build(codegen): main fails `cargo check --all-targets` — two ImportedClass test initializers miss constructor_has_synthetic_arguments

[proggeramlug]

What happened

cargo check --workspace --all-targets does not compile on origin/main (6092204, merge train 218 / v0.5.1596), before any branch is applied. Lint gate 83, "warnings: rustc warnings (host-compatible, all targets)", fails for every branch that validates against main.

error[E0063]: missing field `constructor_has_synthetic_arguments` in initializer of `ImportedClass`
  --> crates/perry-codegen/src/expr/instanceof_imported_rhs_tests.rs:46:5
error[E0063]: missing field `constructor_has_synthetic_arguments` in initializer of `ImportedClass`
   --> crates/perry-codegen/src/lower_call/new_builtin_shadow_tests.rs:118:32
error: could not compile `perry-codegen` (lib test) due to 2 previous errors

Cause

016e8ecf9 ("fix(hir,codegen,runtime): make arguments in class constructors reflect the call site") added pub constructor_has_synthetic_arguments: bool to ImportedClass (crates/perry-codegen/src/codegen/opts.rs:576).

It updated two of the four test initializers of that struct:

• crates/perry-codegen/src/expr/readonly_collection_tests.rs:185 — constructor_has_synthetic_arguments: false,
• crates/perry-codegen/src/lower_call/typed_shape_bake_tests.rs:495 — constructor_has_synthetic_arguments: false,

and missed the two named in the errors.

Suggested fix

Add the same line to both remaining initializers, matching what that commit already did for its siblings:

constructor_has_synthetic_arguments: false,

Why it got through the train

It only appears under --all-targets. cargo build, cargo build --release and cargo test -p perry-runtime --lib are all clean on main — I ran all three — so nothing in an ordinary build or a targeted test run surfaces it. scripts/run_lint_gates.sh catches it, and validate-017.sh reports it as one of its 83 gates.

Not mine

Found while validating an unrelated regex branch whose entire diff is 19 lines in crates/perry-runtime/src/regex/perex_replace_storage.rs, which cannot reach perry-codegen. Reproduced on pristine origin/main at the same commit with no changes applied.

Separately, and longer-standing: the gate "Public benchmark evidence freshness" (python3 benchmarks/ci_public_baseline_check.py) also fails on pristine main — public artifact benchmark inputs changed; regenerate it with ./benchmarks/run_public_baseline.sh. That one has been red for at least a day and is a different problem; mentioning it so the two are not confused when someone runs the gates and sees 2 of 83 failing.

[proggeramlug]

Already fixed on main by #10682 ("fix(test): unbreak the perry-codegen lib-test build on main (-D warnings --all-targets)"), which is merged.

Verified by content on origin/main @ 4715bc2fa — both initializers named in the error now set the field:

crates/perry-codegen/src/expr/instanceof_imported_rhs_tests.rs    constructor_has_synthetic_arguments present
crates/perry-codegen/src/lower_call/new_builtin_shadow_tests.rs   constructor_has_synthetic_arguments present
changelog.d/10682-codegen-testbuild-missing-field.md              present

I came at this from the other direction: an agent of mine independently hit the same broken build while working on #10689, wrote the identical two-line fix, and I was about to open a PR for it. Re-checking main first is what caught that #10682 had already landed it — so there is nothing to push.

Scope check, since I am closing someone else's issue: this issue names exactly those two initializers as its cause and suggested fix, and both are now fixed, so its content is fully addressed. What I have not done is re-run cargo check --workspace --all-targets on current main to prove the whole gate is green — that is a long workspace build and the box is busy. If lint gate 83 is still failing for a different missing initializer, that is a new instance rather than this one, and reopening is the right call.

Closing as fixed by #10682.