[llvm] [LLVM][ADT] Explicitly convert size_t values to SmallVector's size type (PR #77939)

[Andrew Browne via llvm-commits]

================
@@ -100,7 +100,18 @@ template <class Size_T> class SmallVectorBase {
   /// This does not construct or destroy any elements in the vector.
   void set_size(size_t N) {
     assert(N <= capacity());
-    Size = N;
+    Size = static_cast<Size_T>(N);
+    assert(size() == N);
+  }
+
+  /// Set the array capacity to \p M, which the current array size must
+  /// not be greater than.
+  ///
+  /// This does not construct or destroy any elements in the vector.
+  void set_capacity(size_t M) {
+    assert(size() <= M);
----------------
browneee wrote:

No, if the existing `assert(N <= capacity());` passes then I think `assert(M <= SizeTypeMax());` must be true.

I think:
* `set_size` should only check `assert(N <= capacity());`
* `set_capacity` should only check `assert(M <= SizeTypeMax());`

Using set_capacity to shrink (regardless of the size) doesn't make sense, because `set_capacity` would not actually handle shrinking (only a small piece of shrinking - switching to a smaller allocation).

https://github.com/llvm/llvm-project/pull/77939