[Lldb-commits] [lldb] [LLDB] Add more helper functions to ValueObject class. (PR #87197)

[Greg Clayton via lldb-commits]

================
@@ -1089,6 +1089,116 @@ int64_t ValueObject::GetValueAsSigned(int64_t fail_value, bool *success) {
   return fail_value;
 }
 
+llvm::APSInt ValueObject::GetValueAsAPSInt() {
+  lldb::TargetSP target = GetTargetSP();
+  uint64_t byte_size = 0;
+  if (auto temp = GetCompilerType().GetByteSize(target.get()))
+    byte_size = temp.value();
+
+  unsigned bit_width = static_cast<unsigned>(byte_size * CHAR_BIT);
+  bool success = true;
+  uint64_t fail_value = 0;
+  uint64_t ret_val = GetValueAsUnsigned(fail_value, &success);
+  uint64_t new_value = fail_value;
+  if (success)
+    new_value = ret_val;
+  bool is_signed = GetCompilerType().IsSigned();
+
+  return llvm::APSInt(llvm::APInt(bit_width, new_value, is_signed), !is_signed);
+}
+
+llvm::APFloat ValueObject::GetValueAsFloat() {
+  lldb::BasicType basic_type =
+      GetCompilerType().GetCanonicalType().GetBasicTypeEnumeration();
+  lldb::DataExtractorSP data_sp(new DataExtractor());
+  Status error;
+
+  switch (basic_type) {
+  case lldb::eBasicTypeFloat: {
+    float v = 0;
----------------
clayborg wrote:

We really don't want to use native float types, we want to use real APFloat values that are correctly encoded. For instance if we are debugging from x86_64, we don't want to create a 10 byte Intel float and do the math with the native float, we want to use the value from the Scalar that is already set correctly. If we are cross debugging to an iPhone, we want the floating point values to match exactly, which requires us to use the APFloat and not do any manually create a native float and convert it. This all works for the value objects when they get the float as a string, and will be similar where we resolve the scalar and then can just use the Scalar::m_float directly.

https://github.com/llvm/llvm-project/pull/87197