[Coverage] Don't emit mappings for functions in dependent contexts (fixes PR32679)

[vsk]

Summary

The coverage implementation marks functions which won't be emitted as
'deferred', so that it can emit empty coverage regions for them later
(once their linkages are known).

Functions in dependent contexts are an exception: if there isn't a full
instantiation of a function, it shouldn't be marked 'deferred'. We've
been breaking that rule without much consequence because we just ended
up with useless, extra, empty coverage mappings. With PR32679, this
behavior finally caused a crash, because clang marked a partial template
specialization as 'deferred', causing the MS mangler to choke in its
delayed-template-parsing mode:

error: cannot mangle this template type parameter type yet
(http://bugs.llvm.org/show_bug.cgi?id=32679)

Fix this by checking if a decl's context is a dependent context before
marking it 'deferred'.

Based on a patch by Adam Folwarczny!

[adamf]

!MD->getParent()->getDescribedClassTemplate() looks like redundant now. This check is already done in isDependentContext.

[arphaman]

I agree with Adam, the old check is now redundant.

LGTM after removing it.

test/CoverageMapping/pr32679.cpp

2

Interesting, I didn't know about -implicit-check-not. Looks useful though.

lib/CodeGen/ModuleBuilder.cpp

void HandleInlineFunctionDefinition(FunctionDecl *D) override {
       //     void bar();
       //     void foo() { bar(); }
       //   } A;
       DeferredInlineMethodDefinitions.push_back(MD);
 
       // Provide some coverage mapping even for methods that aren't emitted.
       // Don't do this for templated classes though, as they may not be
       // instantiable.
-      if (!MD->getParent()->getDescribedClassTemplate())
+      if (!MD->getParent()->isDependentContext() &&
+          !MD->getParent()->getDescribedClassTemplate())
         Builder->AddDeferredUnusedCoverageMapping(MD);
     }
 
     /// HandleTagDeclDefinition - This callback is invoked each time a TagDecl
     /// to (e.g. struct, union, enum, class) is completed. This allows the
     /// client hack on the type, which can occur at any point in the file
     /// (because these can be defined in declspecs).
     void HandleTagDeclDefinition(TagDecl *D) override {

test/CoverageMapping/pr32679.cpp

+// RUN: %clang_cc1 -cc1 -triple i686-pc-windows-msvc19.0.0 -emit-obj -fprofile-instrument=clang -std=c++14 -fdelayed-template-parsing -fcoverage-mapping -dump-coverage-mapping -emit-llvm-only -main-file-name pr32679.cpp -o - %s | FileCheck %s -check-prefix=MSABI -implicit-check-not=f2
+// RUN: %clang_cc1 -cc1 -triple %itanium_abi_triple -emit-obj -fprofile-instrument=clang -std=c++14 -fcoverage-mapping -dump-coverage-mapping -emit-llvm-only -main-file-name pr32679.cpp -o - %s | FileCheck %s -check-prefix=ITANIUM -implicit-check-not=f2
+
+template <typename T, int S1>
+struct CreateSpecialization;
+
+template <typename T>
+struct CreateSpecialization<T, 0> {
+  static constexpr T f1() { return 0; }
+  static constexpr T f2() { return 0; }
+};
+
+int main() {
+  CreateSpecialization<int, 0>::f1();
+
+  // Don't emit coverage mapping info for functions in dependent contexts.
+  //
+  // E.g we never fully instantiate CreateSpecialization<T, 0>::f2(), so there
+  // should be no mapping for it.
+
+  return 0;
+}
+
+// MSABI: main:
+// MSABI-NEXT:   File 0, [[@LINE-12]]:12 -> [[@LINE-3]]:2 = #0
+// MSABI-NEXT: ?f1@?$CreateSpecialization@H$0A@@@SAHXZ:
+// MSABI-NEXT:   File 0, [[@LINE-18]]:27 -> [[@LINE-18]]:40 = #0
+
+// ITANIUM: main:
+// ITANIUM-NEXT:   File 0, [[@LINE-17]]:12 -> [[@LINE-8]]:2 = #0
+// ITANIUM-NEXT: _ZN20CreateSpecializationIiLi0EE2f1Ev:
+// ITANIUM-NEXT:   File 0, [[@LINE-23]]:27 -> [[@LINE-23]]:40 = #0