Bug 250019

Comment 1 Conrad Meyer 2020-10-02 16:22:36 UTC

Hi Samee,

This isn’t a bug and the change does not impact the meaning of the code in any way.

Comment 2 Warner Losh 2020-10-02 16:25:47 UTC

Can you explain the headaches?

sizeof(char) == 1, so 1 * 256 is 256 and wouldn't change behavior at all.

What's the bug this is trying to fix?

sizeof(char) == 1

and if you:

#include <stdio.h>
#include <string.h>

int main(int argc, char **argv) {
char s1[256];
char s2[256];

memset(s1, 0, 256);
memset(s2, 0, (sizeof(char)*256));

printf("%s", s1);
printf("%s", s2);


        return(0);
}

everything will be fine but if you:

int main(int argc, char **argv) {
char s1[256];
char s2[256];

blah(s1, s2);

        return(0);
}

void blah(char *s1, char *s2) {

memset(s1, 0, 256);
memset(s2, 0, (sizeof(char)*256));

printf("%s", s1);
printf("%s", s2);

}

the one without the sizeof will frequently spit out garbage. 

I dont know why but i think sometimes the compiler mistakes
(char*) for (char).

and even memset((char*)s1, 0, 256) wont fix it.

I think the compiler is doing this:

memset((char*)s2, 0, (sizeof(char*)*256));

but this should result in a core dump and not a buffer filled with garbage.

I dont know. Something to think about.

Comment 4 Ed Maste 2020-10-02 19:05:21 UTC

(In reply to Samee Shahzada from comment #3)
Turn on compiler warnings and it will tell you what the problem is:

$ cc -Wall pr250019.c 
pr250019.c:5:1: warning: implicit declaration of function 'blah' is invalid in C99 [-Wimplicit-function-declaration]
blah(s1, s2);
^
pr250019.c:10:6: error: conflicting types for 'blah'
void blah(char *s1, char *s2) {
     ^
pr250019.c:5:1: note: previous implicit declaration is here
blah(s1, s2);
^
pr250019.c:12:1: warning: implicitly declaring library function 'memset' with type 'void *(void *, int, unsigned long)' [-Wimplicit-function-declaration]
memset(s1, 0, 256);
^
pr250019.c:12:1: note: include the header <string.h> or explicitly provide a declaration for 'memset'
pr250019.c:15:1: warning: implicitly declaring library function 'printf' with type 'int (const char *, ...)' [-Wimplicit-function-declaration]
printf("%s", s1);
^
pr250019.c:15:1: note: include the header <stdio.h> or explicitly provide a declaration for 'printf'

that was the compiler complaining that I didnt first declare blah and need to include stdio. Thats because the code was incomplete.

I was showing you where and how the bug manifests. 

I promise you. 

memset(s1, 0, 256); gonna frequently confuse the compiler and will frequently not memset the buffer correctly. The compiler will not give you a warning and you are going to pull out your hair trying to figure out what's wrong. 

I really think something like memset(s1, 0, (sizeof(char*)*256)); is happening.

Its technically a bug in the compiler and not with FreeBSD but that fact does not put the hair back in your head.

Comment 6 Warner Losh 2020-10-02 19:52:47 UTC

> I really think something like memset(s1, 0, (sizeof(char*)*256)); is happening.

No. That's simply not possible. A compiler that randomly screws up simple integer constants is too broken to use. And 1*constant really isn't going to help things much unless it's a super specific bug.

Since there's nothing specific and actionable listed as a root cause, we should reject this submission.

Comment 7 Ed Maste 2020-10-02 20:52:35 UTC

(In reply to Samee Shahzada from comment #5)
There is no compiler bug here. If you want to convince yourself, compile your source in "memset(foo, 0, 256)" and "memset(foo, 0, sizeof(char)*256)" versions to two separate object files and compare the results. You will find that they are identical.