fix: isolate validation and handle empty initialization state

Makefile.cbm

@@ -1053,10 +1053,12 @@ MIMALLOC_OBJ_TEST = $(BUILD_DIR)/mimalloc.o
 MIMALLOC_OBJ_TSAN = $(BUILD_DIR)/tsan_mimalloc.o
 MIMALLOC_OBJ_PROD = $(BUILD_DIR)/prod_mimalloc.o
 
-$(BUILD_DIR)/mimalloc.o: $(MIMALLOC_SRC) | $(BUILD_DIR)
+$(BUILD_DIR)/mimalloc.o: $(MIMALLOC_SRC) $(wildcard vendored/mimalloc/src/*.c) \
+                         $(wildcard vendored/mimalloc/include/*.h) Makefile.cbm | $(BUILD_DIR)
 	$(CC) $(MIMALLOC_CFLAGS_TEST) -c -o $@ $<
 
-$(BUILD_DIR)/tsan_mimalloc.o: $(MIMALLOC_SRC) | $(BUILD_DIR)
+$(BUILD_DIR)/tsan_mimalloc.o: $(MIMALLOC_SRC) $(wildcard vendored/mimalloc/src/*.c) \
+                              $(wildcard vendored/mimalloc/include/*.h) Makefile.cbm | $(BUILD_DIR)
 	$(CC) $(MIMALLOC_CFLAGS_TEST) $(TSAN_SANITIZE) -c -o $@ $<
 
 # static.c is an AMALGAMATION: it #includes options.c and friends, none of which

internal/cbm/callable_sig.c

@@ -79,7 +79,9 @@ static void sig_raw(sig_ctx_t *c, const char *s, size_t n) {
             c->overflow = true;
             return;
         }
-        memcpy(grown, c->buf, c->len);
+        if (c->len != 0) {
+            memcpy(grown, c->buf, c->len);
+        }
         c->buf = grown;
         c->cap = cap;
     }

internal/cbm/doclink_perl.c

@@ -1538,6 +1538,9 @@ static int pl_pkg_order_cmp(const void *a, const void *b) {
 /* Give every declaration the index of its name's P record (names in order of
  * their first declaration) and write the P records. */
 static void pl_number_packages(pl_walk_t *w, pl_sb_t *sb) {
+    if (w->npkgs == 0) {
+        return;
+    }
     qsort(w->pkgs, (size_t)w->npkgs, sizeof(pl_pkg_t), pl_pkg_name_cmp);
     for (int i = 0; i < w->npkgs; i++) { /* index = order of the name's first declaration */
         bool repeat =
@@ -1572,7 +1575,9 @@ static int pl_subdecl_cmp(const void *a, const void *b) {
 /* One S record per sub definition of the file, with the package the tree
  * places it in. */
 static void pl_put_subs(CBMExtractCtx *ctx, pl_walk_t *w, pl_sb_t *sb) {
-    qsort(w->subs, (size_t)w->nsubs, sizeof(pl_subdecl_t), pl_subdecl_cmp);
+    if (w->nsubs > SKIP_ONE) {
+        qsort(w->subs, (size_t)w->nsubs, sizeof(pl_subdecl_t), pl_subdecl_cmp);
+    }
     const CBMDefArray *defs = &ctx->result->defs;
     for (int i = 0; i < defs->count; i++) {
         const CBMDefinition *d = &defs->items[i];

scripts/test.sh

@@ -492,6 +492,7 @@ fi
 # release artifacts free of it; scripts/ci/check-binary-composition.sh proves
 # they stay that way.
 echo "=== Step 5: parent-death watchdog regression (#406/#407) ==="
+CC="$CC" bash "$ROOT/tests/test_mimalloc_preinit.sh"
 make -j"$NPROC" -f Makefile.cbm cbm TEST_SEAMS=1 ${MAKE_ARGS[@]+"${MAKE_ARGS[@]}"}
 WATCHDOG_BINARY="$ROOT/$BUILD_DIR/codebase-memory-mcp"
 CBM_TEST_BINARY="$WATCHDOG_BINARY" bash "$ROOT/tests/test_parent_watchdog.sh"

