I've tried replicating the deep-matching by saying qualType(hasType(hasUnqualifiedDesugaredType(hasDeclaration(... but this doesn't work because hasDeclaration only returns a matcher for a specific type from a subset of subclasses of Type - this is incompatible with expectations of the proposed hasUnqualifiedDesugaredType which takes a Matcher<Type>.

Do we also need to update the documentation (e.g. to say that ElaboratedType is covered by hasDeclaration)?

Forcing shallow matching means that unit test below will stop passing after this CL.

FWIW, a non-owner LGTM:

• CXXNewExpr seems very similar to CallExpr, so it makes sense that hasDeclaration would behave similarily for both of these expressions (i.e. matching the "callee")
• The issues we've been trying to work through in https://reviews.llvm.org/D24361 mainly revolve around Type and QualType, so I think those issues should not apply to CXXNewExpr matching.

Reverted changes in the patch that are not related to the issue of hasDeclaration not matching *anything* in some cases.

Richard - what are the next steps for this patch?

• Added test where both TemplateSpecializationType and TypedefType are present and both should match regardless of code order inside HasDeclarationMatcher::matchesSpecialized(const QualType &Node...). Removed significance of order inside this matchesSpecialized method.

Richard, could you please take a look?

I am assuming that the expectation I expressed in the unit test is a valid/correct expectation. I guess this is the first thing to validate/ack in the review :-)

In D24268#535463, @rsmith wrote:

This patch looks great, thank you! Do you have an SVN account or do you need someone to commit this for you?

Addressing CR feedbackfrom rsmith@.