Thread safety question on co/instanceCache.cpp

[hernando]

Is it thread safe to return a reference in:

const InstanceCache::Data& InstanceCache::operator[]( const UUID& id )

?

I don't see the lock being returned as part of InstanceCache::Data.

[hernando]

OK. I see that the semantics of release and operator[] guarantee that the data should not be deallocated while in use. Just a minor question then, does stde::hash_map guarantee that the iterators are not invalidated by insertions?

[eile]

Good question: I haven't paid attention. Can you look into it?

[hernando]

The MSVC version doesn't invalidate them:

Each element contains a separate key and a mapped value. The sequence is represented in a way that permits lookup, insertion, and removal of an arbitrary element with a number of operations that is independent of the number of elements in the sequence (constant time) -- at least in the best of cases. Moreover, inserting an element invalidates no iterators, and removing an element invalidates only those iterators which point at the removed element.

SGI's documentation on std::hash_map says that resizing doesn't invalidate either:

[3] Resizing does not invalidate iterators; however, it does not necessarily preserve the ordering relation between iterators...

If gcc agrees with the original STL specification (not the C++ Standard Library), it should be save.

[eile]

Thanks. Iirc the gcc STL is the SGI/HP STL.

[eile]

Can be closed?

[hernando]

I think so