scripts/vendored-checksums.txt

@@ -525,14 +525,14 @@ c5fa4d1d5820cc59f39cb3cb9a9a171e5144b2f63d285cf3450be0a8e549ede8  internal/cbm/v
 85b458dc150c4d17873dd5e73d01cfa23c797997a09d841884e08378242e929c  internal/cbm/vendored/grammars/objectscript_routine/LICENSE
 f3a047fba6d238da258c0370878e703504b347aad13add09a021935daa891bd3  internal/cbm/vendored/grammars/objectscript_routine/objectscript_common.h
 fb9320a77f7c3c110540e9a569c2a8dd7ff74784ae05aac962f4bcbab1343640  internal/cbm/vendored/grammars/objectscript_routine/parser.c
-2fde85097c7df0180b7a204eeae97c32a541f1fcc1b44a3837a57f26ac9966fa  internal/cbm/vendored/grammars/objectscript_routine/scanner.c
+1da66e46ba02fcf2f9ca42b2a80c8658e150ae7b5154d81b8968117b959b5af7  internal/cbm/vendored/grammars/objectscript_routine/scanner.c
 b29c1c9fb7cc82f58c84b376df1297d6e2737a1d655fd356db0859e3c29c2fea  internal/cbm/vendored/grammars/objectscript_routine/tree_sitter/alloc.h
 31e60a1bff6f715afacce03b5b70efe42b58371b4f9595dd4af52a577ff9608c  internal/cbm/vendored/grammars/objectscript_routine/tree_sitter/array.h
 180b893c8734778fd32f372dfbc27bd6ad1cd2221f26150b31256ff6716320d2  internal/cbm/vendored/grammars/objectscript_routine/tree_sitter/parser.h
 85b458dc150c4d17873dd5e73d01cfa23c797997a09d841884e08378242e929c  internal/cbm/vendored/grammars/objectscript_udl/LICENSE
 f3a047fba6d238da258c0370878e703504b347aad13add09a021935daa891bd3  internal/cbm/vendored/grammars/objectscript_udl/objectscript_common.h
 2003ffb1a07ec30a2f7176c65aae958b804074af8356a81a3d42417a240f2e15  internal/cbm/vendored/grammars/objectscript_udl/parser.c
-2e80c4939726948b0ebf8af56103a9c0278a038eeb5be0b769c59d2b4e2d36b1  internal/cbm/vendored/grammars/objectscript_udl/scanner.c
+537be452c5079531e08edc302b65c63f45956510f6ee96ec88c1d7531d203f1e  internal/cbm/vendored/grammars/objectscript_udl/scanner.c
 b29c1c9fb7cc82f58c84b376df1297d6e2737a1d655fd356db0859e3c29c2fea  internal/cbm/vendored/grammars/objectscript_udl/tree_sitter/alloc.h
 31e60a1bff6f715afacce03b5b70efe42b58371b4f9595dd4af52a577ff9608c  internal/cbm/vendored/grammars/objectscript_udl/tree_sitter/array.h
 180b893c8734778fd32f372dfbc27bd6ad1cd2221f26150b31256ff6716320d2  internal/cbm/vendored/grammars/objectscript_udl/tree_sitter/parser.h
@@ -1016,7 +1016,7 @@ eca1fe77d72ab2aaccc447c8119056e9c4e8851bda9a95472b573539f76107ad  vendored/mimal
 ebd95b475f15d5387e6ef3c88a9cb47b88e64d7e55998ae1dc656614676a443b  vendored/mimalloc/src/libc.c
 192ce06ac9076445d223205fb340d8a1fc818dbf2b4c4a801dfd2c10bc6c1162  vendored/mimalloc/src/options.c
 011e965f28c0e2c885a6834d51bd756a458dd37ebfed6fd4d9ad6607ce0390ed  vendored/mimalloc/src/os.c
-1be3a208c678fc34187c2a8ffb75f3fe1eecc674bbf8b2af4d1f329506b1fad5  vendored/mimalloc/src/page-map.c
+e68eee14cb53e13eb1b7e003cfe747c353e3bac849afd0242a6d1862b8f0b895  vendored/mimalloc/src/page-map.c
 7ec275805c9f00eb66710ffa99817b3ed1dba9826435d08f4866f7935152f6ae  vendored/mimalloc/src/page-queue.c
 d3614a59f2407a32e75b051deb1aff31b233f587d325dcc5420ba508551b687e  vendored/mimalloc/src/page.c
 7488e9e562ca85d2d757ec48e4d334935263a2333aa05b6061d8520f427c5890  vendored/mimalloc/src/prim/emscripten/prim.c

