[mypyc] Fix separate=True under free-threading (Py_GIL_DISABLED)

Two gaps in the preceding commit that only showed up on the py314t CI matrix (free-threading Python + compiled-mypy test harness): 1. The ensure_deps capsule call was only added to module_shim.tmpl, not to module_shim_no_gil_multiphase.tmpl. Free-threaded builds use the latter template, so cross-group exports tables were never populated on py314t and consumers called through NULL function pointers on the first cross-group invocation. Mirroring the non-GIL-disabled template fixes it. 2. The `_free_instance` slot (free-list for per-class fast allocation) was declared with needs_export=True, which puts it into the exports-table struct. Under Py_GIL_DISABLED, CPyThreadLocal expands to __thread, which can't legally appear inside a struct field (clang: "type name does not allow storage class to be specified"). The slot is only read/written by the class's own setup/dealloc code inside the defining group -- no cross-group access is needed -- so dropping needs_export is the right fix.
python · VaggelisD · Apr 23, 2026 · Apr 23, 2026 · Apr 23, 2026 · Apr 23, 2026
commit 0269b7a46afecc9cffdcf4661b45e065356de631
diff --git a/mypyc/codegen/emitclass.py b/mypyc/codegen/emitclass.py
@@ -228,9 +228,11 @@ def generate_class_reuse(
     context = c_emitter.context
     name = cl.name_prefix(c_emitter.names) + "_free_instance"
     struct_name = cl.struct_name(c_emitter.names)
-    context.declarations[name] = HeaderDeclaration(
-        f"CPyThreadLocal {struct_name} *{name};", needs_export=True
-    )
+    # Not exported: the free-instance slot is only read/written by the class's
+    # own setup/dealloc code, which lives in the defining group. Exporting it
+    # also trips a C diagnostic under `Py_GIL_DISABLED`, where `CPyThreadLocal`
+    # expands to `__thread` and can't legally appear inside the exports struct.
+    context.declarations[name] = HeaderDeclaration(f"CPyThreadLocal {struct_name} *{name};")
 
 
 def generate_class(cl: ClassIR, module: str, emitter: Emitter) -> None:

diff --git a/mypyc/lib-rt/module_shim_no_gil_multiphase.tmpl b/mypyc/lib-rt/module_shim_no_gil_multiphase.tmpl
@@ -3,7 +3,38 @@
 static int {modname}_exec(PyObject *module)
 {{
     PyObject *tmp;
-    if (!(tmp = PyImport_ImportModule("{libname}"))) return -1;
+    // Use ImportModuleLevel with a non-empty fromlist so Python returns the
+    // leaf submodule via sys.modules lookup, avoiding the top-down getattr
+    // walk done by PyImport_ImportModule. That walk fails when the shim runs
+    // during a parent package's __init__.py, because the parent's submodule
+    // attribute isn't set until after __init__ returns.
+    static PyObject *fromlist = NULL;
+    if (!fromlist) {{
+        fromlist = Py_BuildValue("(s)", "*");
+        if (!fromlist) return -1;
+    }}
+    tmp = PyImport_ImportModuleLevel("{libname}", NULL, NULL, fromlist, 0);
+    if (!tmp) return -1;
+    // Populate cross-group export tables lazily, just before we instantiate
+    // the per-module real init. Deferring this out of the shared lib's own
+    // PyInit keeps separate-mode compiled modules from recursively triggering
+    // sibling-package __init__.py mid-bootstrap.
+    PyObject *deps_capsule = PyObject_GetAttrString(tmp, "ensure_deps");
+    if (deps_capsule != NULL) {{
+        int (*deps_func)(void) = (int (*)(void))PyCapsule_GetPointer(
+            deps_capsule, "{libname}.ensure_deps");
+        Py_DECREF(deps_capsule);
+        if (deps_func == NULL) {{
+            Py_DECREF(tmp);
+            return -1;
+        }}
+        if (deps_func() < 0) {{
+            Py_DECREF(tmp);
+            return -1;
+        }}
+    }} else {{
+        PyErr_Clear();
+    }}
     PyObject *capsule = PyObject_GetAttrString(tmp, "exec_{full_modname}");
     Py_DECREF(tmp);
     if (capsule == NULL) return -1;