chore: replace regression_5202 with more manageably sized program (#…

…5345)

# Description

## Problem\*

Resolves <!-- Link to GitHub Issue -->

## Summary\*

This PR replaces `regression_5202` with another program which expresses
the same issue but is much smaller and so easier to debug if we get a
regression in future.

Thanks to @nventuro for contributing this minimal reproduction.

## Additional Context



## Documentation\*

Check one:
- [x] No documentation needed.
- [ ] Documentation included in this PR.
- [ ] **[For Experimental Features]** Documentation to be submitted in a
separate PR.

# PR Checklist\*

- [x] I have tested the changes locally.
- [x] I have formatted the changes with [Prettier](https://prettier.io/)
and/or `cargo fmt` on default settings.

---------

Co-authored-by: jfecher <jake@aztecprotocol.com>

test_programs/execution_success/regression_5202/Nargo.toml → test_programs/execution_failure/regression_5202/Nargo.toml

@@ -2,7 +2,6 @@
 name = "regression_5202"
 type = "bin"
 authors = [""]
-compiler_version = ">=0.30.0"
+compiler_version = ">=0.29.0"
 
 [dependencies]
-fraction = { path = "fraction" }

test_programs/execution_failure/regression_5202/src/main.nr

@@ -0,0 +1,41 @@
+trait ToField {
+    fn to_field(self) -> Field;
+}
+
+impl ToField for bool { fn to_field(self) -> Field { self as Field } }
+
+unconstrained fn get_unconstrained_option() -> Option<u32> {
+    Option::some(13)
+}
+
+unconstrained fn should_i_assert() -> bool {
+    false
+}
+
+fn get_magical_boolean() -> bool {
+    let option = get_unconstrained_option();
+
+    let pre_assert = option.is_some().to_field();
+
+    if should_i_assert() {
+        // Note that `should_i_assert` is unconstrained, so Noir should not be able to infer
+        // any behavior from the contents of this block. In this case it is actually false, so the
+        // assertion below is not even executed (if it did it'd fail since the values are not equal).
+
+        assert_eq(option, Option::some(42)); /// <- this seems to be the trigger for the bug
+    }
+
+    // In my testing, the `option` value exhibits weird behavior from this point on, as if it had been mutated
+    let post_assert = option.is_some().to_field();
+
+    // The following expression should be true, but I can get it to evaluate to false depending on how I call it
+    pre_assert == post_assert
+}
+
+fn main() {
+    let magic = get_magical_boolean();
+
+    // One of these asserts should fail. Before #5202, they would both pass
+    assert_eq(magic, true);
+    assert_eq(magic, false);
+}