[cling] Update policy for GetFullyQualifiedName in cling utils
#19292
Conversation
Just like in CppInterOp, set the `SuppressElaboration` and `FullyQualifiedName` policies when getting the fully qualified name of a QualType. Closes the following JIRA ticket: https://its.cern.ch/jira/browse/ROOT-10150
Test Results 20 files 20 suites 3d 4h 50m 49s ⏱️ For more details on these failures, see this check. Results for commit ea1957a. |
| EXPECT_EQ("fHash", hashSubfields[0]->GetFieldName()); | ||
| EXPECT_EQ("std::string", hashSubfields[0]->GetTypeName()); | ||
| EXPECT_EQ("EdmHash<1>::value_type", hashSubfields[0]->GetTypeAlias()); | ||
| EXPECT_EQ("EdmHash::value_type", hashSubfields[0]->GetTypeAlias()); |
There was a problem hiding this comment.
This change is incorrect: EdmHash is a template class and requires template arguments. Clang stores the inner class outside the template instantiation as an optimization, but the type must have arguments.
There was a problem hiding this comment.
Yep it's incorrect :( I realized the printing policies are quite brittle in general, and I can't make all cases work. I'll report on the problems I found soon, in a separate GitHub issue I guess.
| // FIXME: This should really be EdmHash<1>::value_typeT<EdmHash<1>::value_type>, but this is the value we get from | ||
| // TDataMember::GetFullTypeName right now... | ||
| EXPECT_EQ("value_typeT<EdmHash<1>::value_type>", hashSubfields[1]->GetTypeAlias()); | ||
| EXPECT_EQ("EdmHash::value_typeT<EdmHash::value_type>", hashSubfields[1]->GetTypeAlias()); |
There was a problem hiding this comment.
This is equally wrong, the expected outcome is even documented in the comment.
| const auto &fdesc = desc.GetFieldDescriptor(desc.FindFieldId("m", fooId)); | ||
| EXPECT_EQ(fdesc.GetTypeName(), "std::vector<std::int32_t>"); | ||
| EXPECT_EQ(fdesc.GetTypeAlias(), "MyVec<std::int32_t>"); | ||
| EXPECT_EQ(fdesc.GetTypeAlias(), "CustomStruct::MyVec<std::int32_t>"); |
There was a problem hiding this comment.
The old version may be correct, depending on how you argue: it's the type alias as the user wrote it, inside of CustomStruct which happens to have a typedef...
|
Related: #999 |
|
Superseeded by the following PR by @devajithvs, coincidentally opened later the same day, but independently: Dev should be the better expert to bring this to conclusion 🙂 |
3 participants
Just like in CppInterOp, set the
SuppressElaborationandFullyQualifiedNamepolicies when getting the fully qualified name of a QualType.Closes the following JIRA ticket:
https://its.cern.ch/jira/browse/ROOT-10150