry));
}
Does not rehash for resizing purpose.
Another solution
If rehashing is necessary for some reason. to change in the following
void rehash(Hash* hash) {
Entry* tempArr;
int i;
tempArr = malloc(hash->size * sizeof(Entry));//Initialization isn't required because it is replaced by memcpy.
//fprintf(stderr,"size: %d\n",hash->size * sizeof(Entry));
memcpy(tempArr,hash->arr, hash->size * sizeof(Entry));
//To replicate word
for(i = 0; i < hash->size; i++) {
if(hash->arr[i].count) {
tempArr[i].word = malloc(strlen(hash->arr[i].word) + 1);
strcpy(tempArr[i].word, hash->arr[i].word);
}
}
removeAllEntries(hash);
resize(hash);
for(i = 0; i < (hash->size / 2); i++) {
if(tempArr[i].count > 0) {
addEntry(hash,tempArr[i].word, tempArr[i].count);
/*printf("Added %s with count %d\n",tempArr[i].word,tempArr[i].count);*/
free(tempArr[i].word);
}
}
free(tempArr);
}
If numValid is the represent the valid registration number, I think that it is sufficient to save only a word and count.