[NFC][Clang] Fix potential deref of end iterator

[Keenuts]

This was found by doing bound-checking on SmallVector iterator usage. When the count is 0, the end iterator is dereferenced to get its address. This doesn't seem to be an issue in practice as most of the time we should be allowed to deref this address, but I don't think this is correct.

[Keenuts]

CI seems OK, except clang format which complains about an unrelated line. Marking as ready.

[llvmbot]

@llvm/pr-subscribers-clang

Author: Nathan Gauër (Keenuts)

Changes

This was found by doing bound-checking on SmallVector iterator usage. When the count is 0, the end iterator is dereferenced to get its address. This doesn't seem to be an issue in practice as most of the time we should be allowed to deref this address, but I don't think this is correct.

---

Full diff: https://github.com/llvm/llvm-project/pull/70193.diff

1 Files Affected:

• (modified) clang/include/clang/Sema/CXXFieldCollector.h (+1-1)

diff --git a/clang/include/clang/Sema/CXXFieldCollector.h b/clang/include/clang/Sema/CXXFieldCollector.h
index f6ecd9f46e5ebdb..ce066581c93fda7 100644
--- a/clang/include/clang/Sema/CXXFieldCollector.h
+++ b/clang/include/clang/Sema/CXXFieldCollector.h
@@ -65,7 +65,7 @@ class CXXFieldCollector {
 
   /// getCurFields - Pointer to array of fields added to the currently parsed
   /// class.
-  FieldDecl **getCurFields() { return &*(Fields.end() - getCurNumFields()); }
+  FieldDecl **getCurFields() { return Fields.end() - getCurNumFields(); }
 
   /// FinishClass - Called by Sema::ActOnFinishCXXClassDef.
   void FinishClass() {

[shafik]

I think this is basically cwg232 this is ok b/c we are not doing an lvalue-to-rvalue conversion but only using it to take the address of.

CC @zygoloid

[Keenuts]

Thanks for the review and linked pointers!
This seems to be the case. and generated code does only the pointer arithmetic, no actual load. Closing this as this is not an issue.

btw, I like your C++ trivia on twitter 😊