| Message ID | 4F0F0DD3.60405@gjlay.de |
|---|---|
| State | New |
| Headers | show |
On Jan 12, 2012, at 8:44 AM, Georg-Johann Lay wrote: > Again: A test case that fails because it incorrectly assumes int is 32 bits wide. > > Ok to apply? I'd think the testcase is more naturally: > return ~((uint32_t) (arg > -3)); Does that work for you? Anyone want to weigh in on which version is better (if the second version works)? I think I have a preference for the form above. > - return ~(arg > -3); > + return ~((unsigned) (arg > -3));
Mike Stump schrieb: > On Jan 12, 2012, at 8:44 AM, Georg-Johann Lay wrote: > >>Again: A test case that fails because it incorrectly assumes int is 32 bits wide. >> >>Ok to apply? > > I'd think the testcase is more naturally: > >>return ~((uint32_t) (arg > -3)); That not equivalent to the proposed patch: It extends to 32 and then inverts all bits. > Does that work for you? Anyone want to weigh in on which version is better (if the second version works)? I think I have a preference for the form above. > >>- return ~(arg > -3); >>+ return ~((unsigned) (arg > -3)); This inverts the 16 bits of unsigned and then zero-extends to 32 bits (the original code sign extends to 32 bits as "arg > 3" is int (16 bits).
Index: gcc.c-torture/execute/20120111-1.c =================================================================== --- gcc.c-torture/execute/20120111-1.c (revision 183128) +++ gcc.c-torture/execute/20120111-1.c (working copy) @@ -6,7 +6,7 @@ uint32_t f0a (uint64_t arg2) __attribute uint32_t f0a (uint64_t arg) { - return ~(arg > -3); + return ~((unsigned) (arg > -3)); } int main() {