fix(parsers): extract C/C++ functions returning a pointer or reference - #1582

Fixes #1524. Full suite: 1139 passed, 7 skipped, no failures.

The bug

Both queries nested pointer_declarator inside function_declarator. The real AST is the reverse — a pointer return type wraps the function declarator from the outside:

char* f() { }
function_definition > pointer_declarator > function_declarator > identifier

Verified against the grammar:

'char* data() { ... }'      -> pointer_declarator > function_declarator > identifier
'int size() { ... }'        -> function_declarator > identifier
'char** dd() { ... }'       -> pointer_declarator > pointer_declarator > function_declarator > identifier
'int& ref() { ... }'        -> reference_declarator > function_declarator
'char* Buf::at() { ... }'   -> pointer_declarator > function_declarator > qualified_identifier

So c.py's second alternative matched nothing, and cpp.py had no alternative at all.

The half that mattered more

pre_scan_c and pre_scan_cpp carry their own copies of the same inverted shape. That is the cross-file name map, so these functions weren't just missing as nodes — they were missing as call targets, and every call to them stayed unresolved. Both copies are fixed here.

C   pre_scan : ['dup_str', 'grid', 'plain']            # was ['plain']
C++ pre_scan : ['Buf', 'Buf::data', 'Buf::size', 'data',
                'make', 'plain', 'ref', 'size']

Accessors returning T*/T&, factory functions and operator[] are pervasive in C and C++, so this was a large silent gap.

Covered forms

T*, T**, T& (C++ only), and the qualified out-of-line method char* Buf::data() — whose class_context survives the extra wrap, which is asserted separately since that is the case most likely to break.

Tests

tests/unit/parsers/test_c_cpp_pointer_returns.py — 11 tests. Reverting the source changes fails 9 of 11; the 2 that still pass are the plain-int control cases, which is the correct behaviour.

Why no golden moved

Worth stating, since an unchanged golden usually means a change didn't apply. The only pointer-returning function in the C/C++ fixtures is const char* get_color_name in tough_macros.c, and it sits inside an X-macro region tree-sitter cannot parse:

ERROR/missing nodes: [('ERROR', 13), ('ERROR', 66), ('ERROR', 73), ('ERROR', 75), ('ERROR', 79)]

Line 79 is that function. It is unreachable for reasons unrelated to this fix, so the goldens are correctly unchanged. A fixture exercising a plain T* return would be a reasonable follow-up.

🤖 Generated with Claude Code