fix(semantic): add salsa cycle recovery to the item generic-params queries

[orizi]

Summary

Adds cycle_fn and cycle_initial handlers to all *_generic_params_data salsa queries (enum, struct, free_function, extern_function, impl_def, module_type_alias, trait_function). The shared handlers (generic_params_data_cycle, generic_params_data_initial) are defined in generics.rs and return the freshly computed value on cycle recovery rather than freezing at the initial provisional value, allowing salsa to fixpoint-iterate to the correct result. Also fixes module_type_alias_semantic_data_cycle to short-circuit with skip_diagnostic() when in_cycle is true, preventing infinite re-entry.

---

Type of change

Please check one:

• Bug fix (fixes incorrect behavior)
• New feature
• Performance improvement
• Documentation change with concrete technical impact
• Style, wording, formatting, or typo-only change

---

Why is this change needed?

When an item-level macro call uses a path that names an enum variant (e.g. E::A!(());), the compiler resolves the path through macro_call_module_id / priv_macro_call_data, which reads the enum's variants, which resolves the variant's type. If that type references a generic item with a trait bound (e.g. W<felt252> where W<T, +Drop<T>>), resolving it re-enters the *_generic_params_data query for that item while it is still on the stack. Without cycle handlers, salsa would panic with an ICE (internal compiler error) instead of recovering gracefully.

---

What was the behavior or documentation before?

The compiler would ICE (panic) when encountering item-level macro calls whose path resolved through an enum variant whose type referenced a generic item with a trait bound, because the resulting query cycle had no registered handler.

---

What is the behavior or documentation after?

The compiler recovers from the cycle, emits a E2028: Cycle detected while resolving generic param diagnostic (and E2156: Inline macro not found for the unresolvable macro), and continues without panicking. Golden test files are added for enum, struct, free_function, extern_function, impl_def, trait, and type_alias to cover each affected query.

---

Related issue or discussion (if any)

---

Additional context

Convergence is bounded: the fixpoint settles within three iterations on the cycle-triggering inputs covered by the golden tests, well within salsa's MAX_ITERATIONS ceiling of 200.

[reviewable-StarkWare]

This change is 

[orizi]

• fix(semantic): add salsa cycle recovery to the item generic-params queries #10272 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[cursor]

PR Summary

Medium Risk
Touches core semantic query cycling for many item kinds; incorrect cycle recovery could hide or mis-report diagnostics, but behavior is bounded by fixpoint iteration and covered by new golden tests.

Overview
Adds Salsa cycle recovery for all *_generic_params_data queries (enums, structs, free/extern functions, impl defs, module type aliases, trait functions) so item-level macro paths that re-enter generic-param resolution no longer ICE.

Shared handlers in generics.rs — generic_params_data_initial (provisional Err(skip_diagnostic())) and generic_params_data_cycle (return the freshly computed value for fixpoint iteration) — are wired via cycle_fn / cycle_initial on each tracked query. That preserves real diagnostics (e.g. E2028) instead of freezing at the initial provisional result.

Module type alias cycle handling now short-circuits with skip_diagnostic() when in_cycle is true to avoid infinite re-entry through module_type_alias_generic_params_data. If module_type_alias_semantic_data fails during recovery, diagnostics fall back to the generic-params query so cycle-related errors still surface.

Golden tests cover the macro → enum variant → generic item with trait bound pattern across the affected item kinds.

^{Reviewed by Cursor Bugbot for commit 7a5f5b7. Bugbot is set up for automated code reviews on this repo. Configure here.}

[eytan-starkware]

@eytan-starkware reviewed 16 files and all commit messages, and made 2 comments.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on orizi and TomerStarkware).

---

crates/cairo-lang-semantic/src/items/generics.rs line 295 at r1 (raw file):

/// silently drop them.
///
/// Convergence is bounded and observed: the fixpoint settles within three iterations on the

Why 3 iterations?
I believe this comment can be less prose and more concise if it had some structure x -> y -> z

[orizi]

Re: generics.rs line 295 — "Why 3 iterations? I believe this comment can be less prose and more concise if it had some structure x -> y -> z"

Why 3: iteration i recomputes the cycle head with the re-entrant read seeing iteration i-1's provisional value, and the computed result only depends on whether that read saw the initial Err or a real value. So the sequence is: initial Err -> value computed against that Err -> value computed against a real value (changes once) -> unchanged (fixpoint). Salsa needs the last iteration to observe equality, hence 3. The number was also measured directly via Cycle::iteration() on the golden tests.

Restructured the whole doc comment into arrow chains (both the cycle path and the convergence argument) and pushed it to claude/issue-10272-response-sa2zw4 (9db7950) for cherry-picking into this PR.

---

Generated by Claude Code

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 9db7950. Configure here.}

[eytan-starkware]

@eytan-starkware reviewed 1 file and all commit messages, made 1 comment, and resolved 1 discussion.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on orizi and TomerStarkware).

[TomerStarkware]

@TomerStarkware reviewed 2 files and all commit messages, and made 1 comment.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on eytan-starkware).

[orizi]

@orizi resolved 1 discussion.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on orizi).