Merge pull request #811 from DeusData/distill/793-walk-exclusions

fix(pipeline): honor discovery exclusions in pkgmap/path-alias/envscan walks

Author: DeusData

src/pipeline/pass_envscan.c

@@ -363,16 +363,22 @@ static int scan_env_file(const char *full_path, const char *rel, file_type_t ft,
 /* Process a single directory entry for env scanning. Returns bindings added. */
 static int process_env_entry(cbm_dirent_t *ent, const char *dir_path, const char *root_path,
                              cbm_env_binding_t *out, int max_out, char path_stack[][CBM_SZ_512],
-                             int *stack_top) {
+                             int *stack_top, char **excluded_dirs, int excluded_count) {
     char full_path[CBM_SZ_512];
     snprintf(full_path, sizeof(full_path), "%s/%s", dir_path, ent->name);
+    const char *rel = full_path + strlen(root_path);
+    while (*rel == '/') {
+        rel++;
+    }
 
     struct stat st;
     if (stat(full_path, &st) != 0) {
         return 0;
     }
     if (S_ISDIR(st.st_mode)) {
-        if (!is_ignored_dir(ent->name) && *stack_top < CBM_SZ_256) {
+        if (!is_ignored_dir(ent->name) &&
+            !cbm_pipeline_relpath_is_excluded(rel, excluded_dirs, excluded_count) &&
+            *stack_top < CBM_SZ_256) {
             strncpy(path_stack[*stack_top], full_path, sizeof(path_stack[0]) - 1);
             path_stack[*stack_top][sizeof(path_stack[0]) - SKIP_ONE] = '\0';
             (*stack_top)++;
@@ -386,14 +392,11 @@ static int process_env_entry(cbm_dirent_t *ent, const char *dir_path, const char
     if (ft == FT_UNKNOWN) {
         return 0;
     }
-    const char *rel = full_path + strlen(root_path);
-    while (*rel == '/') {
-        rel++;
-    }
     return scan_env_file(full_path, rel, ft, out, max_out);
 }
 
-int cbm_scan_project_env_urls(const char *root_path, cbm_env_binding_t *out, int max_out) {
+int cbm_scan_project_env_urls_excluded(const char *root_path, cbm_env_binding_t *out, int max_out,
+                                       char **excluded_dirs, int excluded_count) {
     if (!root_path || !out || max_out <= 0) {
         return 0;
     }
@@ -418,9 +421,13 @@ int cbm_scan_project_env_urls(const char *root_path, cbm_env_binding_t *out, int
         cbm_dirent_t *ent;
         while ((ent = cbm_readdir(d)) && count < max_out) {
             count += process_env_entry(ent, dir_path, root_path, out + count, max_out - count,
-                                       path_stack, &stack_top);
+                                       path_stack, &stack_top, excluded_dirs, excluded_count);
         }
         cbm_closedir(d);
     }
     return count;
 }
+
+int cbm_scan_project_env_urls(const char *root_path, cbm_env_binding_t *out, int max_out) {
+    return cbm_scan_project_env_urls_excluded(root_path, out, max_out, NULL, 0);
+}

src/pipeline/pass_parallel.c

@@ -818,7 +818,7 @@ static void merge_pkg_entries(cbm_pipeline_ctx_t *ctx, cbm_pkg_entries_t *pkg_en
      * by the main discoverer (package.json, composer.json — in
      * IGNORED_JSON_FILES) still feed pkgmap. Append into worker 0's
      * array so the existing merge below sees them. */
-    cbm_pkgmap_scan_repo(ctx->repo_path, &pkg_entries[0]);
+    cbm_pkgmap_scan_repo(ctx->repo_path, &pkg_entries[0], ctx->excluded_dirs, ctx->excluded_count);
     cbm_pipeline_set_pkgmap(cbm_pkgmap_build(pkg_entries, worker_count, ctx->project_name));
     for (int i = 0; i < worker_count; i++) {
         cbm_pkg_entries_free(&pkg_entries[i]);

