================
@@ -0,0 +1,37 @@
+// RUN: %clang_cc1 -std=c++20 -Wunsafe-buffer-usage -verify %s
+
+namespace std {
+ typedef __SIZE_TYPE__ size_t;
+
+ template <typename CharT>
+ class basic_string_view {
+ public:
+ basic_string_view(const CharT *s, size_t count);
+ basic_string_view(const CharT *s);
+ template <typename It>
+ basic_string_view(It begin, It end);
+ };
+
+ using string_view = basic_string_view<char>;
+}
+
+void test_string_view_scenarios(const char *ptr, std::size_t n) {
+ // --- UNSAFE CASES ---
+ std::string_view sv1(ptr, n);
+ // expected-warning@-1 {{the two-parameter std::string_view construction is
unsafe as it can introduce mismatch between buffer size and the bound
information}}
+
+ // --- SAFE CASES (Should NOT warn for string_view construction) ---
+ std::string_view sv2("hello", 5); // Safe: String literal match
+ std::string_view sv3(ptr, 0); // Safe: Zero size
+ std::string_view sv4(ptr); // Safe: Single parameter
+
+ // Note: 'ptr + 5' triggers general pointer arithmetic warning,
+ // but the string_view constructor itself is ignored by your gadget.
+ std::string_view sv5(ptr, ptr + 5); // expected-warning {{unsafe pointer
arithmetic}} \
----------------
rohanjr wrote:
Can we add a test for this constructor that doesn't trigger an error? E.g. with
`s.begin()` and `s.end()` as arguments? I think that would also exercise the
code path where the argument types are record/object types (not pointers).
https://github.com/llvm/llvm-project/pull/180471
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
