fix(pipeline): a bare Go reference never binds a struct Field

The READS/WRITES resolvers and the USAGE registry fallback hand bare
reference text to the short-name registry, which contains Field nodes -
so once Go struct fields exist (#1935), every Go local err := ... binds
whichever struct field is named err, project-wide: 21308 USAGE and 5191
WRITES onto Go fields on the measured repo, top target a test struct's
field T collecting 3013 edges. In Go that binding is impossible by
construction: a field is only reachable through a selector expression
(x.f), and selector references resolve on the LSP path - every
Field-targeted reference edge in the census carried dot-less text.

Add cbm_go_suppress_bare_field_ref() next to the #1928 predicate and
consult it at the same four sites (both READS/WRITES resolvers, both
USAGE registry fallbacks): drop the bind when the file is Go, the
target label is Field, and the reference text has no '.'. Go-gated
because C#/Java/C++/Python method bodies legitimately reference their
own members bare (cp_reads_writes_cs_static_field pins that shape).

Field-validated on the #1940 stack: USAGE onto Go fields 21308 -> 0,
WRITES 5191 -> 0; the only remaining field-targeted edges are 2466
CALLS, which are #1906/#1907's selector-guard territory. Reproduce-
first pipeline probes (sequential + parallel twins) were RED on the
stack without this commit.

