The minimal code needed to reproduce the ZDICT_trainFromBuffer crash

zstd_crash_reader.c

#include <assert.h>
#include <fcntl.h>
#include <malloc.h>
#include <stdio.h>
#include <string.h>
#include <unistd.h>

#include <sys/stat.h>

#include <zdict.h>
#include <zdict_static.h>

typedef struct my_buffer {
    char* data;
    size_t size;
} my_buffer_t;

size_t get_size(int fd) {
    struct stat s;
    memset(&s, 0, sizeof(s));
    assert(fstat(fd, &s) != -1);
    return s.st_size;
}

my_buffer_t read_buffer(char* fname) {
    int fd = open(fname, O_RDONLY);
    assert(fd >= 0);
    struct my_buffer b;
    b.size = get_size(fd);
    b.data = malloc(b.size);
    assert(b.data);
    assert(read(fd, b.data, b.size) == b.size);
    close(fd);
    return b;
}

int main() {
    my_buffer_t samples = read_buffer("ZDICT_trainFromBuffer_samplesBuffer.bin");
    my_buffer_t lengths = read_buffer("ZDICT_trainFromBuffer_samplesSizes.64bitLE.bin");

    size_t nbSamples = lengths.size / sizeof(size_t);
    size_t capacity = samples.size + 8 * lengths.size;

    char* buffer = malloc(capacity);
    assert(buffer);
    printf("%zu %zu %zu %zu\n", samples.size, lengths.size, nbSamples, capacity);

    ZDICT_params_t params;
    memset(&params, 0, sizeof(params));
    params.notificationLevel = 3;
    ZDICT_trainFromBuffer_advanced(buffer, capacity, samples.data, (const size_t*)lengths.data, nbSamples, params);
}