missing input validation for arith overflow

[bennetyee]

https://github.com/oasislabs/parity/blob/80238a2bc4af5026c3e4e55570de85e916a6223f/ethcore/wasm/src/env.rs#L231

u32.max_value() should be illegal here for max_memory due to arithmetic overflow later. Because <u32>::max_value() - 1 was used at the call site at lib.rs, it made it obvious that there are landmines here.

[willscott]

Is this still a problem that needs to be fixed, @nhynes?

[nhynes]

Yeah the - 1 spurious. It's benign, though. Of course, the whole thing will abort if the host machine ooms, but that should be handled by the scheduler. How does the scheduler even know how much memory an invocation will use?