Fixes #1942

Signed-off-by: Ilya Brykau <ilya.brykau@orca.security>
This commit is contained in:
Ilya Brykau
2026-08-31 11:09:27 +02:00
parent e07a12b669
commit cc00a0275e
6 changed files with 169 additions and 0 deletions
+9
View File
@@ -2660,6 +2660,11 @@ static void resolve_file_usages(resolve_ctx_t *rc, resolve_worker_state_t *ws,
if (tgt && cbm_suppress_cross_language_ref(lang, tgt->file_path)) {
continue;
}
/* #1942: a bare Go reference can never denote a struct field. */
if (tgt &&
cbm_go_suppress_bare_field_ref(lang == CBM_LANG_GO, usage->ref_name, tgt->label)) {
continue;
}
if (usage->semantic_reference_blocked && (usage->semantic_reference_local_shadow ||
cbm_pipeline_node_is_callable_target(tgt))) {
continue;
@@ -2745,6 +2750,10 @@ static void resolve_file_rw(resolve_ctx_t *rc, resolve_worker_state_t *ws, CBMFi
if (cbm_suppress_cross_language_ref(lang, tgt->file_path)) {
continue;
}
/* #1942: a bare Go reference can never denote a struct field. */
if (cbm_go_suppress_bare_field_ref(lang == CBM_LANG_GO, rw->var_name, tgt->label)) {
continue;
}
const char *etype = rw->is_write ? "WRITES" : "READS";
cbm_gbuf_insert_edge(ws->local_edge_buf, src->id, tgt->id, etype, "{}");
}
+9
View File
@@ -216,6 +216,11 @@ static int resolve_usage_edges(cbm_pipeline_ctx_t *ctx, const CBMFileResult *res
if (tgt && cbm_suppress_cross_language_ref(lang, tgt->file_path)) {
continue;
}
/* #1942: a bare Go reference can never denote a struct field. */
if (tgt &&
cbm_go_suppress_bare_field_ref(lang == CBM_LANG_GO, usage->ref_name, tgt->label)) {
continue;
}
if (usage->semantic_reference_blocked && (usage->semantic_reference_local_shadow ||
cbm_pipeline_node_is_callable_target(tgt))) {
continue;
@@ -304,6 +309,10 @@ static int resolve_rw_edges(cbm_pipeline_ctx_t *ctx, const CBMFileResult *result
if (cbm_suppress_cross_language_ref(lang, tgt->file_path)) {
continue;
}
/* #1942: a bare Go reference can never denote a struct field. */
if (cbm_go_suppress_bare_field_ref(lang == CBM_LANG_GO, rw->var_name, tgt->label)) {
continue;
}
const char *edge_type = rw->is_write ? "WRITES" : "READS";
cbm_gbuf_insert_edge(ctx->gbuf, src->id, tgt->id, edge_type, "{}");
+8
View File
@@ -296,6 +296,14 @@ bool cbm_suppress_cross_language_suffix_match(CBMLanguage caller_lang, const cha
* unit-tested in test_registry.c. */
bool cbm_suppress_cross_language_ref(CBMLanguage caller_lang, const char *target_file_path);
/* #1942: a bare (dot-less) Go reference can never denote a struct field —
* field access is always a selector expression, and selector references
* resolve on the LSP path. Drops a READS/WRITES/USAGE bind whose target is a
* Field when the reference text carries no '.'. Go only: other OO languages
* legitimately reference their own members bare inside method bodies. Pure;
* unit-tested in test_registry.c. */
bool cbm_go_suppress_bare_field_ref(bool is_go, const char *ref_name, const char *target_label);
/* Get the label of a qualified name, or NULL if not found. */
const char *cbm_registry_label_of(const cbm_registry_t *r, const char *qn);
+18
View File
@@ -546,6 +546,24 @@ bool cbm_suppress_cross_language_ref(CBMLanguage caller_lang, const char *target
return true;
}
bool cbm_go_suppress_bare_field_ref(bool is_go, const char *ref_name, const char *target_label) {
/* #1942: a bare (dot-less) Go reference can never denote a struct field —
* field access is always a selector expression (x.f), and selector
* references resolve through the LSP join, never through the bare-name
* registry fallback. Every Field-targeted reference edge in the field
* census carried dot-less text, so dropping the bind loses nothing real.
* Go-gated: a C#/Java/C++/Python method body legitimately references its
* own members bare (cp_reads_writes_cs_static_field pins that shape as
* required), so a global veto would break those languages. */
if (!is_go || !ref_name || !ref_name[0] || !target_label) {
return false;
}
if (strcmp(target_label, "Field") != 0) {
return false;
}
return strchr(ref_name, '.') == NULL;
}
/* ── Lifecycle ──────────────────────────────────────────────────── */
cbm_registry_t *cbm_registry_new(void) {
+104
View File
@@ -4832,6 +4832,108 @@ TEST(pipeline_go_rw_usage_never_cross_into_c_parallel) {
PASS();
}
static int count_nodes_named(cbm_store_t *s, const char *project, const char *name);
/* Fixture for the #1942 bare-reference-vs-Field probes: a Go struct field
* named like the commonest local (`err`), and a function whose local of the
* same name must NOT bind it — in Go a field is only reachable through a
* selector expression, never a bare identifier. Needs Go Field extraction
* (#1935) to have anything to falsely bind. */
static void write_go_bare_field_fixture(const char *tmp, int pad_files) {
write_temp_file(tmp, "go.mod", "module example.com/fxbare\n\ngo 1.22\n");
write_temp_file(tmp, "state/state.go",
"package state\n"
"\n"
"type Tracker struct {\n"
"\terr error\n"
"\tn int\n"
"}\n");
write_temp_file(tmp, "app/app.go",
"package app\n"
"\n"
"import \"errors\"\n"
"\n"
"func Run() error {\n"
"\terr := errors.New(\"x\")\n"
"\treturn err\n"
"}\n");
