Weak references are objects too

[oldnewthing]

If the source contributes to the module object count, then the weak reference also contributes, so that we do not unload prematurely.

Factored out module lock updating into helper base class module_lock_updater with specialization update_module_lock. This avoids a lot of code duplication.

[oldnewthing]

Wondering why weak_source_producer was defined as

    template <bool Agile>
    struct weak_source_producer
    {
    protected:
        weak_source<Agile> m_source;
    };

instead of

    template <bool Agile>
    struct weak_source_producer : protected weak_source<Agile>
    {
    };

This alternate definition would greatly simplify weak_ref::that():

        weak_ref<Agile>* that() noexcept
        {
            return static_cast<weak_ref<Agile>*>(this);
        }

[kennykerr]

@oldnewthing no particular reason - feel free to fix.

[oldnewthing]

Oh, it's because weak_source is final because it wants to have a different QI from its parent object.

[kennykerr]

/azp run

[azure-pipelines]

No pipelines are associated with this pull request.

[kennykerr]

Thanks!