https://bugs.llvm.org/show_bug.cgi?id=48735
Bug ID: 48735
Summary: Difference between GCC and Clang on a defined program:
GCC's VLA is Clang fixed-size automatic array
Product: clang
Version: 11.0
Hardware: PC
OS: All
Status: NEW
Severity: enhancement
Priority: P
Component: C
Assignee: [email protected]
Reporter: [email protected]
CC: [email protected], [email protected],
[email protected], [email protected],
[email protected]
Consider the compilation unit:
int t[12];
#define ADDR_FIRST (&(t[0]))
#define ADDR_LAST (&(t[11]))
int Y, Z;
void f(void) {
// Define u the same size as t:
long u[ADDR_LAST - ADDR_FIRST + 1];
Y = sizeof *(Z=1, &u);
}
Compiler Explorer link: https://gcc.godbolt.org/z/fa8frb
When u is an automatic variable, GCC 10.2 treats u, declared as above, as a
VLA, and therefore generates code for the function f that assigns Z:
f:
movl $1, Z(%rip)
movl $96, Y(%rip)
ret
Meanwhile Clang 11.0.0 treats the expression “ADDR_LAST - ADDR_FIRST + 1” as a
constant expression, which I think it is entitled to per C17 6.6:10
(https://cigix.me/c17#6.6.p10 ), meaning that u is a fixed-size array and that
f should not assign Z:
f: # @f
movl $96, Y(%rip)
retq
Note that the two compilers's behaviors agree when u is a static-lifetime
variable. In that case GCC emits a somewhat misleading warning about u having
file scope (whereas it should be the lifetime of u that's the problem) and
variable size, and subsequently treats u as a fixed-size array. Changing the
recognition of constant expressions to be stricter for all array sizes would
make Clang reject a static-lifetime definition for u, which GCC accepts.
--
You are receiving this mail because:
You are on the CC list for the bug._______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
