fix: folding of bitwise not literals

[tserg]

What I did

Fix #3218.

How I did it

Reuse (2 ** 256 - 1) - value from BitwiseNot instead of operator.inv, which can return a negative number.

How to verify it

See updated tests.

Commit message

fix: folding of bitwise not literals

Description for the changelog

Fix folding of bitwise not literals.

Cute Animal Picture

[codecov-commenter]

Codecov Report

Merging #3222 (3bce207) into master (ae7d6d2) will increase coverage by 16.14%.
The diff coverage is 100.00%.

@@             Coverage Diff             @@
##           master    #3222       +/-   ##
===========================================
+ Coverage   72.73%   88.87%   +16.14%     
===========================================
  Files          84       84               
  Lines       10580    10589        +9     
  Branches     2139     2208       +69     
===========================================
+ Hits         7695     9411     +1716     
+ Misses       2271      769     -1502     
+ Partials      614      409      -205     

Impacted Files	Coverage Δ
vyper/ast/nodes.py	93.78% <100.00%> (+14.68%)	⬆️
vyper/codegen/ir_node.py	91.03% <0.00%> (+1.72%)	⬆️
vyper/abi_types.py	83.33% <0.00%> (+1.78%)	⬆️
vyper/ast/signatures/function_signature.py	88.28% <0.00%> (+1.80%)	⬆️
vyper/codegen/arithmetic.py	84.71% <0.00%> (+2.02%)	⬆️
vyper/codegen/context.py	91.91% <0.00%> (+2.94%)	⬆️
vyper/builtins/_signatures.py	97.64% <0.00%> (+3.52%)	⬆️
vyper/semantics/types/primitives.py	92.53% <0.00%> (+3.53%)	⬆️
vyper/codegen/self_call.py	95.65% <0.00%> (+4.34%)	⬆️
vyper/codegen/module.py	97.70% <0.00%> (+5.74%)	⬆️
... and 45 more

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[tserg]

If 0.4.0 is the next release, will it be a good time to remove the set of bitwise builtin functions that have been deprecated since 0.3.4?

[charles-cooper]

If 0.4.0 is the next release, will it be a good time to remove the set of bitwise builtin functions that have been deprecated since 0.3.4?

let's worry about that later!

[charles-cooper]

this is only valid for uint256 btw

[charles-cooper]

it's going to be clearer imo as

Suggested change

	return (2 ** 256 - 1) - value
	return (2 ** 256 - 1) ^ value

the xor reflects what's going on with the bits. it's also going to generalize to other integer types a bit better. @tserg what do you think?

[tserg]

I agree, I started with a xor mask before changing to its current form to match what was already in vyper.utils.