ByteBuffer cleanup

[mdavidsaver]

Check for allocation failure in ctors and add const qualifier where possible. Also document that externally provided buffers will be passed to free() (so can't use new[]).

[mdavidsaver]

oops, push the wrong branch

[dhickin]

1. When constructing a ByteBuffer around an existing buffer it's not true that the buffer is "free'd with free()". The wrapping constructor sets _wrapped to true and free is then not called in the destructor. Presumably whoever created the buffer is responsible for deleting it using the correct method.
2. Is std::bad_alloc the right exception to throw in the wrapping constructor? No allocation has been attempted. If it's not valid to call this constructor with a null pointer it seems reasonable to throw an exception, but std::invalid_argument might be a more correct choice.

[mdavidsaver]

1. This was a stupid mistake on my part. reverted.
2. Constructing invalid_argument involves memory allocation, bad_alloc doesn't. So if an actual low memory situation exists, bad_alloc will be thrown. So the caller must be prepared for bad_alloc anyway (as most all c++ code should). That said, I don't feel that strongly about this. If it's a blocker in your mind I'll change it.

[dhickin]

My preference is for std::invalid_argument. It's logically more correct, as no allocation might have been performed. If the buffer is allocated with new[] or some similar mechanism it will (usually, unless exceptions are disabled) throw the bad_alloc.If you allocate with malloc, I would say that the client code should check the pointer and throw the bad_alloc. The case when a null pointer is used (e.g. if it's initialised to null and never assigned to with the result of a allocation) will then return invalid_argument .

[mdavidsaver]

Changed to invalid_argument. Unless there are further objections I'll merge this on Monday.