patch 9.0.1776: No support for stable Python 3 ABIv9.0.1776

Problem:  No support for stable Python 3 ABI
Solution: Support Python 3 stable ABI

Commits:
1) Support Python 3 stable ABI to allow mixed version interoperatbility

Vim currently supports embedding Python for use with plugins, and the
"dynamic" linking option allows the user to specify a locally installed
version of Python by setting `pythonthreedll`. However, one caveat is
that the Python 3 libs are not binary compatible across minor versions,
and mixing versions can potentially be dangerous (e.g. let's say Vim was
linked against the Python 3.10 SDK, but the user sets `pythonthreedll`
to a 3.11 lib). Usually, nothing bad happens, but in theory this could
lead to crashes, memory corruption, and other unpredictable behaviors.
It's also difficult for the user to tell something is wrong because Vim
has no way of reporting what Python 3 version Vim was linked with.

For Vim installed via a package manager, this usually isn't an issue
because all the dependencies would already be figured out. For prebuilt
Vim binaries like MacVim (my motivation for working on this), AppImage,
and Win32 installer this could potentially be an issue as usually a
single binary is distributed. This is more tricky when a new Python
version is released, as there's a chicken-and-egg issue with deciding
what Python version to build against and hard to keep in sync when a new
Python version just drops and we have a mix of users of different Python
versions, and a user just blindly upgrading to a new Python could lead to
bad interactions with Vim.

Python 3 does have a solution for this problem: stable ABI / limited API
(see https://docs.python.org/3/c-api/stable.html). The C SDK limits the
API to a set of functions that are promised to be stable across
versions. This pull request adds an ifdef config that allows us to turn
it on when building Vim. Vim binaries built with this option should be
safe to freely link with any Python 3 libraies without having the
constraint of having to use the same minor version.

Note: Python 2 has no such concept and this doesn't change how Python 2
integration works (not that there is going to be a new version of Python
2 that would cause compatibility issues in the future anyway).

---

Technical details:
======

The stable ABI can be accessed when we compile with the Python 3 limited
API (by defining `Py_LIMITED_API`). The Python 3 code (in `if_python3.c`
and `if_py_both.h`) would now handle this and switch to limited API
mode. Without it set, Vim will still use the full API as before so this
is an opt-in change.

The main difference is that `PyType_Object` is now an opaque struct that
we can't directly create "static types" out of, and we have to create
type objects as "heap types" instead. This is because the struct is not
stable and changes from version to version (e.g. 3.8 added a
`tp_vectorcall` field to it). I had to change all the types to be
allocated on the heap instead with just a pointer to them.

Other functions are also simply missing in limited API, or they are
introduced too late (e.g. `PyUnicode_AsUTF8AndSize` in 3.10) to it that
we need some other ways to do the same thing, so I had to abstract a few
things into macros, and sometimes re-implement functions like
`PyObject_NEW`.

One caveat is that in limited API, `OutputType` (used for replacing
`sys.stdout`) no longer inherits from `PyStdPrinter_Type` which I don't
think has any real issue other than minor differences in how they
convert to a string and missing a couple functions like `mode()` and
`fileno()`.

Also fixed an existing bug where `tp_basicsize` was set incorrectly for
`BufferObject`, `TabListObject, `WinListObject`.

Technically, there could be a small performance drop, there is a little
more indirection with accessing type objects, and some APIs like
`PyUnicode_AsUTF8AndSize` are missing, but in practice I didn't see any
difference, and any well-written Python plugin should try to avoid
excessing callbacks to the `vim` module in Python anyway.

I only tested limited API mode down to Python 3.7, which seemes to
compile and work fine. I haven't tried earlier Python versions.

2) Fix PyIter_Check on older Python vers / type##Ptr unused warning

For PyIter_Check, older versions exposed them as either macros (used in
full API), or a function (for use in limited API). A previous change
exposed PyIter_Check to the dynamic build because Python just moved it
to function-only in 3.10 anyway. Because of that, just make sure we
always grab the function in dynamic builds in earlier versions since
that's what Python eventually did anyway.

3) Move Py_LIMITED_API define to configure script

Can now use --with-python-stable-abi flag to customize what stable ABI
version to target. Can also use an env var to do so as well.

4) Show +python/dyn-stable in :version, and allow has() feature query

Not sure if the "/dyn-stable" suffix would break things, or whether we
should do it another way. Or just don't show it in version and rely on
has() feature checking.

5) Documentation first draft. Still need to implement v:python3_version

6) Fix PyIter_Check build breaks when compiling against Python 3.8

7) Add CI coverage stable ABI on Linux/Windows / make configurable on Windows

This adds configurable options for Windows make files (both MinGW and
MSVC). CI will also now exercise both traditional full API and stable
ABI for Linux and Windows in the matrix for coverage.

Also added a "dynamic" option to Linux matrix as a drive-by change to
make other scripting languages like Ruby / Perl testable under both
static and dynamic builds.

8) Fix inaccuracy in Windows docs

Python's own docs are confusing but you don't actually want to use
`python3.dll` for the dynamic linkage.

