[Lldb-commits] [lldb] [LLDB] Add array subscription and integer parsing to DIL (PR #138551)

[Ilia Kuklin via lldb-commits]

================
@@ -272,4 +272,67 @@ Interpreter::Visit(const UnaryOpNode *node) {
       m_expr, "invalid ast: unexpected binary operator", node->GetLocation());
 }
 
+llvm::Expected<lldb::ValueObjectSP>
+Interpreter::Visit(const ArraySubscriptNode *node) {
+  auto lhs_or_err = Evaluate(node->GetBase());
+  if (!lhs_or_err) {
+    return lhs_or_err;
+  }
+  lldb::ValueObjectSP base = *lhs_or_err;
+  const llvm::APInt *index = node->GetIndex();
+
+  Status error;
+  if (base->GetCompilerType().IsReferenceType()) {
+    base = base->Dereference(error);
+    if (error.Fail())
+      return error.ToError();
+  }
+
+  // Check to see if 'base' has a synthetic value; if so, try using that.
+  uint64_t child_idx = index->getZExtValue();
+  if (base->HasSyntheticValue()) {
+    lldb::ValueObjectSP synthetic = base->GetSyntheticValue();
+    if (synthetic && synthetic != base) {
+      uint32_t num_children = synthetic->GetNumChildrenIgnoringErrors();
----------------
kuilpd wrote:

> Is it possible that some (simple) data formatter implements GetChildAtIndex as return foo without checking whether the index argument is in range?

One thing I can add that I noticed: that code worked when I tried it from console lldb, but didn't work on the same executable and input string when called from Python tests. I'm not sure why.

I changed the call to `GetNumChildren`, I think I'd prefer a vector telling me that I'm out of bounds instead of returning 0 like it's a correct value from the vector.

https://github.com/llvm/llvm-project/pull/138551