Tasmota/lib/libesp32/berry/tests/property

Property Tests

Python-side tests for the Berry Python port. These validate correctness properties of the interpreter — from low-level bit manipulation up to end-to-end solidification output equivalence with the C implementation.

Running

# All tests
python3 -m pytest tests/property/ -v --tb=short

# A single file
python3 -m pytest tests/property/test_decoder_roundtrip.py -v

# Tests matching a keyword
python3 -m pytest tests/property/ -k "solidification" -v

Test Patterns

Tests use two main styles:

Hypothesis property tests

Many files use Hypothesis to generate random inputs and verify invariants hold across the input space. These are the "property" tests in the formal sense — e.g., "for any valid opcode/RA/RKB/RKC, encoding then decoding produces the original values."

@given(op=opcodes, ra=ra_vals, rkb=rkb_vals, rkc=rkc_vals)
def test_abc_roundtrip(op, ra, rkb, rkc):
    ins = ISET_OP(op) | ISET_RA(ra) | ISET_RKB(rkb) | ISET_RKC(rkc)
    assert IGET_OP(ins) == op
    ...

End-to-end Berry execution tests

Other files compile and run Berry source code through the Python port, checking output against expected values. When the compiled C berry binary is available in the repo root, some tests also cross-validate against it.

def test_python_port_output(desc, source, expected):
    result = _run_python(source)
    assert result.stdout.strip() == expected

Test Categories

File	What it tests
test_decoder_roundtrip.py	Instruction encoding/decoding round-trip (ABC, ABx, sBx formats)
test_vector_growth.py	Vector capacity doubling strategy matches C be_nextsize
test_string_hash.py	FNV-1a hash equivalence with C be_strhash
test_string_interning.py	Short string interning (≤64 bytes → same object identity)
test_map_compaction.py	Map slot layout after be_map_compact
test_class_member_resolution.py	Instance member lookup through inheritance chain
test_upvalue_lifecycle.py	Open/closed upvalue transitions
test_lexer_token_equivalence.py	Tokenizer output matches C lexer
test_enum_preservation.py	Python enum constants match C integer values
test_bytecode_roundtrip.py	Save/load bytecode produces equivalent protos
test_exception_unwinding.py	try/except/raise stack unwinding
test_memory_tracking.py	gc.usage counter consistency
test_instruction_disassembly.py	be_print_inst text output matches C
test_native_stack_layout.py	Native function call argument layout
test_toidentifier.py	String-to-C-identifier _X escape encoding
test_solidification_output.py	Solidified C source is character-for-character identical to C output
test_compilation_equivalence.py	Bytecode/constants match between Python and C compilers
test_vm_execution.py	End-to-end program output equivalence
test_json_module.py	JSON load/dump correctness and edge cases
test_int64lib.py	int64 class operations
test_os_module.py	os module (getcwd, chdir, mkdir, listdir, path.*)
test_sys_module.py	sys.path() returns module search paths
test_re_module.py, test_re15_*.py	Regex engine (re1.5 port)
test_gc_collect.py	GC mark-sweep cycle
test_vararg.py	Variadic function argument handling
test_cli_repl.py	CLI entry point and REPL behavior
test_file_lib.py, test_file_read_unicode.py	File I/O operations
test_global_module.py, test_getgbl_builtins.py	Global/builtin variable access
test_* (remaining)	Bug regression tests and specific edge cases

C Cross-Validation

Tests in test_vm_execution.py, test_solidification_output.py, and test_compilation_equivalence.py can optionally compare output against the compiled C berry binary. If ./berry exists and is executable, these tests run the same input through both implementations and assert identical results. If the binary isn't available, those specific cross-validation tests are skipped — the Python-only assertions still run.