9) Add generated autoconf file

10) Add v:python3_version support

This variable indicates the version of Python3 that Vim was built
against (PY_VERSION_HEX), and will be useful to check whether the Python
library you are loading in dynamically actually fits it. When built with
stable ABI, it will be the limited ABI version instead
(`Py_LIMITED_API`), which indicates the minimum version of Python 3 the
user should have, rather than the exact match. When stable ABI is used,
we won't be exposing PY_VERSION_HEX in this var because it just doesn't
seem necessary to do so (the whole point of stable ABI is the promise
that it will work across versions), and I don't want to confuse the user
with too many variables.

Also, cleaned up some documentation, and added help tags.

11) Fix Python 3.7 compat issues

Fix a couple issues when using limited API < 3.8

- Crash on exit: In Python 3.7, if a heap-allocated type is destroyed
  before all instances are, it would cause a crash later. This happens
  when we destroyed `OptionsType` before calling `Py_Finalize` when
  using the limited API. To make it worse, later versions changed the
  semantics and now each instance has a strong reference to its own type
  and the recommendation has changed to have each instance de-ref its
  own type and have its type in GC traversal. To avoid dealing with
  these cross-version variations, we just don't free the heap type. They
  are static types in non-limited-API anyway and are designed to last
  through the entirety of the app, and we also don't restart the Python
  runtime and therefore do not need it to have absolutely 0 leaks.

  See:
  - https://docs.python.org/3/whatsnew/3.8.html#changes-in-the-c-api
  - https://docs.python.org/3/whatsnew/3.9.html#changes-in-the-c-api

- PyIter_Check: This function is not provided in limited APIs older than
  3.8. Previously I was trying to mock it out using manual
  PyType_GetSlot() but it was brittle and also does not actually work
  properly for static types (it will generate a Python error). Just
  return false. It does mean using limited API < 3.8 is not recommended
  as you lose the functionality to handle iterators, but from playing
  with plugins I couldn't find it to be an issue.

- Fix loading of PyIter_Check so it will be done when limited API < 3.8.
  Otherwise loading a 3.7 Python lib will fail even if limited API was
  specified to use it.

12) Make sure to only load `PyUnicode_AsUTF8AndSize` in needed in limited API

We don't use this function unless limited API >= 3.10, but we were
loading it regardless. Usually it's ok in Unix-like systems where Python
just has a single lib that we load from, but in Windows where there is a
separate python3.dll this would not work as the symbol would not have
been exposed in this more limited DLL file. This makes it much clearer
under what condition is this function needed.

closes: #12032

Signed-off-by: Christian Brabandt <cb@256bit.org>
Co-authored-by: Yee Cheng Chin <ychin.git@gmail.com>

1 files changed, 431 insertions, 118 deletions

diff --git a/src/if_py_both.h b/src/if_py_both.h
index 110de234fd..ff18098605 100644
--- a/src/if_py_both.h
+++ b/src/if_py_both.h
@@ -30,9 +30,285 @@ static const char *vim_special_path = "_vim_path_";
 #define PyErr_FORMAT2(exc, str, arg1, arg2) PyErr_Format(exc, _(str), arg1,arg2)
 #define PyErr_VIM_FORMAT(str, arg) PyErr_FORMAT(VimError, str, arg)
 
