ir: Fix is_in_non_fully_specialized_template check.

[emilio]

Fixes #462

r? @upsuper

[upsuper]

I didn't write the patch because I'm not fully sure how this should be handled. I guess is_template_like is the right way, but I'm not confident enough on what does that mean. Do we have a reliably way to detect whether one is a template (rather than a template specialization)?

[upsuper]

Probably we should add a new method called is_template and check whether kind is one of CXCursor_FunctionTemplate, CXCursor_ClassTemplate, CXCursor_ClassTemplatePartialSpecialization, and CXCursor_TypeAliasTemplateDecl.

is_template_like sounds vague here as well... And I'm not sure whether the lack of function template there could be a problem here.

[upsuper]

I suppose we should still check recursively whether parent is_in_non_fully_specialized_template, otherwise something like

template<typename... T>
struct Outer {
  struct Inner {
    static constexpr int value[] = { T::value, ... };
  };
};

may crash again?

[emilio]

Indeed, great catch.

[emilio]

The lack of function templates doesn't matter at all because we don't inspect their bodies.

[emilio]

Indeed, great catch.

[upsuper]

Yeah, I agree lack of that doesn't matter here. I just feel is_template_like sounds vague... anyway, this looks fine.

[upsuper]

@bors-servo r+ (if I can)

[bors-servo]

@upsuper: 🔑 Insufficient privileges: Not in reviewers

[emilio]

@bors-servo r=upsuper

[bors-servo]

📌 Commit eb0a8af has been approved by upsuper

[emilio]

@bors-servo retry

• Travis is drunk lately

[emilio]

@bors-servo r=upsuper

[bors-servo]

📌 Commit 993a959 has been approved by upsuper

[bors-servo]

⌛ Testing commit 993a959 with merge 44119eb...

[bors-servo]

☀️ Test successful - status-travis
Approved by: upsuper
Pushing 44119eb to master...