[clang-tidy] ProTypeMemberInitCheck - check that field decls do not have in-class initializer

[flx]

Summary

This fixes bug https://llvm.org/bugs/show_bug.cgi?id=26747

[JVApen]

I'm not sure if I'm the right person to do reviews, as I'm mostly a user of the tool and have not yet done any developments or bug fixes to it.
However, from what I know from coding, this looks good to me.

[aaron.ballman]

test/clang-tidy/cppcoreguidelines-pro-type-member-init.cpp

94 ↗

(On Diff #51123)

Why is this checking a fix? I thought that point was that this should not generate any diagnostic (and hence, no fix is required)?

[flx]

test/clang-tidy/cppcoreguidelines-pro-type-member-init.cpp

94 ↗

(On Diff #51123)

If I understood Alex correctly, CHECK-FIXES is needed to check for the absence of a change made by ClangTidy. But I agree it seems unlikely to have a change without a warning message which will trigger a test failure at any rate.

[aaron.ballman]

test/clang-tidy/cppcoreguidelines-pro-type-member-init.cpp

94 ↗

(On Diff #51123)

@alexfh, what are your thoughts on this? It seems like we should be able to test negative fixes, like with CHECK-NOT, except CHECK-FIXES-NOT or some such?

[alexfh]

Missed this patch somehow: it was in a wrong part of my dashboard =\

LG. Thanks for the fix.

test/clang-tidy/cppcoreguidelines-pro-type-member-init.cpp

94 ↗

(On Diff #51123)

I usually ask people to add CHECK-FIXES in case when there is a warning that could change some code, but shouldn't. When there's no warning that is likely to touch some code, there's no need for a CHECK-FIXES to verify this.

[alexfh]

Felix, do you have time to commit the patch or should I do this for you?

[flx]

I should free up again soon, just caught the flu though. Maybe next week?
Feel free to commit before then.

[alexfh]

In D18300#412003, @flx wrote:

I should free up again soon, just caught the flu though. Maybe next week?
Feel free to commit before then.

Just tried it, and the patch doesn't apply cleanly. Leaving this to you.

clang-tools-extra/trunk/clang-tidy/cppcoreguidelines/ProTypeMemberInitCheck.cpp

 namespace cppcoreguidelines {
 
 namespace {
 
 void fieldsRequiringInit(const RecordDecl::field_range &Fields,
                          ASTContext &Context,
                          SmallPtrSetImpl<const FieldDecl *> &FieldsToInit) {
   for (const FieldDecl *F : Fields) {
+    if (F->hasInClassInitializer())
+      continue;
     QualType Type = F->getType();
     if (!F->hasInClassInitializer() &&
         type_traits::isTriviallyDefaultConstructible(Type, Context))
       FieldsToInit.insert(F);
   }
 }
 
 void removeFieldsInitializedInBody(

clang-tools-extra/trunk/test/clang-tidy/cppcoreguidelines-pro-type-member-init.cpp

struct NegativeConstructorDelegated {
   NegativeConstructorDelegated() : NegativeConstructorDelegated(0) {}
 };
 
 struct NegativeInitializedInBody {
   NegativeInitializedInBody() { I = 0; }
   int I;
 };
 
+struct A {};
+template <class> class AA;
+template <class T> class NegativeTemplateConstructor {
+  NegativeTemplateConstructor(const AA<T> &, A) {}
+  bool Bool{false};
+  // CHECK-FIXES: bool Bool{false};
+};
+
 #define UNINITIALIZED_FIELD_IN_MACRO_BODY(FIELD) \
   struct UninitializedField##FIELD {             \
     UninitializedField##FIELD() {}               \
     int FIELD;                                   \
   };                                             \
 // Ensure FIELD is not initialized since fixes inside of macros are disabled.
 // CHECK-FIXES: int FIELD;