-#define Py_TYPE_NAME(obj) ((obj)->ob_type->tp_name == NULL \
+#ifdef USE_LIMITED_API
+// Limited Python API. Need to call only exposed functions and remap macros.
+// PyTypeObject is an opaque struct.
+
+typedef struct {
+    lenfunc sq_length;
+    binaryfunc sq_concat;
+    ssizeargfunc sq_repeat;
+    ssizeargfunc sq_item;
+    void *was_sq_slice;
+    ssizeobjargproc sq_ass_item;
+    void *was_sq_ass_slice;
+    objobjproc sq_contains;
+
+    binaryfunc sq_inplace_concat;
+    ssizeargfunc sq_inplace_repeat;
+} PySequenceMethods;
+
+typedef struct {
+    lenfunc mp_length;
+    binaryfunc mp_subscript;
+    objobjargproc mp_ass_subscript;
+} PyMappingMethods;
+
+// This struct emulates the concrete _typeobject struct to allow the code to
+// work the same way in both limited and full Python APIs.
+struct typeobject_wrapper {
+    const char *tp_name;
+    Py_ssize_t tp_basicsize;
+    unsigned long tp_flags;
+
+    // When adding new slots below, also need to make sure we add ADD_TP_SLOT
+    // call in AddHeapType for it.
+
+    destructor tp_dealloc;
+    reprfunc tp_repr;
+
+    PySequenceMethods *tp_as_sequence;
+    PyMappingMethods *tp_as_mapping;
+
+    ternaryfunc tp_call;
+    getattrofunc tp_getattro;
+    setattrofunc tp_setattro;
+
+    const char *tp_doc;
+
+    traverseproc tp_traverse;
+
+    inquiry tp_clear;
+
+    getiterfunc tp_iter;
+    iternextfunc tp_iternext;
+
+    struct PyMethodDef *tp_methods;
+    struct _typeobject *tp_base;
+    allocfunc tp_alloc;
+    newfunc tp_new;
+    freefunc tp_free;
+};
+
+# define DEFINE_PY_TYPE_OBJECT(type) \
+    static struct typeobject_wrapper type; \
+    static PyTypeObject* type##Ptr = NULL
+
+// PyObject_HEAD_INIT_TYPE and PyObject_FINISH_INIT_TYPE need to come in pairs
+// We first initialize with NULL because the type is not allocated until
+// init_types() is called later. It's in FINISH_INIT_TYPE where we fill the
+// type in with the newly allocated type.
+# define PyObject_HEAD_INIT_TYPE(type) PyObject_HEAD_INIT(NULL)
+# define PyObject_FINISH_INIT_TYPE(obj, type) obj.ob_base.ob_type = type##Ptr
+
+# define Py_TYPE_GET_TP_ALLOC(type) ((allocfunc)PyType_GetSlot(type, Py_tp_alloc))
+# define Py_TYPE_GET_TP_METHODS(type) ((PyMethodDef *)PyType_GetSlot(type, Py_tp_methods))
+
+// PyObject_NEW is not part of stable ABI, but PyObject_Malloc/Init are.
+PyObject* Vim_PyObject_New(PyTypeObject *type, size_t objsize)
+{
+    PyObject *obj = (PyObject *)PyObject_Malloc(objsize);
+    if (obj == NULL)
+        return PyErr_NoMemory();
+    return PyObject_Init(obj, type);
+}
+# undef PyObject_NEW
+# define PyObject_NEW(type, typeobj) ((type *)Vim_PyObject_New(typeobj, sizeof(type)))
+
+// This is a somewhat convoluted because limited API doesn't expose an easy way
+// to get the tp_name field, and so we have to manually reconstruct it as
+// "__module__.__name__" (with __module__ omitted for builtins to emulate
+// Python behavior). Also, some of the more convenient functions like
+// PyUnicode_AsUTF8AndSize and PyType_GetQualName() are not available until
+// late Python 3 versions, and won't be available if you set Py_LIMITED_API too
+// low.
+# define PyErr_FORMAT_TYPE(msg, obj) \
+    do { \
+	PyObject* qualname = PyObject_GetAttrString((PyObject*)(obj)->ob_type, "__qualname__"); \
+	if (qualname == NULL) \
+	{ \
+	    PyErr_FORMAT(PyExc_TypeError, msg, "(NULL)"); \
+	    break; \
+	} \
+	PyObject* module = PyObject_GetAttrString((PyObject*)(obj)->ob_type, "__module__"); \
+	PyObject* full; \
+	if (module == NULL || PyUnicode_CompareWithASCIIString(module, "builtins") == 0 \
+		|| PyUnicode_CompareWithASCIIString(module, "__main__") == 0) \
+	{ \
+	    full = qualname; \
+	    Py_INCREF(full); \
+	} \
+	else \
+	    full = PyUnicode_FromFormat("%U.%U", module, qualname); \
+	PyObject* full_bytes = PyUnicode_AsUTF8String(full); \
+	const char* full_str = PyBytes_AsString(full_bytes); \
+	full_str = full_str == NULL ? "(NULL)" : full_str; \
+	PyErr_FORMAT(PyExc_TypeError, msg, full_str); \
+	Py_DECREF(qualname); \
+	Py_XDECREF(module); \
+	Py_XDECREF(full); \
+	Py_XDECREF(full_bytes); \
+    } while(0)
+
+# define PyList_GET_ITEM(list, i) PyList_GetItem(list, i)
+# define PyList_GET_SIZE(o) PyList_Size(o)
+# define PyTuple_GET_ITEM(o, pos) PyTuple_GetItem(o, pos)
+# define PyTuple_GET_SIZE(o) PyTuple_Size(o)
+
+// PyList_SET_ITEM and PyList_SetItem have slightly different behaviors. The
+// former will leave the old item dangling, and the latter will decref on it.
+// Since we only use this on new lists, this difference doesn't matter.
+# define PyList_SET_ITEM(list, i, item) PyList_SetItem(list, i, item)
+
+# if Py_LIMITED_API < 0x03080000
+// PyIter_check only became part of stable ABI in 3.8, and there is no easy way
+// to check for it in the API. We simply return false as a compromise. This