src/mcp/mcp.c

@@ -19018,13 +19018,17 @@ static char *handle_detect_changes(cbm_mcp_server_t *srv, const char *args) {
         char hunk_cmd[CBM_SZ_2K];
 #ifdef _WIN32
         snprintf(hunk_cmd, sizeof(hunk_cmd),
-                 "git -C \"%s\" diff --relative --unified=0 \"%s\" \"%s\" -- 2>NUL && "
-                 "git -C \"%s\" diff --relative --unified=0 -- 2>NUL",
+                 "git -C \"%s\" " CBM_TEST_IMPACT_GIT_PINS
+                 " diff --relative " CBM_TEST_IMPACT_DIFF_FLAGS " \"%s\" \"%s\" -- 2>NUL && "
+                 "git -C \"%s\" " CBM_TEST_IMPACT_GIT_PINS
+                 " diff --relative " CBM_TEST_IMPACT_DIFF_FLAGS " -- 2>NUL",
                  root_path, merge_base, head_oid, root_path);
 #else
         snprintf(hunk_cmd, sizeof(hunk_cmd),
-                 "git -C '%s' diff --relative --unified=0 '%s' '%s' -- 2>/dev/null && "
-                 "git -C '%s' diff --relative --unified=0 -- 2>/dev/null",
+                 "git -C '%s' " CBM_TEST_IMPACT_GIT_PINS
+                 " diff --relative " CBM_TEST_IMPACT_DIFF_FLAGS " '%s' '%s' -- 2>/dev/null && "
+                 "git -C '%s' " CBM_TEST_IMPACT_GIT_PINS
+                 " diff --relative " CBM_TEST_IMPACT_DIFF_FLAGS " -- 2>/dev/null",
                  root_path, merge_base, head_oid, root_path);
 #endif
         char hunk_output_path[CBM_SZ_2K] = {0};

src/pipeline/doc_links_perl.c

@@ -244,7 +244,9 @@ static bool pr_index_packages(pr_index_t *ix, const cbm_doclink_build_in_t *in)
     if (ix->oom) {
         return false;
     }
-    qsort(ix->decls, (size_t)ix->ndecls, sizeof(pr_decl_t), pr_decl_cmp);
+    if (ix->ndecls > SKIP_ONE) {
+        qsort(ix->decls, (size_t)ix->ndecls, sizeof(pr_decl_t), pr_decl_cmp);
+    }
     ix->pkg = cbm_ht_create_in(CBM_MEM_CLASS_OTHER, (uint32_t)(ix->ndecls + SKIP_ONE));
     if (!ix->pkg) {
         return false;

src/pipeline/doc_links_rst.c

@@ -405,6 +405,9 @@ static void rr_destroy(void *index) {
 /* The documentation set a document belongs to: the nearest conf.py at or
  * above its directory; NULL for none. */
 static const rr_conf_t *rr_conf_for(const rr_index_t *x, const char *doc) {
+    if (x->nconfs == 0) {
+        return NULL;
+    }
     char dir[RR_PATH_CAP];
     const char *b = rr_base(doc);
     size_t dl = b > doc ? (size_t)(b - doc - SKIP_ONE) : 0;

