[clang-tools-extra] [clang-tidy] Make `P +- BS / sizeof(*P)` opt-outable in `bugprone-sizeof-expression` (PR #111178)

[via cfe-commits]

================
@@ -352,21 +352,30 @@ void good13(void) {
   int Buffer[BufferSize];
 
   int *P = &Buffer[0];
-  while (P < (Buffer + sizeof(Buffer) / sizeof(int))) {
+  while (P < Buffer + sizeof(Buffer) / sizeof(int)) {
     // NO-WARNING: Calculating the element count of the buffer here, which is
     // safe with this idiom (as long as the types don't change).
     ++P;
   }
 
-  while (P < (Buffer + sizeof(Buffer) / sizeof(Buffer[0]))) {
+  while (P < Buffer + sizeof(Buffer) / sizeof(Buffer[0])) {
     // NO-WARNING: Calculating the element count of the buffer here, which is
     // safe with this idiom.
     ++P;
   }
 
-  while (P < (Buffer + sizeof(Buffer) / sizeof(*P))) {
+  while (P < Buffer + sizeof(Buffer) / sizeof(*P)) {
     // NO-WARNING: Calculating the element count of the buffer here, which is
     // safe with this idiom.
     ++P;
   }
 }
+
+void situational14(int *Buffer, size_t BufferSize) {
+  int *P = &Buffer[0];
+  while (P < Buffer + BufferSize / sizeof(*Buffer)) {
+    // CHECK-MESSAGES: :[[@LINE-1]]:21: warning: suspicious usage of 'sizeof(...)' in pointer arithmetic; this scaled value will be scaled again by the '+' operator
----------------
whisperity wrote:

I am not saying "_with_ the", I am saying "**by** the", which highlights the difference. Pointer arithmetic operator `+` does an internal scaling, cathing the misuse of this is the goal of the improvement that was done in the previous patch.

> scaled value will be used again with the other operand

This does not explain why it is an issue, only reiterates what is visible otherwise from the code.

https://github.com/llvm/llvm-project/pull/111178