for (int i = 0; i < pad_files; i++) {
char name[64];
char body[128];
snprintf(name, sizeof(name), "pad/filler%d.go", i);
snprintf(body, sizeof(body), "package pad\n\nfunc filler%d() int { return %d }\n", i, i);
write_temp_file(tmp, name, body);
}
}
TEST(pipeline_go_bare_ref_never_binds_field) {
/* #1942: the READS/WRITES resolvers and the USAGE registry fallback hand
* bare reference text to the short-name registry, which contains Field
* nodes — so every Go local `err := …` bound whichever struct field was
* named err (21308 USAGE / 5191 WRITES onto Go fields on the measured
* repo, top target a test struct's field collecting 3013 edges).
* Sequential-path twin; the parallel twin follows. */
char tmp[256];
snprintf(tmp, sizeof(tmp), "/tmp/cbm_go_bare_XXXXXX");
if (!cbm_mkdtemp(tmp)) {
FAIL("tmpdir");
}
write_go_bare_field_fixture(tmp, 0);
char db_path[512];
snprintf(db_path, sizeof(db_path), "%s/go_bare.db", tmp);
cbm_pipeline_t *p = cbm_pipeline_new(tmp, db_path, CBM_MODE_FULL);
ASSERT_NOT_NULL(p);
ASSERT_EQ(cbm_pipeline_run(p), 0);
const char *project = cbm_pipeline_project_name(p);
cbm_store_t *s = cbm_store_open_path(db_path);
ASSERT_NOT_NULL(s);
/* The field must exist for the probe to mean anything (#1935's fix). */
ASSERT_TRUE(count_nodes_named(s, project, "err") >= 1);
/* Reproduce-first: RED before the fix — the bare local binds the field. */
ASSERT_FALSE(cross_file_edge_exists(s, project, "Run", "err", "WRITES"));
ASSERT_FALSE(cross_file_edge_exists(s, project, "Run", "err", "READS"));
ASSERT_FALSE(cross_file_edge_exists(s, project, "Run", "err", "USAGE"));
cbm_store_close(s);
cbm_pipeline_free(p);
th_rmtree(tmp);
PASS();
}
TEST(pipeline_go_bare_ref_never_binds_field_parallel) {
/* Parallel twin: resolve_file_rw / resolve_file_usages are independent
* resolvers and must consult the same predicate (#1928's lesson). */
char tmp[256];
snprintf(tmp, sizeof(tmp), "/tmp/cbm_go_barep_XXXXXX");
if (!cbm_mkdtemp(tmp)) {
FAIL("tmpdir");
}
write_go_bare_field_fixture(tmp, 52);
char db_path[512];
snprintf(db_path, sizeof(db_path), "%s/go_barep.db", tmp);
cbm_pipeline_t *p = cbm_pipeline_new(tmp, db_path, CBM_MODE_FULL);
ASSERT_NOT_NULL(p);
ASSERT_EQ(cbm_pipeline_run(p), 0);
const char *project = cbm_pipeline_project_name(p);
cbm_store_t *s = cbm_store_open_path(db_path);
ASSERT_NOT_NULL(s);
ASSERT_TRUE(count_nodes_named(s, project, "err") >= 1);
ASSERT_FALSE(cross_file_edge_exists(s, project, "Run", "err", "WRITES"));
ASSERT_FALSE(cross_file_edge_exists(s, project, "Run", "err", "READS"));
ASSERT_FALSE(cross_file_edge_exists(s, project, "Run", "err", "USAGE"));
cbm_store_close(s);
cbm_pipeline_free(p);
th_rmtree(tmp);
PASS();
}
/* Count nodes with the given exact name in the project (e.g. a Route path). */
static int count_nodes_named(cbm_store_t *s, const char *project, const char *name) {
cbm_node_t *ns = NULL;
@@ -12951,6 +13053,8 @@ SUITE(pipeline) {
RUN_TEST(pipeline_python_receiver_suppresses_weak_method_edge);
RUN_TEST(pipeline_go_rw_usage_never_cross_into_c);
RUN_TEST(pipeline_go_rw_usage_never_cross_into_c_parallel);
RUN_TEST(pipeline_go_bare_ref_never_binds_field);
RUN_TEST(pipeline_go_bare_ref_never_binds_field_parallel);
RUN_TEST(pipeline_tsjs_receiver_parallel_keeps_service_edges);
RUN_TEST(pipeline_python_receiver_parallel_suppresses_weak_method_edges);
RUN_TEST(pipeline_parallel_python_cross_only_dunder_gets_synthetic_carrier);
+21
View File
@@ -839,6 +839,26 @@ TEST(cross_language_ref_drops_go_vs_c) {
PASS();
}
TEST(go_bare_ref_never_binds_field) {
/* #1942: a bare (dot-less) Go reference can never denote a struct field —
* field access is always a selector expression. */
ASSERT_TRUE(cbm_go_suppress_bare_field_ref(true, "err", "Field"));
ASSERT_TRUE(cbm_go_suppress_bare_field_ref(true, "config", "Field"));
/* A selector-shaped reference may bind a field. */
ASSERT_FALSE(cbm_go_suppress_bare_field_ref(true, "t.err", "Field"));
/* Bare references to non-fields are untouched. */
ASSERT_FALSE(cbm_go_suppress_bare_field_ref(true, "err", "Variable"));
ASSERT_FALSE(cbm_go_suppress_bare_field_ref(true, "err", "Function"));
/* Other languages reference their own members bare inside methods —
* never suppressed (cp_reads_writes_cs_static_field pins the C# shape). */
ASSERT_FALSE(cbm_go_suppress_bare_field_ref(false, "_count", "Field"));
/* Degenerate inputs → nothing to judge. */
ASSERT_FALSE(cbm_go_suppress_bare_field_ref(true, NULL, "Field"));
ASSERT_FALSE(cbm_go_suppress_bare_field_ref(true, "", "Field"));
ASSERT_FALSE(cbm_go_suppress_bare_field_ref(true, "err", NULL));
PASS();
}
TEST(dynamic_suppress_drops_weak_method_matches) {
/* #592/#606/#1276: a member call whose receiver the LSP could not type, that
* landed via a WEAK short-name strategy, is generic-resolver noise → drop.
@@ -974,6 +994,7 @@ SUITE(registry) {
RUN_TEST(perl_suppress_keeps_high_confidence_and_genuine_calls);
RUN_TEST(cross_language_suffix_match_drops_py_vs_js);
RUN_TEST(cross_language_ref_drops_go_vs_c);
RUN_TEST(go_bare_ref_never_binds_field);
RUN_TEST(dynamic_suppress_drops_weak_method_matches);
RUN_TEST(dynamic_suppress_keeps_high_confidence_and_non_methods);
}