src/pipeline/pass_pkgmap.c

@@ -845,7 +845,7 @@ static bool pkgmap_is_reparse_point(const char *abs_path) {
  * it hang. On Windows we additionally skip reparse points before
  * descending as a best-effort early-out. */
 static int pkgmap_walk_dir(const char *abs_dir, const char *rel_dir, cbm_pkg_entries_t *entries,
-                           int depth) {
+                           int depth, char **excluded_dirs, int excluded_count) {
     if (depth >= PKGMAP_WALK_MAX_DEPTH) {
         cbm_log_info("pkgmap.walk", "depth_cap", rel_dir && rel_dir[0] ? rel_dir : ".");
         return 0;
@@ -874,7 +874,8 @@ static int pkgmap_walk_dir(const char *abs_dir, const char *rel_dir, cbm_pkg_ent
             continue;
         }
         if (S_ISDIR(st.st_mode)) {
-            if (cbm_should_skip_dir(name, CBM_MODE_FULL)) {
+            if (cbm_should_skip_dir(name, CBM_MODE_FULL) ||
+                cbm_pipeline_relpath_is_excluded(rel_path, excluded_dirs, excluded_count)) {
                 continue;
             }
 #ifdef _WIN32
@@ -886,7 +887,8 @@ static int pkgmap_walk_dir(const char *abs_dir, const char *rel_dir, cbm_pkg_ent
                 continue;
             }
 #endif
-            parsed += pkgmap_walk_dir(abs_path, rel_path, entries, depth + 1);
+            parsed += pkgmap_walk_dir(abs_path, rel_path, entries, depth + 1, excluded_dirs,
+                                      excluded_count);
             continue;
         }
         if (!S_ISREG(st.st_mode)) {
@@ -920,11 +922,12 @@ static int pkgmap_walk_dir(const char *abs_dir, const char *rel_dir, cbm_pkg_ent
  * Windows reparse points, so it cannot hang on directory junctions.
  * This is what lets bare workspace imports (e.g. "@org/pkg" declared in
  * an ignored package.json) resolve on Windows as well as POSIX. */
-int cbm_pkgmap_scan_repo(const char *repo_path, cbm_pkg_entries_t *entries) {
+int cbm_pkgmap_scan_repo(const char *repo_path, cbm_pkg_entries_t *entries, char **excluded_dirs,
+                         int excluded_count) {
     if (!repo_path || !entries) {
         return 0;
     }
-    int parsed = pkgmap_walk_dir(repo_path, "", entries, 0);
+    int parsed = pkgmap_walk_dir(repo_path, "", entries, 0, excluded_dirs, excluded_count);
     cbm_log_info("pkgmap.scan_repo", "manifests", pkgmap_itoa(parsed));
     return parsed;
 }
@@ -961,7 +964,8 @@ CBMHashTable *cbm_pkgmap_build_from_files(const cbm_file_info_t *files, int file
  * (the canonical case: package.json, which is in IGNORED_JSON_FILES).
  * Falls back to the files[]-only behaviour if repo_path is NULL. */
 CBMHashTable *cbm_pkgmap_build_from_repo(const char *repo_path, const cbm_file_info_t *files,
-                                         int file_count, const char *project_name) {
+                                         int file_count, const char *project_name,
+                                         char **excluded_dirs, int excluded_count) {
     cbm_pkg_entries_t entries;
     cbm_pkg_entries_init(&entries);
 
@@ -985,7 +989,7 @@ CBMHashTable *cbm_pkgmap_build_from_repo(const char *repo_path, const cbm_file_i
         free(source);
     }
 
-    int from_walk = cbm_pkgmap_scan_repo(repo_path, &entries);
+    int from_walk = cbm_pkgmap_scan_repo(repo_path, &entries, excluded_dirs, excluded_count);
     cbm_log_info("pkgmap.scan", "manifests_from_files", pkgmap_itoa(from_files),
                  "manifests_from_walk", pkgmap_itoa(from_walk), "entries",
                  pkgmap_itoa(entries.count));

src/pipeline/path_alias.c

@@ -20,6 +20,8 @@
 
 #include "pipeline/path_alias.h"
 
+#include "pipeline/pipeline_internal.h"
+
 #include "foundation/compat.h"
 #include "foundation/compat_fs.h"
 #include "foundation/constants.h"
@@ -380,7 +382,8 @@ static const char *const TS_CONFIG_NAMES[] = {"tsconfig.json", "jsconfig.json"};
 enum { TS_CONFIG_NAMES_COUNT = 2 };
 
 static void find_alias_files(const char *abs_dir, const char *rel_dir, alias_config_hit_t *out,
-                             int *count, int max_count, int depth) {
+                             int *count, int max_count, int depth, char **excluded_dirs,
+                             int excluded_count) {
     if (*count >= max_count || depth > CBM_PATH_ALIAS_MAX_DEPTH) {
         return;
     }
@@ -422,12 +425,18 @@ static void find_alias_files(const char *abs_dir, const char *rel_dir, alias_con
         } else {
             snprintf(child_rel, sizeof(child_rel), "%s/%s", rel_dir, name);
         }
-        find_alias_files(child_abs, child_rel, out, count, max_count, depth + 1);
+        if (cbm_pipeline_relpath_is_excluded(child_rel, excluded_dirs, excluded_count)) {
+            continue;
+        }
+        find_alias_files(child_abs, child_rel, out, count, max_count, depth + 1, excluded_dirs,
+                         excluded_count);
     }
     cbm_closedir(d);
 }
 
