search

aionnetwork / AVM

Enabling Java code to run in a blockchain environment
https://theoan.com/
MIT License
49 stars 25 forks source link

Specify that hashCodes are unique, Fixes #404 #405

Closed fulldecent closed 5 years ago

jeff-aion commented 5 years ago

This isn't really correct, though. The only reason why these are unique for Enum instances is that they are all allocated as constants, which are assigned at deployment time such that a duplication would require an impossibly expensive overflow. Also, the hashCode() can (and should) be over-ridden by user-defined classes and various other types such that this statement wouldn't hold.

This is probably the kind of thing which should be documented elsewhere, as it is more of a consequence of how/when Enum constants are initialized, as opposed to anything else.

fulldecent commented 5 years ago

How about this construction?

Enum types (i.e. the constants inside an enum declaration) are static instances and Aion assigns static instances unique values for hashCode. This depends on a contract not having more than Integer.MAX_VALUE, which is already impossible due to energy limits. Java defines Enum.hashCode as a final method and therefore it cannot be overridden. (Source https://docs.oracle.com/javase/6/docs/api/java/lang/Enum.html.) Therefore the hashCode for each Enum type is globally unique in a specific contract deployment.

Also I'm not sure if that is the specific reference for Java since than is an API and not the language specification.

How do I update this construction so that a company with $1B of NFT assets at risk will be sufficiently convinced that the statement in bold is true?

jeff-aion commented 5 years ago

The problem is that this isn't true of hashCode(), in general, just the hashCode() among AVM Enum types. Also, the user is more likely to be looking at the normal JDK documentation than the AVM-specific variant. These 2 points combined lead me to think that this should be documented elsewhere and more specifically to Enums.

fulldecent commented 5 years ago

We agree that the location the PR puts it is wrong.

Right now I’m trying to develop the statement just relating to Enums. It seems we agree the statement is true. Please confirm explicitly if this is the case or propose an alternative.

Next we can find a good place to put the statement.

jeff-aion commented 5 years ago

The "Consequences" section of this page describes what you are looking for: https://github.com/aionnetwork/AVM/wiki/Hash-Code

fulldecent commented 5 years ago

Thank you, that resolves the issue.

Used at : https://github.com/fulldecent/aion-aip040/commit/635d84363e9a4ef46bcb1ccfd3ce3bb8505b3d28