Fix/msg data

[skellet0r]

What I did

Refactored the msg.data environment variable. Instead of handling each distinctive use case in expr.py, uses are handled in the respective functions which operate on msg.data.

• Len builtin simply returns calldatasize
• Slice builtin returns a bytes array containing the sliced portion of calldata
  • The length argument to slice should be a compile time literal as calldatasize is unbounded, and we need a maxsize
    on the return type of the slice function instead of defaulting to the maxsize of msg.data

How I did it

Followed charles' guidance. Mainly just made msg.data return an empty LLLnode with location='calldata', and then handled it in the builtin functions

How to verify it

The test suite for msg.data is unchanged, so passing tests verify this refactoring worked.

Description for the changelog

• Fix: msg.data memory allocation removed in favor of builtin function handling

Cute Animal Picture

[codecov-commenter]

Codecov Report

Merging #2419 (531296e) into master (31dd776) will increase coverage by 0.89%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #2419      +/-   ##
==========================================
+ Coverage   84.61%   85.51%   +0.89%     
==========================================
  Files          91       91              
  Lines        9034     9027       -7     
  Branches     2151     2151              
==========================================
+ Hits         7644     7719      +75     
+ Misses        882      804      -78     
+ Partials      508      504       -4     

Impacted Files	Coverage Δ
vyper/builtin_functions/functions.py	88.82% <100.00%> (+5.48%)	⬆️
vyper/old_codegen/expr.py	80.11% <100.00%> (+4.12%)	⬆️
vyper/semantics/validation/annotation.py	95.52% <100.00%> (+0.55%)	⬆️
vyper/semantics/validation/local.py	89.92% <100.00%> (-0.08%)	⬇️
vyper/ast/nodes.py	93.58% <0.00%> (+1.46%)	⬆️
vyper/semantics/types/value/numeric.py	86.02% <0.00%> (+5.37%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 31dd776...531296e. Read the comment docs.

[charles-cooper]

Overall looks great, I suggested a couple style/defensive changes.

As an aside, it would be really nice if we could get rid of the copy altogether, and just pass around a pointer to calldata. But that requires some rethinking of how we handle bytestrings (i.e. handling the length section separately from the data section) in general, and at that point we could really revamp slice -- including the case where the bytestring is in memory.

[charles-cooper]

Looks good. Thanks