Adding label support to assembler

[lialan]

I would like to use this little tool to do some easy assembling in my project but found that this does not support label syntax, such as:

    ...
some_label:
    JUMPDEST
    ...
    PUSH some_label
    JUMP

So I just added the labeling syntax.

Another thing is that it seems that EVM memory space address starts with 1 rather than 0. So I changed pc according.

[CLAassistant]

All committers have signed the CLA.

[feliam]

alternative instr.semantics == "PUSH"

Also if you really need an is_push make one like is_branch (you can use opcodes instead of semantics if you want)

pyevmasm/pyevmasm/evmasm.py

Lines 305 to 308 in 0933d39

	@property
	def is_branch(self):
	""" True if the instruction is a jump """
	return self.semantics in {'JUMP', 'JUMPI'}

[feliam]

Being it so simple I vote to do this inline so we do not need to document, maintain, etc and we save 1 func call.

[feliam]

line.statrswith ?

[feliam]

no need for keys(). label not in fillins is enough I think, no?

[feliam]

Here you can check if the instruction has any operan referring to a label and fix them.

Add an Instruction.uses_labels() that encapsulates that?

[feliam]

Idea:

Not sure we need the fillings dict?
Maybe just write whatever there is to the instruction.operand. An int or label(string).
And then consider the instructions with a string operand to need fixup. ?

Instruction will not be able to generate a bytecode until it is "fixedup"

[feliam]

Instead of iterating over al linstructions for each label that qualifies maybe just rework all this do a single iteration over the instructions and fix the ones that needs it ? The instructions with Instruction.uses_labels() true

[feliam]

Hey @lialan thank you for this. The idea of having labels and other more high level constructs is very interesting. I was secretly expecting for someone to use this in some compiler-ish project.

I gave it a first pass.
We may be on the verge of needing a Contract class that contains the more high level stuff. (labels, directives)

I hope we do not loss the filter like structure of the disassembler/assembler. cat bytecode | disassemble | assemble
We'll see.

[lialan]

hi @feliam, thanks for the reply (and suggestions!), I will update accordingly to your comments later this week.

so I basically am using this assembler in my LLVM project for EVM backend, and I need an easy way to generate some LLVM IR for EVM target, such as https://github.com/etclabscore/evm_llvm/blob/EVM/test/CodeGen/EVM/runtime_tests/fib.ll

Because I have not figured out how to represent contract creation code in LLVM code so here is what I am doing: using a contract template, instantiate it with with generated EVM assembly, then emit a real contract assembly, then use this assembler to emit EVM binary for testing on Geth.

Here is how I integrate it into my testing workflow:
https://github.com/etclabscore/evm_llvm/blob/EVM/tools/evm-test/evm_test.py

Yes, this is a real use case of your code, kudo for that, and thanks for this little tool! I will try to see if I can figure out your comments and make proper changes to this pull request.