from src.cpg.annotate import annotate_sample

STRCPY_FUNC = """void copy_input(char *src) {
    char buf[64];
    strcpy(buf, src);
}
"""


def test_annotate_sample_vulnerable_row():
    ann = annotate_sample("s1", STRCPY_FUNC, patch_lines_vuln=[3], cwe="CWE-119")
    assert ann.l1_lines == [3]
    assert 3 in ann.l2_lines
    assert ann.l3_lines == [1, 2, 3, 4]
    assert ann.trace_unverifiable is False
    assert len(ann.trace_gt) >= 1
    assert ann.annotation_backend == "tree_sitter"


def test_annotate_sample_benign_row_has_no_localization():
    ann = annotate_sample("s2", STRCPY_FUNC, patch_lines_vuln=[], cwe=None)
    assert ann.l1_lines == []
    assert ann.l2_lines == []
    assert ann.l3_lines == [1, 2, 3, 4]
    assert ann.trace_unverifiable is True
    assert ann.trace_reason == "target_0_no_patch_lines"


MULTI_HOP_FUNC = """void process(char *input) {
    int n = strlen(input);
    char buf[128];
    memcpy(buf, input, n);
}
"""


def test_annotate_sample_l2_includes_backward_slice():
    ann = annotate_sample("s3", MULTI_HOP_FUNC, patch_lines_vuln=[4], cwe="CWE-119")
    assert 4 in ann.l1_lines
    # L2 should pull in line 2 (n = strlen(input)) via the def-use slice from
    # the memcpy call that uses n.
    assert 2 in ann.l2_lines


NO_DANGEROUS_CALL_FUNC = """int add(int a, int b) {
    return a + b;
}
"""


def test_annotate_sample_traces_even_without_a_dangerous_call():
    # Amendment (DECISIONS.md "Phase 2 trace-mining redefinition"): the sink
    # is L1 itself, not a dangerous-call-table lookup. This function has no
    # "dangerous call" at all, but L1 (the return statement) directly uses
    # both parameters with no in-function predecessor, so it's a valid
    # zero-hop trace — previously this was unconditionally trace_unverifiable.
    ann = annotate_sample("s4", NO_DANGEROUS_CALL_FUNC, patch_lines_vuln=[2], cwe="CWE-119")
    assert ann.trace_unverifiable is False
    assert ann.trace_gt != []
    assert ann.trace_lengths == [1]
    assert ann.used_fallback_sink_search is False


NO_SOURCE_NO_SINK_FUNC = """void f(void) {
    do_something();
}
"""


def test_annotate_sample_genuinely_unverifiable_when_l1_has_no_inputs():
    # L1 statement uses no identifiers at all (no params, no in-function
    # defs) -> genuinely no source to trace back to.
    ann = annotate_sample("s5", NO_SOURCE_NO_SINK_FUNC, patch_lines_vuln=[2], cwe=None)
    assert ann.trace_unverifiable is True
    assert ann.trace_reason == "no_source_found"