+// does mean we should avoid compiling with stable ABI < 3.8.
+#  undef PyIter_Check
+#  define PyIter_Check(obj) (FALSE)
+# endif
+
+PyTypeObject* AddHeapType(struct typeobject_wrapper* type_object)
+{
+    PyType_Spec type_spec;
+    type_spec.name = type_object->tp_name;
+    type_spec.basicsize = type_object->tp_basicsize;
+    type_spec.itemsize = 0;
+    type_spec.flags = type_object->tp_flags;
+
+    // We just need to statically allocate a large enough buffer that can hold
+    // all slots. We need to leave a null-terminated slot at the end.
+    PyType_Slot slots[40] = { {0, NULL} };
+    size_t slot_i = 0;
+
+# define ADD_TP_SLOT(slot_name) \
+    if (slot_i >= 40) return NULL; /* this should never happen */ \
+    if (type_object->slot_name != NULL) \
+    { \
+	slots[slot_i].slot = Py_##slot_name; \
+	slots[slot_i].pfunc = (void*)type_object->slot_name; \
+	++slot_i; \
+    }
+# define ADD_TP_SUB_SLOT(sub_slot, slot_name) \
+    if (slot_i >= 40) return NULL; /* this should never happen */ \
+    if (type_object->sub_slot != NULL && type_object->sub_slot->slot_name != NULL) \
+    { \
+	slots[slot_i].slot = Py_##slot_name; \
+	slots[slot_i].pfunc = (void*)type_object->sub_slot->slot_name; \
+	++slot_i; \
+    }
+
+    ADD_TP_SLOT(tp_dealloc)
+    ADD_TP_SLOT(tp_repr)
+    ADD_TP_SLOT(tp_call)
+    ADD_TP_SLOT(tp_getattro)
+    ADD_TP_SLOT(tp_setattro)
+    ADD_TP_SLOT(tp_doc)
+    ADD_TP_SLOT(tp_traverse)
+    ADD_TP_SLOT(tp_clear)
+    ADD_TP_SLOT(tp_iter)
+    ADD_TP_SLOT(tp_iternext)
+    ADD_TP_SLOT(tp_methods)
+    ADD_TP_SLOT(tp_base)
+    ADD_TP_SLOT(tp_alloc)
+    ADD_TP_SLOT(tp_new)
+    ADD_TP_SLOT(tp_free)
+
+    ADD_TP_SUB_SLOT(tp_as_sequence, sq_length)
+    ADD_TP_SUB_SLOT(tp_as_sequence, sq_concat)
+    ADD_TP_SUB_SLOT(tp_as_sequence, sq_repeat)
+    ADD_TP_SUB_SLOT(tp_as_sequence, sq_item)
+    ADD_TP_SUB_SLOT(tp_as_sequence, sq_ass_item)
+    ADD_TP_SUB_SLOT(tp_as_sequence, sq_contains)
+    ADD_TP_SUB_SLOT(tp_as_sequence, sq_inplace_concat)
+    ADD_TP_SUB_SLOT(tp_as_sequence, sq_inplace_repeat)
+
+    ADD_TP_SUB_SLOT(tp_as_mapping, mp_length)
+    ADD_TP_SUB_SLOT(tp_as_mapping, mp_subscript)
+    ADD_TP_SUB_SLOT(tp_as_mapping, mp_ass_subscript)
+# undef ADD_TP_SLOT
+# undef ADD_TP_SUB_SLOT
+
+    type_spec.slots = slots;
+
+    PyObject* newtype = PyType_FromSpec(&type_spec);
+    return (PyTypeObject*)newtype;
+}
+
+// Add a heap type, since static types do not work in limited API
+// Each PYTYPE_READY is paired with PYTYPE_CLEANUP.
+//
+// Note that we don't call Py_DECREF(type##Ptr) in clean up. The reason for
+// that in 3.7, it's possible to de-allocate a heap type before all instances
+// are cleared, leading to a crash, whereas in 3.8 the semantics were changed
+// and instances hold strong references to types. Since these types are
+// designed to be static, just keep them around to avoid having to write
+// version-specific handling. Vim does not re-start the Python runtime so there
+// will be no long-term leak.
+# define PYTYPE_READY(type) \
+    type##Ptr = AddHeapType(&(type)); \
+    if (type##Ptr == NULL) \
+	return -1;
+# define PYTYPE_CLEANUP(type) \
+    type##Ptr = NULL;
+
+// Limited API does not provide PyRun_* functions. Need to implement manually
+// using PyCompile and PyEval.
+PyObject* Vim_PyRun_String(const char *str, int start, PyObject *globals, PyObject *locals)
+{
+    // Just pass "" for filename for now.
+    PyObject* compiled = Py_CompileString(str, "", start);
+    if (compiled == NULL)
+	return NULL;
+
+    PyObject* eval_result = PyEval_EvalCode(compiled, globals, locals);
+    Py_DECREF(compiled);
+    return eval_result;
+}
+int Vim_PyRun_SimpleString(const char *str)
+{
+    // This function emulates CPython's implementation.
+    PyObject* m = PyImport_AddModule("__main__");
+    if (m == NULL)
+        return -1;
+    PyObject* d = PyModule_GetDict(m);
+    PyObject* output = Vim_PyRun_String(str, Py_file_input, d, d);
+    if (output == NULL)
+    {
+	PyErr_PrintEx(TRUE);
+	return -1;
+    }
+    Py_DECREF(output);
+    return 0;
+}
+#define PyRun_String Vim_PyRun_String
+#define PyRun_SimpleString Vim_PyRun_SimpleString
+
+#else // !defined(USE_LIMITED_API)
+
+// Full Python API. Can make use of structs and macros directly.
+# define DEFINE_PY_TYPE_OBJECT(type) \
+    static PyTypeObject type; \
+    static PyTypeObject* type##Ptr = &type
+# define PyObject_HEAD_INIT_TYPE(type) PyObject_HEAD_INIT(&type)
+
+# define Py_TYPE_GET_TP_ALLOC(type) type->tp_alloc
+# define Py_TYPE_GET_TP_METHODS(type) type->tp_methods
+
+# define Py_TYPE_NAME(obj) ((obj)->ob_type->tp_name == NULL \
 	? "(NULL)" \
 	: (obj)->ob_type->tp_name)
+# define PyErr_FORMAT_TYPE(msg, obj) \
+    PyErr_FORMAT(PyExc_TypeError, msg, \
+	    Py_TYPE_NAME(obj))
+
+// Add a static type
+# define PYTYPE_READY(type) \
+    if (PyType_Ready(type##Ptr)) \
+	return -1;
+
+#endif
+
 
