[NFC] Remove dead conditionals

[schittir]

Summary

Patch contains changes to remove obvious dead code.

[aaron.ballman]

The other two changes LGTM, but the objc changes bring up a question about what the correct fix is.

clang/lib/CodeGen/CGObjCMac.cpp

3812

I agree this is dead code (see line 3777-3778 above), but @rjmccall is it a bug that we bail out with a null value (e.g, is the correct fix to remove the above lines)?

[rjmccall]

clang/lib/CodeGen/CGObjCMac.cpp

3812

No, I don't think this is a bug, and that would not be the right fix anyway. ForClass means we're building the metaclass, and subclasses do not introduce new ivars in the metaclass. The comment is saying that GCC declares ivars for the class structure when emitting metaclasses for root classes, but if we haven't been doing that for 15 years, I don't see a reason to start. Certainly the runtime doesn't require it in order to work. So the patch as written looks good.

For what it's worth, this code is all but dead for more basic reasons. CGObjCMac is the legacy ObjC ABI, which is no longer used on any platform that's even vaguely supported by Apple. The last platform we shipped using this ABI was i386 macOS; we haven't sold Macs that only supported i386 since 2006, and we dropped the i386 userspace entirely in macOS Catalina (2019). So the only reason for this code to exist is to allow people to either compile programs for very old hardware or intentionally use a 32-bit ABI on newer hardware running an old OS.

[aaron.ballman]

clang/lib/CodeGen/CGObjCMac.cpp

3812

Should we consider removing this code at this point? Folks who need to compile programs for very old hardware or have esoteric ABI needs can stick with an older release of Clang (not ideal, but I don't think it makes much sense to continue to carry around unsupported targets).

If so, we could add a release note/diagnostic for Clang 17 and then rip the code out in Clang 18 unless users have compelling use cases.

[schittir]

Fix clang-format issue in SemaOverload.cpp

[schittir]

Thank you for the review, @aaron.ballman and @rjmccall
The latest patch fixes clang-format issue causing build failure. Could you please take a look at this again? Thanks!

[aaron.ballman]

LGTM!

[schittir]

Patch landed with commit id 0e444c57c85853c86e4f4c94d9a51fb459fd9922

clang/lib/CodeGen/CGObjCMac.cpp

if (count == 0) {
     ivarList.abandon();
     return llvm::Constant::getNullValue(ObjCTypes.IvarListPtrTy);
   }
 
   ivars.finishAndAddTo(ivarList);
   ivarList.fillPlaceholderWithInt(countSlot, ObjCTypes.IntTy, count);
 
   llvm::GlobalVariable *GV;
-  if (ForClass)
-    GV =
-        CreateMetadataVar("OBJC_CLASS_VARIABLES_" + ID->getName(), ivarList,
-                          "__OBJC,__class_vars,regular,no_dead_strip",
-                          CGM.getPointerAlign(), true);
-  else
-    GV = CreateMetadataVar("OBJC_INSTANCE_VARIABLES_" + ID->getName(), ivarList,
-                           "__OBJC,__instance_vars,regular,no_dead_strip",
-                           CGM.getPointerAlign(), true);
   return llvm::ConstantExpr::getBitCast(GV, ObjCTypes.IvarListPtrTy);
 }
 
 /// Build a struct objc_method_description constant for the given method.
 ///
 /// struct objc_method_description {
 ///   SEL method_name;
 ///   char *method_types;

clang/lib/Sema/SemaOverload.cpp

if (Conv.Bad.Kind == BadConversionSequence::lvalue_ref_to_rvalue ||
     return;
   }
 
   // Special diagnostic for failure to convert an initializer list, since
   // telling the user that it has type void is not useful.
   if (FromExpr && isa<InitListExpr>(FromExpr)) {
     S.Diag(Fn->getLocation(), diag::note_ovl_candidate_bad_list_argument)
         << (unsigned)FnKindPair.first << (unsigned)FnKindPair.second << FnDesc
-        << (FromExpr ? FromExpr->getSourceRange() : SourceRange()) << FromTy
-        << ToTy << (unsigned)isObjectArgument << I + 1
+        << FromExpr->getSourceRange() << FromTy << ToTy
+        << (unsigned)isObjectArgument << I + 1
         << (Conv.Bad.Kind == BadConversionSequence::too_few_initializers ? 1
             : Conv.Bad.Kind == BadConversionSequence::too_many_initializers
                 ? 2
                 : 0);
     MaybeEmitInheritedConstructorNote(S, Cand->FoundDecl);
     return;
   }
 

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

CanonicalLoopInfo *OpenMPIRBuilder::createCanonicalLoop(
 
   Value *CountIfLooping;
   if (InclusiveStop) {
     CountIfLooping = Builder.CreateAdd(Builder.CreateUDiv(Span, Incr), One);
   } else {
     // Avoid incrementing past stop since it could overflow.
     Value *CountIfTwo = Builder.CreateAdd(
         Builder.CreateUDiv(Builder.CreateSub(Span, One), Incr), One);
-    Value *OneCmp = Builder.CreateICmp(
-        InclusiveStop ? CmpInst::ICMP_ULT : CmpInst::ICMP_ULE, Span, Incr);
+    Value *OneCmp = Builder.CreateICmp(CmpInst::ICMP_ULE, Span, Incr);
     CountIfLooping = Builder.CreateSelect(OneCmp, One, CountIfTwo);
   }
   Value *TripCount = Builder.CreateSelect(ZeroCmp, Zero, CountIfLooping,
                                           "omp_" + Name + ".tripcount");
 
   auto BodyGen = [=](InsertPointTy CodeGenIP, Value *IV) {
     Builder.restoreIP(CodeGenIP);
     Value *Span = Builder.CreateMul(IV, Step);