-cbm_path_alias_collection_t *cbm_load_path_aliases(const char *repo_path) {
+cbm_path_alias_collection_t *cbm_load_path_aliases_excluded(const char *repo_path,
+                                                            char **excluded_dirs,
+                                                            int excluded_count) {
     if (!repo_path) {
         return NULL;
     }
@@ -436,7 +445,8 @@ cbm_path_alias_collection_t *cbm_load_path_aliases(const char *repo_path) {
         return NULL;
     }
     int count = 0;
-    find_alias_files(repo_path, "", hits, &count, CBM_PATH_ALIAS_MAX_FILES, 0);
+    find_alias_files(repo_path, "", hits, &count, CBM_PATH_ALIAS_MAX_FILES, 0, excluded_dirs,
+                     excluded_count);
     if (count >= CBM_PATH_ALIAS_MAX_FILES) {
         cbm_log_warn("path_alias.files.cap_hit", "repo", repo_path, "kept", "256_of_more");
     }
@@ -479,6 +489,10 @@ cbm_path_alias_collection_t *cbm_load_path_aliases(const char *repo_path) {
     return coll;
 }
 
+cbm_path_alias_collection_t *cbm_load_path_aliases(const char *repo_path) {
+    return cbm_load_path_aliases_excluded(repo_path, NULL, 0);
+}
+
 const cbm_path_alias_map_t *cbm_path_alias_find_for_file(const cbm_path_alias_collection_t *coll,
                                                          const char *rel_path) {
     if (!coll || !rel_path) {

src/pipeline/path_alias.h

@@ -72,6 +72,9 @@ typedef struct {
  * are found (also NULL on out-of-memory). Caller frees with
  * cbm_path_alias_collection_free. */
 cbm_path_alias_collection_t *cbm_load_path_aliases(const char *repo_path);
+cbm_path_alias_collection_t *cbm_load_path_aliases_excluded(const char *repo_path,
+                                                            char **excluded_dirs,
+                                                            int excluded_count);
 
 /* Free a collection produced by cbm_load_path_aliases. NULL-safe. */
 void cbm_path_alias_collection_free(cbm_path_alias_collection_t *coll);

src/pipeline/pipeline.c

@@ -737,8 +737,9 @@ static int run_sequential_pipeline(cbm_pipeline_t *p, cbm_pipeline_ctx_t *ctx,
      * Use the repo-walking variant so manifests filtered out by the main
      * discoverer (package.json, composer.json) still feed pkgmap and let
      * workspace imports like `@my/pkg` resolve to their target Module. */
-    cbm_pipeline_set_pkgmap(
-        cbm_pkgmap_build_from_repo(ctx->repo_path, files, file_count, ctx->project_name));
+    cbm_pipeline_set_pkgmap(cbm_pkgmap_build_from_repo(ctx->repo_path, files, file_count,
+                                                       ctx->project_name, ctx->excluded_dirs,
+                                                       ctx->excluded_count));
 
     CBMFileResult **seq_cache = (CBMFileResult **)calloc(file_count, sizeof(CBMFileResult *));
     if (seq_cache) {
@@ -1261,7 +1262,8 @@ int cbm_pipeline_run(cbm_pipeline_t *p) {
 
     /* Phase 2b: Load build-tool path aliases (tsconfig/jsconfig today). NULL
      * when no usable configs are found — non-TS projects pay nothing. */
-    path_aliases = cbm_load_path_aliases(p->repo_path);
+    path_aliases =
+        cbm_load_path_aliases_excluded(p->repo_path, p->excluded_dirs, p->excluded_count);
 
