[llvm] [AMDGPU] Add disassembler diagnostics for invalid kernel descriptors (PR #87400)

[Emma Pilkington via llvm-commits]

================
@@ -160,7 +161,7 @@ class MCDisassembler {
   ///                   case.
   virtual std::optional<DecodeStatus>
   onSymbolStart(SymbolInfoTy &Symbol, uint64_t &Size, ArrayRef<uint8_t> Bytes,
-                uint64_t Address, raw_ostream &CStream) const;
+                uint64_t Address, raw_ostream &ErrS) const;
----------------
epilk wrote:

Using an `Error` to represent errors seems like a nice refactor, but an Error return isn't sufficient for this function. It has three cases: a) nullopt, meaning the disassembler didn't have any handling for this symbol, b) Fail, meaning an error was encountered, and c) Success. b & c both indicate to objdump that some bytes were consumed, so `Size` must be set. In the update I changed the return type to bool, where true means the target disassembler had some handling for this symbol, with an Error out param to indicate whether it contained errors.

https://github.com/llvm/llvm-project/pull/87400