 #define RAISE_NO_EMPTY_KEYS PyErr_SET_STRING(PyExc_ValueError, \
 					    N_("empty keys are not allowed"))
@@ -45,8 +321,7 @@ static const char *vim_special_path = "_vim_path_";
 #define RAISE_KEY_ADD_FAIL(key) \
     PyErr_VIM_FORMAT(N_("failed to add key '%s' to dictionary"), key)
 #define RAISE_INVALID_INDEX_TYPE(idx) \
-    PyErr_FORMAT(PyExc_TypeError, N_("index must be int or slice, not %s"), \
-	    Py_TYPE_NAME(idx));
+    PyErr_FORMAT_TYPE(N_("index must be int or slice, not %s"), idx);
 
 #define INVALID_BUFFER_VALUE ((buf_T *)(-1))
 #define INVALID_WINDOW_VALUE ((win_T *)(-1))
@@ -144,13 +419,11 @@ StringToChars(PyObject *obj, PyObject **todecref)
     else
     {
 #if PY_MAJOR_VERSION < 3
-	PyErr_FORMAT(PyExc_TypeError,
-		N_("expected str() or unicode() instance, but got %s"),
-		Py_TYPE_NAME(obj));
+	PyErr_FORMAT_TYPE(N_("expected str() or unicode() instance, but got %s"),
+		obj);
 #else
-	PyErr_FORMAT(PyExc_TypeError,
-		N_("expected bytes() or str() instance, but got %s"),
-		Py_TYPE_NAME(obj));
+	PyErr_FORMAT_TYPE(N_("expected bytes() or str() instance, but got %s"),
+		obj);
 #endif
 	return NULL;
     }
@@ -198,15 +471,15 @@ NumberToLong(PyObject *obj, long *result, int flags)
     else
     {
 #if PY_MAJOR_VERSION < 3
-	PyErr_FORMAT(PyExc_TypeError,
+	PyErr_FORMAT_TYPE(
 		N_("expected int(), long() or something supporting "
 		   "coercing to long(), but got %s"),
-		Py_TYPE_NAME(obj));
+		obj);
 #else
-	PyErr_FORMAT(PyExc_TypeError,
+	PyErr_FORMAT_TYPE(
 		N_("expected int() or something supporting coercing to int(), "
 		   "but got %s"),
-		Py_TYPE_NAME(obj));
+		obj);
 #endif
 	return -1;
     }
@@ -278,7 +551,7 @@ ObjectDir(PyObject *self, char **attributes)
 	return NULL;
 
     if (self)
-	for (method = self->ob_type->tp_methods ; method->ml_name != NULL ; ++method)
+	for (method = Py_TYPE_GET_TP_METHODS(self->ob_type) ; method->ml_name != NULL ; ++method)
 	    if (add_string(ret, (char *)method->ml_name))
 	    {
 		Py_DECREF(ret);
@@ -308,7 +581,7 @@ ObjectDir(PyObject *self, char **attributes)
 // Function to write a line, points to either msg() or emsg().
 typedef int (*writefn)(char *);
 
-static PyTypeObject OutputType;
+DEFINE_PY_TYPE_OBJECT(OutputType);
 
 typedef struct
 {
@@ -514,14 +787,14 @@ static struct PyMethodDef OutputMethods[] = {
 
 static OutputObject Output =
 {
-    PyObject_HEAD_INIT(&OutputType)
+    PyObject_HEAD_INIT_TYPE(OutputType)
     0,
     0
 };
 
 static OutputObject Error =
 {
-    PyObject_HEAD_INIT(&OutputType)
+    PyObject_HEAD_INIT_TYPE(OutputType)
     0,
     1
 };
@@ -552,7 +825,7 @@ typedef struct
     char	*fullname;
     PyObject	*result;
 } LoaderObject;
-static PyTypeObject LoaderType;
+DEFINE_PY_TYPE_OBJECT(LoaderType);
 
     static void
 LoaderDestructor(LoaderObject *self)
@@ -1243,9 +1516,9 @@ call_load_module(char *name, int len, PyObject *find_module_result)
 
     if (!PyTuple_Check(find_module_result))
     {
-	PyErr_FORMAT(PyExc_TypeError,
+	PyErr_FORMAT_TYPE(
 		N_("expected 3-tuple as imp.find_module() result, but got %s"),
-		Py_TYPE_NAME(find_module_result));
+		find_module_result);
 	return NULL;
     }
     if (PyTuple_GET_SIZE(find_module_result) != 3)
@@ -1367,7 +1640,7 @@ FinderFindModule(PyObject *self, PyObject *args)
 	return NULL;
     }
 
