Memleak in C API

[linas]

The program below leaks 64 bytes per spelling suggestion, or about a gigabyte in a few minutes. As far as I can tell, the program is written in exactly the style that the documentation calls for. Am I missing something?

This is on Debian stable, aspell version 0.60.8-3

I note that issue #600 reports a memleak, but its not clear if its the same issue. From what I can tell, the fix for #600 was never applied.

#include <malloc.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <aspell.h>

// Compile with
// cc aspell-memleak.c -laspell

int main()
{
        AspellConfig *config = new_aspell_config();

        aspell_config_replace(config, "lang", "en_US");

        AspellCanHaveError *spell_err = new_aspell_speller(config);
        AspellSpeller *speller = to_aspell_speller(spell_err);

        size_t k=0;
        char* word = "asdf";
        for (int l=0; l<1000000; l++)
        {
                /* Returns 1 is the word is in dict. */
                int found = aspell_speller_check(speller, word, -1);
                // printf("Found the word: %d\n", found);

                const AspellWordList *list = aspell_speller_suggest(speller, word, -1);
                AspellStringEnumeration *elem = aspell_word_list_elements(list);
                unsigned int size = aspell_word_list_size(list);

                const char *aword = NULL;
                while ((aword = aspell_string_enumeration_next(elem)) != NULL)
                {
                        // printf("Spell suggesion: %s\n", aword);
                        k++;
                }
                delete_aspell_string_enumeration(elem);

                if (0 == l%20000)
                {
                        printf("Loop count= %d spell suggests= %lu\n", l, k);
                        malloc_stats();
                }
        }

        delete_aspell_speller(speller);
        delete_aspell_config(config);
}

Debugged in opencog/link-grammar#1373

[kevina]

Thanks for the test case and sorry for the long delay. I can confirm that this is the same issue as #600 and is fixed by #601.