[clang] [clang][analyzer] Correct SMT Layer for _BitInt cases refutations (PR #143310)

[Balazs Benics via cfe-commits]

================
@@ -570,23 +572,35 @@ class SMTConv {
   // TODO: Refactor to put elsewhere
   static inline QualType getAPSIntType(ASTContext &Ctx,
                                        const llvm::APSInt &Int) {
-    return Ctx.getIntTypeForBitwidth(Int.getBitWidth(), Int.isSigned());
+    QualType Ty;
+    if (!(Ty = Ctx.getIntTypeForBitwidth(Int.getBitWidth(), Int.isSigned()))
+             .isNull())
+      return Ty;
+    // If Ty is Null, could be because the original type was a _BitInt.
+    // Get the bit size and round up to next power of 2, max char size
+    unsigned CharTypeSize = Ctx.getTypeSize(Ctx.CharTy);
+    unsigned Pow2DestWidth =
+        std::max(llvm::bit_ceil(Int.getBitWidth()), CharTypeSize);
+    Ty = Ctx.getIntTypeForBitwidth(Pow2DestWidth, Int.isSigned());
+    return Ty;
----------------
steakhal wrote:

I think we could just directly return this type instead of assigning it and returning what we just assigned.

https://github.com/llvm/llvm-project/pull/143310