-    if (!(loader = PyObject_NEW(LoaderObject, &LoaderType)))
+    if (!(loader = PyObject_NEW(LoaderObject, LoaderTypePtr)))
     {
 	vim_free(fullname);
 	Py_DECREF(result);
@@ -1424,7 +1697,7 @@ static struct PyMethodDef VimMethods[] = {
  * Generic iterator object
  */
 
-static PyTypeObject IterType;
+DEFINE_PY_TYPE_OBJECT(IterType);
 
 typedef PyObject *(*nextfun)(void **);
 typedef void (*destructorfun)(void *);
@@ -1451,7 +1724,7 @@ IterNew(void *start, destructorfun destruct, nextfun next, traversefun traverse,
 {
     IterObject *self;
 
-    self = PyObject_GC_New(IterObject, &IterType);
+    self = PyObject_GC_New(IterObject, IterTypePtr);
     self->cur = start;
     self->next = next;
     self->destruct = destruct;
@@ -1556,7 +1829,7 @@ pyll_add(PyObject *self, pylinkedlist_T *ref, pylinkedlist_T **last)
     *last = ref;
 }
 
-static PyTypeObject DictionaryType;
+DEFINE_PY_TYPE_OBJECT(DictionaryType);
 
 typedef struct
 {
@@ -1567,14 +1840,14 @@ typedef struct
 
 static PyObject *DictionaryUpdate(DictionaryObject *, PyObject *, PyObject *);
 
-#define NEW_DICTIONARY(dict) DictionaryNew(&DictionaryType, dict)
+#define NEW_DICTIONARY(dict) DictionaryNew(DictionaryTypePtr, dict)
 
     static PyObject *
 DictionaryNew(PyTypeObject *subtype, dict_T *dict)
 {
     DictionaryObject	*self;
 
-    self = (DictionaryObject *) subtype->tp_alloc(subtype, 0);
+    self = (DictionaryObject *) Py_TYPE_GET_TP_ALLOC(subtype)(subtype, 0);
     if (self == NULL)
 	return NULL;
     self->dict = dict;
@@ -2238,7 +2511,7 @@ static struct PyMethodDef DictionaryMethods[] = {
     { NULL,	NULL,					0,		NULL}
 };
 
-static PyTypeObject ListType;
+DEFINE_PY_TYPE_OBJECT(ListType);
 
 typedef struct
 {
@@ -2247,7 +2520,7 @@ typedef struct
     pylinkedlist_T	ref;
 } ListObject;
 
-#define NEW_LIST(list) ListNew(&ListType, list)
+#define NEW_LIST(list) ListNew(ListTypePtr, list)
 
     static PyObject *
 ListNew(PyTypeObject *subtype, list_T *list)
@@ -2257,7 +2530,7 @@ ListNew(PyTypeObject *subtype, list_T *list)
     if (list == NULL)
 	return NULL;
 
-    self = (ListObject *) subtype->tp_alloc(subtype, 0);
+    self = (ListObject *) Py_TYPE_GET_TP_ALLOC(subtype)(subtype, 0);
     if (self == NULL)
 	return NULL;
     self->list = list;
@@ -2937,10 +3210,10 @@ typedef struct
     int		auto_rebind;
 } FunctionObject;
 
-static PyTypeObject FunctionType;
+DEFINE_PY_TYPE_OBJECT(FunctionType);
 
 #define NEW_FUNCTION(name, argc, argv, self, pt_auto) \
-    FunctionNew(&FunctionType, (name), (argc), (argv), (self), (pt_auto))
+    FunctionNew(FunctionTypePtr, (name), (argc), (argv), (self), (pt_auto))
 
     static PyObject *
 FunctionNew(PyTypeObject *subtype, char_u *name, int argc, typval_T *argv,
@@ -2948,7 +3221,7 @@ FunctionNew(PyTypeObject *subtype, char_u *name, int argc, typval_T *argv,
 {
     FunctionObject	*self;
 
-    self = (FunctionObject *)subtype->tp_alloc(subtype, 0);
+    self = (FunctionObject *) Py_TYPE_GET_TP_ALLOC(subtype)(subtype, 0);
     if (self == NULL)
 	return NULL;
 
@@ -3311,7 +3584,7 @@ static struct PyMethodDef FunctionMethods[] = {
  * Options object
  */
 
