Regression Fix : __jit_debug_descriptor bind on Windows O/S (updated - formatting)

[steve_w]

Summary

LLDB __jit_debug_descriptor fails to bind to a valid JIT delivery tool because regressed code attempts to find a symbol of type Data which isn't implemented, Windows codepath.

Hence workaround of searching for symbol of type Any is provided to address regression & re-enable Windows based LLDB GDB/JIT development work.

A working example is available (wasmtime), which demonstrates this fix re-enabling GDB/JIT debugging of web assembly client code.

Note that module also had a minor regression (symbols not exported on Windows), so will require trunk of that tree as fix has just landed.

[sgraenitz]

Thanks for working on this. The formatting nits should be straightforward to fix. Apart from that, it would be really nice if the jit-loader_jitlink/rtdyld_elf.test tests could reflect this fix. Otherwise it might regress again any time.

[steve_w]

You're welcome. Agree test(s) should exercise codepath. Our wasmtime test does that. I'm not personally getting into lldb test infrastructure as have huge workload on primary project, so its essentials only from me at this point.

lldb/source/Plugins/JITLoader/GDB/JITLoaderGDB.cpp

   if (jit_addr == LLDB_INVALID_ADDRESS)
     return;
 
-  m_jit_descriptor_addr = GetSymbolAddress(
-      module_list, ConstString("__jit_debug_descriptor"), eSymbolTypeData);
+  m_jit_descriptor_addr =
+      GetSymbolAddress(module_list, ConstString("__jit_debug_descriptor"),
+#ifdef _MSC_VER
+                       // Windows implementation not indicating symbol is Data
+                       // so we use Any as workaround for the time being.
+                       eSymbolTypeAny);
+#else
+                       eSymbolTypeData);
+#endif
+
   if (m_jit_descriptor_addr == LLDB_INVALID_ADDRESS) {
     LLDB_LOGF(log, "JITLoaderGDB::%s failed to find JIT descriptor address",
               __FUNCTION__);