cl newPointer: rm types.Signature ptr

[xushiwei]

No description provided.

[fennoai]

Review: function proto / function pointer type support

The change is correct and well-targeted. It adds a top-level case lc.TypeFunctionProto in toType (handles a decayed bare function-type parameter such as void f(int (callback)(void))) and removes the *types.Signature panic branch in newPointer so a pointer-to-function-pointer lowers to *func().

Traced against the fixtures, all three shapes are handled consistently:

• int (*cmp)(...) → bare func(...) c.Int (single pointer elided at toType:70-72)
• int (callback)(void) → bare func() c.Int (new TypeFunctionProto case)
• void (**callbackPtr)(void) → *func() (recurses; inner func() wrapped by newPointer, which now falls through instead of panicking)

The _testmockc/type in.h/out.go fixtures match this behavior and provide adequate coverage.

Findings: No correctness, performance, or security issues. One optional maintainability note inline.

Note: go test ./cl -run TestMockC could not be executed in the review sandbox (missing llgo/libclang linkage — relocation target C.clang_createIndex not defined); fixture correctness was verified by manual trace-through rather than a live test run.

Additional findings

• cl/type_and_var.go:72: [P3] Consider a comment on the intentional pointer-elision asymmetry: A single-level function pointer (int (*cmp)(...)) is intentionally lowered to a bare func(...) here, while a double pointer (void (**)(void)) becomes *func(). This one-level elision is the central behavior of the PR but is not documented in code. A brief comment on this branch would help future maintainers avoid "fixing" it into *func(...). Optional / non-blocking.