rebis-dev: MachineState::new calling Heap::store_resource_error makes tests non-deterministic

[adri326]

Should Heap::store_resource_error always store the value? Right now, if I run tests with hard-coded offsets within the heap in isolation, they tend to break, which isn't great and defeats the purpose of MockWAM.

[triska]

As I understand it, a resource error term is stored on the heap in advance, so that a resource error can be thrown even if no more memory can be allocated. So, regarding the question, the answer seems yes, also because a test may itself run out of memory (for instance, try running a test with small ulimit) and is then expected to throw a clean resource error instead of crashing or other unexpected behaviour.

Maybe the tests can be adapted to use offsets relative to the first cell (index) that is available on the heap when the test starts?