tests/test_extraction.c

@@ -10250,6 +10250,12 @@ TEST(doc_perl_module_doc_is_the_pod_no_sub_holds) {
     ASSERT_NOT_NULL(r);
     ASSERT_NULL(r->module_doc);
     cbm_free_result(r);
+    /* POD without any declarations leaves both scope sort inputs empty. */
+    r = extract("=head1 NAME\n\nStandalone module.\n\n=cut\n", CBM_LANG_PERL, "t", "Standalone.pm");
+    ASSERT_NOT_NULL(r);
+    ASSERT_EQ(count_defs_with_label(r, "Function"), 0);
+    ASSERT_STR_EQ(r->module_doc, "=head1 NAME\n\nStandalone module.\n\n=cut");
+    cbm_free_result(r);
     PASS();
 }
 

tests/test_main.c

@@ -211,14 +211,12 @@ static void tf_cleanup_cache_sentinel(void) {
     }
 }
 
-/* Client home overrides the CLI honours BEFORE $HOME, so redirecting HOME alone
- * does not isolate them: cbm_codex_config_dir() and its siblings return the
- * ambient path and the suite resolves against the developer's real config —
- * reading its state and writing to it. Same inventory the shell fixtures are
- * already required to neutralize (tests/test_smoke_fixture_contract.sh), kept
- * in one place so a new client cannot be isolated in the smoke scripts and
- * forgotten here. A test that exercises one of these sets it after setup. */
+/* These paths take precedence over HOME and can escape the private fixture.
+ * A test that exercises an override sets it after setup. */
 static const char *const tf_client_home_overrides[] = {
+    "XDG_CONFIG_HOME",
+    "XDG_DATA_HOME",
+    "XDG_CACHE_HOME",
     "CLAUDE_CONFIG_DIR",
     "CODEX_HOME",
     "KIRO_HOME",

tests/test_mcp.c

@@ -17735,7 +17735,8 @@ static bool dc_walk_fixture_open(dc_walk_fixture_t *fx) {
     }
     /* Same git invocation rules as the #1363 fixture above. */
     char cmd[1200];
-    const char *steps[] = {"init -q", "add -A", "commit -q -m init"};
+    const char *steps[] = {"init -q", "config diff.mnemonicPrefix true", "add -A",
+                           "commit -q -m init"};
     for (size_t s = 0; s < sizeof(steps) / sizeof(steps[0]); s++) {
         snprintf(cmd, sizeof(cmd),
                  "git -C \"%s\" -c user.name=t -c user.email=t@t.io -c init.defaultBranch=main "

tests/test_mimalloc_preinit.sh

@@ -0,0 +1,31 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# ELF preinitializers run before allocator constructors, independent of libc version.
+if [[ "$(uname -s)" != Linux ]]; then
+    echo "mimalloc pre-init regression: ELF-only check not applicable"
+    exit 0
+fi
+ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+probe_dir="$(mktemp -d "${TMPDIR:-/tmp}/cbm-mi-preinit.XXXXXX")"
+trap 'rm -rf "$probe_dir"' EXIT
+
+cat >"$probe_dir/probe.c" <<'C'
+#include <stddef.h>
+#include "mimalloc.h"
+static void before_init(void) { mi_free(NULL); }
+__attribute__((used, section(".preinit_array")))
+static void (*const preinit)(void) = before_init;
+int main(void) {
+    void *p = mi_malloc(16);
+    if (!p) return 1;
+    mi_free(p);
+    return 0;
+}
+C
+"${CC:-cc}" -O2 -DMI_DEBUG=0 -DMI_OVERRIDE=0 \
+    -I"$ROOT/vendored/mimalloc/include" -I"$ROOT/vendored/mimalloc/src" \
+    "$probe_dir/probe.c" "$ROOT/vendored/mimalloc/src/static.c" \
+    -lpthread -lm -o "$probe_dir/probe"
+"$probe_dir/probe"
+echo "mimalloc pre-init free(NULL) regression passed"

