I was recently bamboozled by my own code when I switched from compiling for 32-bits (i386) to 64-bits (x86_64). I had a bunch of scanf/printf lines reading and printing size_t variables.
Here’s a test program in C++ that demonstrates the correct usage of the z option for reading in size_t:
#include <cstdio> #include <limits> int main(int argc,char * argv[]) { // Show size of size_t printf("sizeof(size_t): %lu\n",(long unsigned int)(sizeof(size_t))); // Show max size_t value: (2^8)^sizeof(size_t) long unsigned int max_size_t = std::numeric_limits<std::size_t>::max(); printf("max_size_t: %lu\n",max_size_t); // read into size_t size_t a; int nr = sscanf(argv[1],"%zu",&a); // print size_t printf("argv[1]: %zu\n",a); return 0; }