Regression: clang 3.5 causes struct members to be defined again

[jacob-carlborg]

Regression: clang 3.5 causes struct members to be defined again

[mihails-strasuns-sociomantic]

Don't have old clang to see how it worked before but right now the problem is how libclang reports cursors:

struct foo { }; // CXCursor_StructDecl

struct foo var; // CXCursor_VarDecl

struct bar // CXCursor_StructDecl
{
    struct foo var; // CXCursor_StructDecl
}

I am looking for a way to distinguish that latter CXCursor_StructDecl is not really a struct declaration based on cursor metadata but nothing obvious seems to match.

[mihails-strasuns-sociomantic]

One option cold be to store list of all definition cursors (http://clang.llvm.org/doxygen/group__CINDEX__CURSOR__XREF.html#gafcfbec461e561bf13f1e8540bbbd655b) and check it before outputing a new one.

[mihails-strasuns-sociomantic]

#37

[jacob-carlborg]

@mihails-strasuns-sociomantic struct foo var is not reported as a CXCursor_FieldDecl when inside a struct?

[jacob-carlborg]

Don't have old clang to see how it worked before but right now

If you want to, you can run dub --config=test which will download a couple of versions of libclang and run the whole test suite. The actual test suite requires Ruby and two gems.

[mihails-strasuns-sociomantic]

struct foo var is not reported as a CXCursor_FieldDecl when inside a struct?

@jacob-carlborg it is, but is also reported as CXCursor_StructDecl - with a cursor referencing real definition. See linked PR.