Thread: Unusual 'malloc()' behaviour...

Take a look at the following c function:

Code:

struct neuron*
new_neuron(char* name, boolean state, float threshold)
{
  struct neuron* current = malloc((size_t) sizeof(struct neuron*));
  struct synapse** spare = malloc((size_t) sizeof(struct synapse**)); /* <-- #2 */
  
  // If there's no room for the neuron, report to standard error and exit.
  if (current == NULL)
  {
    fprintf(stderr, "ERROR: THE MALLOC() FUNCTION FAILED ON A NEURON!\n");
    fprintf(stderr, "Named '%s'\n.", name);
    exit(NO_MEMORY_FOR_NEURON_ERROR);
  }
  
  // Set aside memory to store the synapses.
  (*current).dendrites = malloc((size_t) sizeof(struct synapse**)); /* <-- #1 */
  (*current).terminals = malloc((size_t) sizeof(struct synapse**));
  
  // If there's no room for the synapses, report to standard error and exit.
  if ((*current).dendrites == NULL || (*current).terminals == NULL)
  {
    fprintf(stderr, "ERROR: THE MALLOC() FUNCTION FAILED ON A SYNAPSE!\n");
    fprintf(stderr, "Within '%s'\n.", name);
    exit(NO_MEMORY_FOR_SYNAPSE_ERROR);
  }
  
  // Define and return a neuron.
  (*current).name = name;
  (*current).state = state;
  (*current).threshold = threshold;
  (*current).dendrite_count = 0;
  (*current).terminal_count = 0;
  return current;
}

It works just fine and does exactly what I need it to, but for some peculiar reason, the first 'struct synapse**' variable I apply 'malloc()' to, showcases some very unusual behaviour and doesn't work properly, despite the fact that all subsequent ones work fine.

Initially, the line marked "#1" was the first time I applied 'malloc()' to a 'struct synapse**', but when that resulted in weird behaviour, I added the line marked "#2" in order to salvage the rest of the function with a quick-and-dirty solution. That particular variable ("spare") doesn't do anything, so the fact that it's broken is irrelevant, but I'm really curious why it's broken! Someone enlighten me please.

You are acting as though current has enough memory to store a struct neuron. Since you allocate enough memory for a pointer to a struct neuron, that assumption is incorrect. You should allocate the correct size in your line #1 (the typical way is to do sizeof(*current), but you can use the type name itself).

Ha ha, thank you both. I just CANNOT believe I made such a hideously simple mistake! I should have known better.

Why was I able to get away with the exact same thing the second, third, or nth time though? That's probably why it snuck under my radar...