     /* Build shared context for pass functions */
     cbm_pipeline_ctx_t ctx = {
@@ -1273,6 +1275,8 @@ int cbm_pipeline_run(cbm_pipeline_t *p) {
         .pipeline = p, /* so passes can record per-file skips (Track B) */
         .mode = (int)p->mode,
         .path_aliases = path_aliases,
+        .excluded_dirs = p->excluded_dirs,
+        .excluded_count = p->excluded_count,
     };
 
     rc = run_extraction_phase(p, &ctx, files, file_count);

src/pipeline/pipeline_incremental.c

@@ -806,7 +806,16 @@ int cbm_pipeline_run_incremental(cbm_pipeline_t *p, const char *db_path, cbm_fil
     cbm_log_info("incremental.registry_seed", "symbols", itoa_buf(cbm_registry_size(registry)),
                  "elapsed_ms", itoa_buf((int)elapsed_ms(t)));
 
-    cbm_path_alias_collection_t *path_aliases = cbm_load_path_aliases(cbm_pipeline_repo_path(p));
+    /* Discovery exclusions (gitignore + skip dirs) captured by the run that
+     * routed here. Borrowed from the pipeline so the auxiliary repo walks
+     * (pkgmap via merge_pkg_entries, path aliases) skip excluded subtrees on
+     * incremental runs too — same borrow as the full path (#792/#804). */
+    char **excluded_dirs = NULL;
+    int excluded_count = 0;
+    cbm_pipeline_get_excluded(p, &excluded_dirs, &excluded_count);
+
+    cbm_path_alias_collection_t *path_aliases =
+        cbm_load_path_aliases_excluded(cbm_pipeline_repo_path(p), excluded_dirs, excluded_count);
 
     cbm_pipeline_ctx_t ctx = {
         .project_name = project,
@@ -817,6 +826,8 @@ int cbm_pipeline_run_incremental(cbm_pipeline_t *p, const char *db_path, cbm_fil
         .pipeline = p, /* so passes can record per-file skips (Track B) */
         .mode = cbm_pipeline_get_mode(p),
         .path_aliases = path_aliases,
+        .excluded_dirs = excluded_dirs,
+        .excluded_count = excluded_count,
     };
 
     for (int i = 0; i < ci; i++) {

src/pipeline/pipeline_internal.h

@@ -16,6 +16,7 @@
 #include "cbm.h"
 #include "lsp/go_lsp.h" /* CBMLSPDef for cbm_parallel_resolve cross-LSP inputs */
 #include <stdatomic.h>
+#include <string.h>
 
 /* ── Shared pipeline constants ─────────────────────────────────── */
 
@@ -81,8 +82,30 @@ typedef struct {
      * configs are an easy follow-on). NULL when no usable configs were found.
      * Owned by pipeline.c / pipeline_incremental.c. */
     const cbm_path_alias_collection_t *path_aliases;
+
+    /* Directory subtrees excluded during discovery. Borrowed from pipeline.c. */
+    char **excluded_dirs;
+    int excluded_count;
 } cbm_pipeline_ctx_t;
 
+static inline int cbm_pipeline_relpath_is_excluded(const char *rel_path, char *const *excluded_dirs,
+                                                   int excluded_count) {
+    if (!rel_path || rel_path[0] == '\0' || !excluded_dirs || excluded_count <= 0) {
+        return 0;
+    }
+    for (int i = 0; i < excluded_count; i++) {
+        const char *excluded = excluded_dirs[i];
+        if (!excluded || excluded[0] == '\0') {
+            continue;
+        }
+        size_t n = strlen(excluded);
+        if (strncmp(rel_path, excluded, n) == 0 && (rel_path[n] == '\0' || rel_path[n] == '/')) {
+            return SKIP_ONE;
+        }
+    }
+    return 0;
+}
+
 /* Get the current pipeline's package map (NULL if none). */
 CBMHashTable *cbm_pipeline_get_pkgmap(void);
 void cbm_pipeline_set_pkgmap(CBMHashTable *map);
@@ -131,9 +154,11 @@ CBMHashTable *cbm_pkgmap_build(cbm_pkg_entries_t *worker_entries, int worker_cou
                                const char *project_name);
 
 /* Build pkgmap by reading manifest files from the files array (sequential path). */
-int cbm_pkgmap_scan_repo(const char *repo_path, cbm_pkg_entries_t *entries);
+int cbm_pkgmap_scan_repo(const char *repo_path, cbm_pkg_entries_t *entries, char **excluded_dirs,
+                         int excluded_count);
 CBMHashTable *cbm_pkgmap_build_from_repo(const char *repo_path, const cbm_file_info_t *files,
-                                         int file_count, const char *project_name);
+                                         int file_count, const char *project_name,
+                                         char **excluded_dirs, int excluded_count);
 CBMHashTable *cbm_pkgmap_build_from_files(const cbm_file_info_t *files, int file_count,
                                           const char *project_name);
 
