https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123256
Bug ID: 123256
Summary: Possible false positive in -Wuninitialized warning
with pointer type conversion and bit-field access
Product: gcc
Version: 14.3.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
Assignee: unassigned at gcc dot gnu.org
Reporter: jiangxuezhi2 at huawei dot com
Target Milestone: ---
I've encountered what appears to be a false positive with the -Wuninitialized
warning.
$ gcc test.c -O2 -Wextra -S -o test.s
#include <stdint.h>
#define MAX_ID 10
typedef struct {
uint32_t id : 6;
} Header;
int checkHeader(uint32_t a, uint32_t b)
{
uint32_t buf[2];
Header *h;
buf[0] = a;
buf[1] = b;
h = (Header *)(void *)buf;
volatile uint32_t result = h->id;
return result > MAX_ID ? 0 : 1;
}
the output:
test.c: In function 'checkHeader':
test.c:39:33: warning: 'buf' is used uninitialized [-Wuninitialized]
39 | volatile uint32_t result = h->id;
| ~^~~~
test.c:33:14: note: 'buf' declared here
33 | uint32_t buf[2];
| ^~~
Both elements of the array data are explicitly initialized, and the bit-field
id accesses only the initialized low 6 bits of data[0]. Given this, I think the
-Wuninitialized warning may be unnecessary for this case.
This seems to be related to the recent enhancements to uninitialized variable
detection (specifically the patch addressing PR middle-end/10138 and PR
middle-end/95136), which I understand were meant to improve detection of
legitimate bugs.
If this is indeed expected behavior, I'd appreciate understanding the
reasoning,Thanks.
