I see your points, you are definitely right :)
G
Il 09/Set/2015 14:45, "Jef Driesen" jef@libdivecomputer.org ha scritto:
On 2015-09-09 14:04, Giorgio Marzano wrote:
I have compiled this sniplet:
main (int argc, char *argv[]) { int i; const mares_iconhd_model_t models[] = { {"Matrix", MATRIX}, // {"Smart Apnea", SMARTAPNEA}, {"Smart", SMART}, {"Icon HD", ICONHD}, {"Icon AIR", ICONHDNET}, {"Puck Pro", PUCKPRO}, {"Nemo Wide 2", NEMOWIDE2}, {"Puck 2", PUCK2}, };
// Check the product name in the version packet against the list // with valid names, and return the corresponding model number. unsigned int model = 0; for ( i = 0; i < 7; ++i) { printf ("i: %d, sizeof: %d, strlen: %d\n",i,sizeof(models[i].name),strlen (models[i].name));
} }
and the corresponding ouput is:
giorgio@giorgio-laptop:~$ ./prova i: 0, sizeof: 17, strlen: 6 i: 1, sizeof: 17, strlen: 5 i: 2, sizeof: 17, strlen: 7 i: 3, sizeof: 17, strlen: 8 i: 4, sizeof: 17, strlen: 8 i: 5, sizeof: 17, strlen: 11 i: 6, sizeof: 17, strlen: 6
So it seems to me that either we use strlen or we use strcmp
No, the memcmp is correct. In the mares header, the name is a 16 byte array. If the name is shorter than 16 bytes, the remaining bytes are padded with zero's. But if there would be a name that's exactly 16 bytes long, then there won't be a terminating zero byte. In that case, strlen and strcmp will result in a buffer overflow! That's why we use memcmp. The other reason is that I want to do an exact match. The "Smart" entry should not match "Smart Apnea". Future models needs to be added explicitly, and not detected by accident.
Jef