There have been various projects where I've wanted to be able to run arbitrary Ethereum code within a smart contract, especially when dealing with interesting techniques, such as Wisps and a variety of Proxy contracts.
The idea of writing an EVM in EVM started as a joke, but then useful use cases started cropping up, and it's been stirring around in my head every once and a while, but without a good excuse to build it... So, perfect hackathon fodder. :)
What it does
calldata, which it then executes, returning the computed result.
How I built it
100% Pure Assembly. It's ridiculous. It's frustrating. It's painful. But oddly rewarding and fun.
No human should be writing assembly.
Challenges I ran into
Another reason I wanted to do this now, was I recently wrote an EVM assembler, and wanted to build a large-is scale project in it to test it out.
The main challenge was the number of non-trivial bugs in the actual assembler. The assembler is much better as a result of this project, if nothing else comes of it. :)
Accomplishments that I'm proud of
It works! All opcodes do their thing and it can, on-chain, execute contracts which are not on-chain.
What I learned
There is always those random little things you learn when you suffer so much. I learned how much I miss having useful debugging tools, like
console.log. Also that writing in assembly is always more time consuming than you think it will be, even if you think it will be really time-consuming.
What's next for Lurch
There is a list of todos, for sure.
- I want a generic hook contract, which will also be evaluated through Lurch (so recursion is needed). This will make it very simple to do some very weird, cool and potentially dangerous things
- Optimize the jump table. Hackathon, time-constraints, yadda yadda. Basically, by using a better multi-tier jump table, a lot of the instructions (especially
PUSH) will have their additional overhead reduced by at least a
- Incorporate it with the Long-Lived Wisp project