Opened IOStream is never closed for COLLADA files about assimp HOT 5 CLOSED

The IOStream can either be deleted explicitly, or closed via IOHandler::Close; in theory, they should accomplish the same thing. However, that seems overly ambiguous, and places an unnecessary burden on implementers of IOStream/IOHandler to properly handle both use-cases.

What if instead, we always returned a scoped_ptr from IOHandler::Open, removed the IOHandler::Close method completely, and made the IOStream destructor solely responsible for cleanup. Thoughts? This enforces consistent usage for all clients of IOHandler.

from assimp.

A std::unique_ptr would probably be a solution, unfortunately we cannot switch to C++11 for any forseeable future. A scoped_ptr is not intended to be returned at all (and without move semantics it makes no sense either).

Personally, I would just drop the Close() and let the destructor do it.

from assimp.

Yeah, that sounds reasonable.

from assimp.

I actually call Close on the destructor. So I also agree Close can go. But the issue remain because the IOStream is never destructed.

I currently worked around it by keeping an instance of the IOStream I create in IOSystem and destroy it when IOSystem is destroyed or before opening a new one...

it's probably wrong for some use case but works for mine because I have to load only one IOStream at the time.

from assimp.

Ok, I take it partially back.

Close is never called.

But the destructor of the IOStream is

from assimp.