On 3/17/20 2:39 AM, Jakub Jelinek wrote:
Hi!The gcc.dg/pr68785.c test which contains: int foo (void) { return *(int *) ""; } has UB in the program if it is ever called, but causes UB in the compiler as well as at least in theory non-reproduceable code generation. The problem is that nbytes is in this case 4, prep is the TREE_STRING_POINTER of a "" string literal with TREE_STRING_LENGTH of 1
At least as important as avoiding the Valgrind error is detecting the bug in the code. -Warray-bounds has all it needs to diagnose it, but, regrettably, it doesn't. I raised PR 94195 to make sure it does. Martin and
we do: 4890 for (const char *p = prep; p != prep + nbytes; ++p) 4891 if (*p) 4892 { 4893 *allnul = false; 4894 break; 4895 } and so read the bytes after the STRING_CST payload, which can be random. I think we should just punt in this case. Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk? 2020-03-16 Jakub Jelinek <ja...@redhat.com> PR tree-optimization/94187 * tree-ssa-strlen.c (count_nonzero_bytes): Punt if nchars - offset < nbytes. --- gcc/tree-ssa-strlen.c.jj 2020-03-14 08:14:47.034742349 +0100 +++ gcc/tree-ssa-strlen.c 2020-03-16 12:23:57.523534887 +0100 @@ -4822,6 +4822,8 @@ count_nonzero_bytes (tree exp, unsigned of the access), set it here to the size of the string, including all internal and trailing nuls if the string has any. */ nbytes = nchars - offset; + else if (nchars - offset < nbytes) + return false;prep = TREE_STRING_POINTER (exp) + offset;} Jakub