-static PyTypeObject OptionsType;
+DEFINE_PY_TYPE_OBJECT(OptionsType);
 
 typedef int (*checkfun)(void *);
 
@@ -3335,7 +3608,7 @@ OptionsNew(int opt_type, void *from, checkfun Check, PyObject *fromObj)
 {
     OptionsObject	*self;
 
-    self = PyObject_GC_New(OptionsObject, &OptionsType);
+    self = PyObject_GC_New(OptionsObject, OptionsTypePtr);
     if (self == NULL)
 	return NULL;
 
@@ -3692,7 +3965,7 @@ typedef struct
 
 static PyObject *WinListNew(TabPageObject *tabObject);
 
-static PyTypeObject TabPageType;
+DEFINE_PY_TYPE_OBJECT(TabPageType);
 
     static int
 CheckTabPage(TabPageObject *self)
@@ -3718,7 +3991,7 @@ TabPageNew(tabpage_T *tab)
     }
     else
     {
-	self = PyObject_NEW(TabPageObject, &TabPageType);
+	self = PyObject_NEW(TabPageObject, TabPageTypePtr);
 	if (self == NULL)
 	    return NULL;
 	self->tab = tab;
@@ -3810,7 +4083,7 @@ static struct PyMethodDef TabPageMethods[] = {
  * Window list object
  */
 
-static PyTypeObject TabListType;
+DEFINE_PY_TYPE_OBJECT(TabListType);
 static PySequenceMethods TabListAsSeq;
 
 typedef struct
@@ -3818,6 +4091,11 @@ typedef struct
     PyObject_HEAD
 } TabListObject;
 
+static TabListObject TheTabPageList =
+{
+    PyObject_HEAD_INIT_TYPE(TabListType)
+};
+
     static PyInt
 TabListLength(PyObject *self UNUSED)
 {
@@ -3857,7 +4135,7 @@ typedef struct
     TabPageObject	*tabObject;
 } WindowObject;
 
-static PyTypeObject WindowType;
+DEFINE_PY_TYPE_OBJECT(WindowType);
 
     static int
 CheckWindow(WindowObject *self)
@@ -3899,7 +4177,7 @@ WindowNew(win_T *win, tabpage_T *tab)
     }
     else
     {
-	self = PyObject_GC_New(WindowObject, &WindowType);
+	self = PyObject_GC_New(WindowObject, WindowTypePtr);
 	if (self == NULL)
 	    return NULL;
 	self->win = win;
@@ -4150,7 +4428,7 @@ static struct PyMethodDef WindowMethods[] = {
  * Window list object
  */
 
-static PyTypeObject WinListType;
+DEFINE_PY_TYPE_OBJECT(WinListType);
 static PySequenceMethods WinListAsSeq;
 
 typedef struct
@@ -4159,12 +4437,18 @@ typedef struct
     TabPageObject	*tabObject;
 } WinListObject;
 
+static WinListObject TheWindowList =
+{
+    PyObject_HEAD_INIT_TYPE(WinListType)
+    NULL
+};
+
     static PyObject *
 WinListNew(TabPageObject *tabObject)
 {
     WinListObject	*self;
 
-    self = PyObject_NEW(WinListObject, &WinListType);
+    self = PyObject_NEW(WinListObject, WinListTypePtr);
     self->tabObject = tabObject;
     Py_INCREF(tabObject);
 
@@ -4259,13 +4543,13 @@ StringToLine(PyObject *obj)
     else
     {
 #if PY_MAJOR_VERSION < 3
-	PyErr_FORMAT(PyExc_TypeError,
+	PyErr_FORMAT_TYPE(
 		N_("expected str() or unicode() instance, but got %s"),
-		Py_TYPE_NAME(obj));
+		obj);
 #else
-	PyErr_FORMAT(PyExc_TypeError,
+	PyErr_FORMAT_TYPE(
 		N_("expected bytes() or str() instance, but got %s"),
-		Py_TYPE_NAME(obj));
+		obj);
 #endif
 	return NULL;
     }
@@ -5028,7 +5312,7 @@ RBAppend(
 
 // Range object
 
-static PyTypeObject RangeType;
+DEFINE_PY_TYPE_OBJECT(RangeType);
 static PySequenceMethods RangeAsSeq;
 static PyMappingMethods RangeAsMapping;
 
@@ -5045,7 +5329,7 @@ RangeNew(buf_T *buf, PyInt start, PyInt end)
 {
     BufferObject *bufr;
     RangeObject *self;
-    self = PyObject_GC_New(RangeObject, &RangeType);
+    self = PyObject_GC_New(RangeObject, RangeTypePtr);
     if (self == NULL)
 	return NULL;
 
@@ -5150,7 +5434,7 @@ static struct PyMethodDef RangeMethods[] = {
     { NULL,	NULL,				0,		NULL}
 };
 
-static PyTypeObject BufferType;
+DEFINE_PY_TYPE_OBJECT(BufferType);
 static PySequenceMethods BufferAsSeq;
 static PyMappingMethods BufferAsMapping;
 
