Handling creation and destruction

[greggman]

IIUC trace/replay never destroys a resource. That seems problematic. I can imagine creating a temp 2gig buffer, then destroying it, then creating another, then destroying it, etc....... If replay never destroys resources it will run out of memory.

[Kangz]

The trace should be correctly destroying resources no? That's why there is the WeakMap too.

[greggman]

The replay does not.

[Kangz]

The replay needs to keep the resources alive otherwise it's not possible to go back. Do you have another suggestion?

[greggman]

The trace should arguably record creation and destruction that happens during the trace. So there's the pool of objects that existed before the trace, and then objects added and destroyed during the trace. At replay, create the initial pool then create/destroy the rest in playback.