@@ -530,8 +555,14 @@ typedef struct {
 /* Scan a project directory for environment variable assignments with URL values.
  * Walks the filesystem, scans Dockerfiles, shell scripts, .env, YAML, TOML,
  * Terraform, and .properties files. Filters out secrets.
- * Returns number of bindings written to out (up to max_out). */
+ * Returns number of bindings written to out (up to max_out).
+ * NOTE: this walker currently has no production callers — it is exercised
+ * only by tests. The _excluded variant honors discovery exclusions for
+ * consistency with the pkgmap/path-alias walks (#792); the plain variant
+ * scans unexcluded (NULL exclusion list). */
 int cbm_scan_project_env_urls(const char *root_path, cbm_env_binding_t *out, int max_out);
+int cbm_scan_project_env_urls_excluded(const char *root_path, cbm_env_binding_t *out, int max_out,
+                                       char **excluded_dirs, int excluded_count);
 
 /* ── Incremental pipeline (pipeline_incremental.c) ───────────────── */
 

tests/test_path_alias.c

@@ -336,6 +336,62 @@ TEST(path_alias_loader_monorepo_dotdot_climb) {
     PASS();
 }
 
+/* ── Loader honors discovery exclusions (#792) ─────────────────── */
+
+/* find_alias_files must not descend into discovery-excluded subtrees.
+ * Control run first (no exclusions → both configs collected) so the
+ * exclusion assertion below cannot pass vacuously. */
+TEST(path_alias_loader_honors_discovery_exclusions) {
+    char tmpl[256];
+    snprintf(tmpl, sizeof(tmpl), "/tmp/cbm_palias_excl_XXXXXX");
+    char *root = cbm_mkdtemp(tmpl);
+    ASSERT_NOT_NULL(root);
+
+    char sub[512];
+    snprintf(sub, sizeof(sub), "%s/big_generated", root);
+    cbm_mkdir(sub);
+
+    char path[512];
+    snprintf(path, sizeof(path), "%s/tsconfig.json", root);
+    ASSERT_EQ(write_file(path,
+                         "{\n  \"compilerOptions\": {\n    \"paths\": {\n"
+                         "      \"@root/*\": [\"shared/*\"]\n    }\n  }\n}\n"),
+              0);
+    snprintf(path, sizeof(path), "%s/big_generated/tsconfig.json", root);
+    ASSERT_EQ(write_file(path,
+                         "{\n  \"compilerOptions\": {\n    \"paths\": {\n"
+                         "      \"@gen/*\": [\"./src/*\"]\n    }\n  }\n}\n"),
+              0);
+
+    /* Control: the unexcluded loader collects BOTH configs. */
+    cbm_path_alias_collection_t *coll = cbm_load_path_aliases(root);
+    ASSERT_NOT_NULL(coll);
+    ASSERT_EQ(coll->count, 2);
+    cbm_path_alias_collection_free(coll);
+
+    /* Excluding big_generated drops its config; the root one survives. */
+    char *excluded[] = {(char *)"big_generated"};
+    coll = cbm_load_path_aliases_excluded(root, excluded, 1);
+    ASSERT_NOT_NULL(coll);
+    ASSERT_EQ(coll->count, 1);
+    const cbm_path_alias_map_t *m = cbm_path_alias_find_for_file(coll, "src/x.ts");
+    ASSERT_NOT_NULL(m);
+    char *r = cbm_path_alias_resolve(m, "@root/utils");
+    ASSERT_NOT_NULL(r);
+    ASSERT_STR_EQ(r, "shared/utils");
+    free(r);
+    cbm_path_alias_collection_free(coll);
+
+    snprintf(path, sizeof(path), "%s/big_generated/tsconfig.json", root);
+    unlink(path);
+    snprintf(path, sizeof(path), "%s/tsconfig.json", root);
+    unlink(path);
+    snprintf(path, sizeof(path), "%s/big_generated", root);
+    rmdir(path);
+    rmdir(root);
+    PASS();
+}
+
 /* ── Loader returns NULL when no configs found ─────────────────── */
 
 TEST(path_alias_loader_no_configs) {
@@ -363,5 +419,6 @@ void suite_path_alias(void) {
     RUN_TEST(path_alias_find_for_file_nearest_ancestor);
     RUN_TEST(path_alias_loader_monorepo);
     RUN_TEST(path_alias_loader_monorepo_dotdot_climb);
+    RUN_TEST(path_alias_loader_honors_discovery_exclusions);
     RUN_TEST(path_alias_loader_no_configs);
 }