We are running a survey to help us improve the experience for all of our members. If you see the survey appear, please take the time to tell us about your experience if you can.
Simulation of following code, compiled with V6.12 and optimzation level -O2
#include <stdio.h> typedef struct { char data[23]; } device_t; static device_t deviceList[100]; static device_t* rotate(device_t *device) { int index; index = (device - &deviceList[0]) + 1; if (index < 0 || index > 30) { index = 0; } printf("index = %d\n", index); return &deviceList[index]; } int main(void) { unsigned int n = 0; while (1) { printf("n = %d: ", n); rotate((device_t*)(n)); n += 1000; } }
prints n = 0: index = 0 n = 1000: index = 0 n = 2000: index = 0 n = 3000: index = 0 n = 4000: index = 0 n = 5000: index = 0 n = 6000: index = 163395741 n = 7000: index = 350133493 n = 8000: index = 536871245 n = 9000: index = 723608997 n = 10000: index = 910346749 n = 11000: index = 1097084501 n = 12000: index = 1283822253 n = 13000: index = 1470560005 n = 14000: index = 1657297757 n = 15000: index = 1844035509 into Debug (printf) Viewer, which is obviously incorrect. Maybe a bug, or do I missunderstand something?
If device does not point to an valid object in list, any list element could be returned.
Except that that part of the intended function is just impossible to implement without causing undefined behaviour via C99 6.5.6 or 6.5.8 along the way. In other words: forget about it.
If that's what you have to achieve, you have to change the signature from handling pointers to handling indices into the array.
In short: in C, there's no such thing as deriving a trustworthy pointer from garbage input. The only checks you can do on untrustworthy pointers are trivial variations on
(ptr != 0) (ptr == 0)
Thanks for your advice.