[Mlir-commits] [mlir] [MLIR][LLVM] Fuse Scope into CallsiteLoc Callee (PR #74546)

[Tori Baker]

================
@@ -330,7 +330,10 @@ llvm::DILocation *DebugTranslation::translateLoc(Location loc,
   if (auto callLoc = dyn_cast<CallSiteLoc>(loc)) {
     // For callsites, the caller is fed as the inlinedAt for the callee.
     auto *callerLoc = translateLoc(callLoc.getCaller(), scope, inlinedAt);
-    llvmLoc = translateLoc(callLoc.getCallee(), scope, callerLoc);
+    llvmLoc = translateLoc(callLoc.getCallee(), nullptr, callerLoc);
----------------
vwbaker wrote:

Hi, I'm working with triton and I'm not sure I fully understand this change - is this scope->nullptr really intentional? It looks like in no place you are propagating the scope? We've found that while we previously were getting the correct inlined location propagated for our debugging, we no longer are - now it propagates the parent scope instead of the inlined function location (ex: https://github.com/openai/triton/blob/main/python/test/unit/language/test_line_info.py#L143, we expect it to correctly be able to return the inlined function location, line 22, but it's now returning line 30, where the inlined function is called). Is it no longer supported that we see the location within the inlined function or is this a bug?

https://github.com/llvm/llvm-project/pull/74546