tests/test_smoke_fixture_contract.sh

@@ -118,7 +118,9 @@ CLIENT_HOME_OVERRIDES = (
 # so it needs the same neutralization — otherwise a green run on a developer
 # machine only proves the ambient config happened to be writable.
 test_main = read("tests/test_main.c")
-for variable in CLIENT_HOME_OVERRIDES:
+for variable in CLIENT_HOME_OVERRIDES + (
+    "XDG_CONFIG_HOME", "XDG_DATA_HOME", "XDG_CACHE_HOME"
+):
     require(
         f'"{variable}"' in test_main,
         f"tests/test_main.c must neutralize ambient {variable}",

tests/test_venue_parity_contract.sh

@@ -409,9 +409,9 @@ test-infrastructure/vm/vm-run-tests.sh
 test-infrastructure/vm/win.sh
 "
     for entry in $HELP_ENTRIES; do
-        out=$(cd "$ROOT" && bash "$entry" --help 2>&1) && rc=0 || rc=$?
         case "$entry" in
         *.py) out=$(cd "$ROOT" && python3 "$entry" --help 2>&1) && rc=0 || rc=$? ;;
+        *) out=$(cd "$ROOT" && bash "$entry" --help 2>&1) && rc=0 || rc=$? ;;
         esac
         # Substring test in-shell: piping into `grep -q` is unsafe under
         # `set -o pipefail` because grep exits at the first match and the

vendored/mimalloc/src/page-map.c

@@ -25,7 +25,10 @@ static void mi_page_map_cannot_commit(void) {
 // A full 256 TiB address space (48 bit) needs a 4 GiB page map.
 // A full 4 GiB address space (32 bit) needs only a 64 KiB page map.
 
-mi_decl_cache_align uint8_t* _mi_page_map = NULL;
+// Use an initial empty page map so `free(NULL)` works even if mimalloc is not yet initialized (issue #1341)
+static uint8_t mi_page_map_empty[1] = { 1 };      // _mi_ptr_page(NULL) == NULL
+
+mi_decl_cache_align uint8_t* _mi_page_map = mi_page_map_empty;
 static void*        mi_page_map_max_address = NULL;
 static mi_memid_t   mi_page_map_memid;
 
@@ -198,7 +201,11 @@ mi_decl_nodiscard mi_decl_export bool mi_is_in_heap_region(const void* p) mi_att
 #define MI_PAGE_MAP_SUB_SIZE          (MI_PAGE_MAP_SUB_COUNT * sizeof(mi_page_t*))
 #define MI_PAGE_MAP_ENTRIES_PER_CBIT  (MI_PAGE_MAP_COUNT < MI_BFIELD_BITS ? 1 : (MI_PAGE_MAP_COUNT / MI_BFIELD_BITS))
 
-mi_decl_cache_align _Atomic(mi_submap_t)* _mi_page_map;
+// Use an initial empty page map so `free(NULL)` works even if mimalloc is not yet initialized (issue #1341)
+static mi_page_t*            mi_submap_empty[1]   = { NULL };
+static _Atomic(mi_submap_t)  mi_page_map_empty[1] = { MI_ATOMIC_VAR_INIT(mi_submap_empty) };
+
+mi_decl_cache_align _Atomic(mi_submap_t)* _mi_page_map = mi_page_map_empty;
 static size_t       mi_page_map_count;
 static void*        mi_page_map_max_address;
 static mi_memid_t   mi_page_map_memid;
@@ -261,6 +268,7 @@ bool _mi_page_map_init(void) {
   _mi_page_map = (_Atomic(mi_page_t**)*)_mi_os_alloc_aligned(reserve_size, 1, commit, true /* allow large */, &mi_page_map_memid);
   if (_mi_page_map==NULL) {
     _mi_error_message(ENOMEM, "unable to reserve virtual memory for the page map (%zu KiB)\n", page_map_size / MI_KiB);
+    _mi_page_map = mi_page_map_empty;
     return false;
   }
   if (mi_page_map_memid.initially_committed && !mi_page_map_memid.initially_zero) {