Created attachment 293449 [details]
test case - nested initializers

This is related to bug #688647 - I first reported it there ( https://bugzilla.gnome.org/show_bug.cgi?id=688647#c5 ), but it's best to open a new one...

I'm using doxygen 1.8.8 on Arch Linux, so bug #688647 is supposed to be fixed, but I'm experiencing the same thing as the one reported by first post. I'll attach a test-case (one header from a project and Doxyfile) that demonstrates this.

Shorter version:

--->8--->8--->8--->8--->8--->8---

template<typename T>
class FifoQueue
{
public:

    ...

    /**
     * \brief FifoQueue's constructor
     */

    FifoQueue(Storage* const storage, const size_t maxElements) :
            fifoQueueBase_{storage, maxElements,
scheduler::FifoQueueBase::TypeTag<T>{}}
    {

    }

    /**
     * \brief Pops the oldest (first) element from the queue.
     */

    int pop(T& value)
    {
        return fifoQueueBase_.pop(value);
    }

    ...

--->8--->8--->8--->8--->8--->8---

No documentation is generated after the constructor. If I replace this:

fifoQueueBase_{storage, maxElements, scheduler::FifoQueueBase::TypeTag<T>{}}

with this:

fifoQueueBase_(storage, maxElements, scheduler::FifoQueueBase::TypeTag<T>{})

then everything is fine.

If I replace the same line with this:

fifoQueueBase_{storage, maxElements}

or this:

fifoQueueBase_{storage, maxElements, scheduler::FifoQueueBase::TypeTag<T>}

it also works.

This is not related to bug #721462 - if I put anything in the TypeTag
initializer, it still doesn't work.

So it seems that doxygen handles the new initializer syntax of C++11 only when
there is no nesting, but doesn't like the idea of one initializer embedded in
another initializer.