https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110373
Bug ID: 110373
Summary: __builtin_object_size does not recognize subarrays in
multi-dimensional arrays
Product: gcc
Version: 13.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: siddhesh at gcc dot gnu.org
Target Milestone: ---
const char example[][32] = {
"abcd",
"xyzt",
"1234",
"5678"
};
__SIZE_TYPE__ foo (void) {
return __builtin_dynamic_object_size(example[1], 1);
}
$ gcc/cc1 -quiet -O -o - ~/src/random/fortify-module/fortify-module1.c
.file "fortify-module1.c"
.text
.globl foo
.type foo, @function
foo:
.LFB0:
.cfi_startproc
movl $96, %eax
ret
.cfi_endproc
.LFE0:
.size foo, .-foo
.globl example
.section .rodata
.align 32
.type example, @object
.size example, 128
example:
.string "abcd"
.zero 27
.string "xyzt"
.zero 27
.string "1234"
.zero 27
.string "5678"
.zero 27
.ident "GCC: (GNU) 14.0.0 20230622 (experimental)"
.section .note.GNU-stack,"",@progbits
Clang returns 32 instead, recognizing the subarray correctly. This appears to
be happening very early on in the frontend; addr_object_size doesn't see that
this is a subarray reference since it only compares the pt_var with the operand
of the ADDR_EXPR. If pt_var is an ARRAY_REF, it should also additionally see
if this is a subarray reference.