[llvm] [CFI] Fix Direct Call Issues in CFI Dispatch Table (PR #69663)

[Eli Friedman via llvm-commits]

================
@@ -1471,6 +1471,11 @@ void LowerTypeTestsModule::createJumpTable(
   SmallVector<Value *, 16> AsmArgs;
   AsmArgs.reserve(Functions.size() * 2);
 
+  // Check if the first entry has the NoUnwind attribute.
----------------
efriedma-quic wrote:

I haven't looked at this in a while, but I think the issue I was running into was that the EH info was causing issues for a linker script that was expecting no EH info.  Currently, the only way to suppress the generation of the ".cfi_startproc" (which genreates EH info) is "nounwind", I think.

Ideally, we would skip emitting EH info for all jump tables functions; even if the callee unwinds, the jump table isn't on the stack at that point anyway.  Maybe worth adding some way to represent that.

https://github.com/llvm/llvm-project/pull/69663