fix: complex arguments to builtin functions

[tserg]

What I did

Fix #3164, and also ecadd and ecmul.

How I did it

Cache the argument.

How to verify it

See tests.

Commit message

prior to this commit, some builtin functions including ceil, would panic
if their arguments were function calls (or otherwise determined to be
complex expressions by `is_complex_ir`). this commit fixes the relevant
builtin functions by using `cache_when_complex` where appropriate.

Description for the changelog

Fix codegen for function calls as argument in builtin functions

Cute Animal Picture

[codecov-commenter]

Codecov Report

Merging #3167 (5323481) into master (f450cb1) will increase coverage by 0.27%.
The diff coverage is 92.59%.

❗ Your organization is not using the GitHub App Integration. As a result you may experience degraded service beginning May 15th. Please install the Github App Integration for your organization. Read more.

@@            Coverage Diff             @@
##           master    #3167      +/-   ##
==========================================
+ Coverage   89.05%   89.32%   +0.27%     
==========================================
  Files          86       84       -2     
  Lines       10792    10777      -15     
  Branches     2449     2459      +10     
==========================================
+ Hits         9611     9627      +16     
+ Misses        781      754      -27     
+ Partials      400      396       -4     

Impacted Files	Coverage Δ
vyper/builtins/functions.py	90.51% <92.59%> (+1.29%)	⬆️

... and 25 files with indirect coverage changes

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

[charles-cooper]

codegen looks good, but i don't think the test cases are testing the case properly. i think a better test would have (non-idempotent) side effects in the called contract so you can check if the side effects are getting invoked once or multiple times.

[charles-cooper]

also, since the test functionality is repeated (here, and also in other places but we don't need to worry about the other places just now), it could be good to have some scaffolding / utilities which make this easier. i'm thinking like, a common contract which is intended to be called for its side effects, and a utility function which checks that the side effect was invoked exactly one time.

[tserg]

I have added two fixtures side_effects_contract and assert_side_effect_invoked_once to standardise the test for side effects. What do you think?

[charles-cooper]

I have added two fixtures side_effects_contract and assert_side_effect_invoked_once to standardise the test for side effects. What do you think?

the fixtures look a bit weird to me, but i'm not sure there is a much cleaner way to do it

[tserg]

I am not sure why this call to foo() reverts if this contract was in the test suite. As long as there is more than one external call in uint256_addmod regardless of the argument position, the transaction reverts even if the gas is set to a high value. If tested in boa, foo() works fine.

@external
def foo(addr: address) -> uint256:
    f: Foo = Foo(addr)
    return uint256_addmod(f.a(), f.b(), f.c())

interface Foo:
    def a() -> uint256: payable
    def b() -> uint256: payable
    def c() -> uint256: payable

I will set the argument for only c to the external call (e.g. uint256_addmod(32, 2, f.c())) for the purpose of this PR.

[charles-cooper]

cleaned up the side effects checker API a little in 5323481