@@ -5184,7 +5468,7 @@ BufferNew(buf_T *buf)
     }
     else
     {
-	self = PyObject_NEW(BufferObject, &BufferType);
+	self = PyObject_NEW(BufferObject, BufferTypePtr);
 	if (self == NULL)
 	    return NULL;
 	self->buf = buf;
@@ -5410,13 +5694,18 @@ static struct PyMethodDef BufferMethods[] = {
  * Buffer list object - Implementation
  */
 
-static PyTypeObject BufMapType;
+DEFINE_PY_TYPE_OBJECT(BufMapType);
 
 typedef struct
 {
     PyObject_HEAD
 } BufMapObject;
 
+static BufMapObject TheBufferMap =
+{
+    PyObject_HEAD_INIT_TYPE(BufMapType)
+};
+
     static PyInt
 BufMapLength(PyObject *self UNUSED)
 {
@@ -5574,11 +5863,11 @@ CurrentSetattr(PyObject *self UNUSED, char *name, PyObject *valObject)
     {
 	int count;
 
-	if (valObject->ob_type != &BufferType)
+	if (valObject->ob_type != BufferTypePtr)
 	{
-	    PyErr_FORMAT(PyExc_TypeError,
+	    PyErr_FORMAT_TYPE(
 		    N_("expected vim.Buffer object, but got %s"),
-		    Py_TYPE_NAME(valObject));
+		    valObject);
 	    return -1;
 	}
 
@@ -5601,11 +5890,11 @@ CurrentSetattr(PyObject *self UNUSED, char *name, PyObject *valObject)
     {
 	int count;
 
-	if (valObject->ob_type != &WindowType)
+	if (valObject->ob_type != WindowTypePtr)
 	{
-	    PyErr_FORMAT(PyExc_TypeError,
+	    PyErr_FORMAT_TYPE(
 		    N_("expected vim.Window object, but got %s"),
-		    Py_TYPE_NAME(valObject));
+		    valObject);
 	    return -1;
 	}
 
@@ -5635,11 +5924,11 @@ CurrentSetattr(PyObject *self UNUSED, char *name, PyObject *valObject)
     }
     else if (strcmp(name, "tabpage") == 0)
     {
-	if (valObject->ob_type != &TabPageType)
+	if (valObject->ob_type != TabPageTypePtr)
 	{
-	    PyErr_FORMAT(PyExc_TypeError,
+	    PyErr_FORMAT_TYPE(
 		    N_("expected vim.TabPage object, but got %s"),
-		    Py_TYPE_NAME(valObject));
+		    valObject);
 	    return -1;
 	}
 
@@ -6180,7 +6469,7 @@ ConvertFromPyMapping(PyObject *obj, typval_T *tv)
     if (!(lookup_dict = PyDict_New()))
 	return -1;
 
-    if (PyType_IsSubtype(obj->ob_type, &DictionaryType))
+    if (PyType_IsSubtype(obj->ob_type, DictionaryTypePtr))
     {
 	tv->v_type = VAR_DICT;
 	tv->vval.v_dict = (((DictionaryObject *)(obj))->dict);
@@ -6193,9 +6482,9 @@ ConvertFromPyMapping(PyObject *obj, typval_T *tv)
 	ret = convert_dl(obj, tv, pymap_to_tv, lookup_dict);
     else
     {
-	PyErr_FORMAT(PyExc_TypeError,
+	PyErr_FORMAT_TYPE(
 		N_("unable to convert %s to a Vim dictionary"),
-		Py_TYPE_NAME(obj));
+		obj);
 	ret = -1;
     }
     Py_DECREF(lookup_dict);
@@ -6211,7 +6500,7 @@ ConvertFromPySequence(PyObject *obj, typval_T *tv)
     if (!(lookup_dict = PyDict_New()))
 	return -1;
 
-    if (PyType_IsSubtype(obj->ob_type, &ListType))
+    if (PyType_IsSubtype(obj->ob_type, ListTypePtr))
     {
 	tv->v_type = VAR_LIST;
 	tv->vval.v_list = (((ListObject *)(obj))->list);
@@ -6222,9 +6511,9 @@ ConvertFromPySequence(PyObject *obj, typval_T *tv)
 	ret = convert_dl(obj, tv, pyseq_to_tv, lookup_dict);
     else
     {
-	PyErr_FORMAT(PyExc_TypeError,
+	PyErr_FORMAT_TYPE(
 		N_("unable to convert %s to a Vim list"),
-		Py_TYPE_NAME(obj));
+		obj);
 	ret = -1;
     }
     Py_DECREF(lookup_dict);
@@ -6247,19 +6536,19 @@ ConvertFromPyObject(PyObject *obj, typval_T *tv)
     static int
 _ConvertFromPyObject(PyObject *obj, typval_T *tv, PyObject *lookup_dict)
 {
-    if (PyType_IsSubtype(obj->ob_type, &DictionaryType))
+    if (PyType_IsSubtype(obj->ob_type, DictionaryTypePtr))
     {
 	tv->v_type = VAR_DICT;
 	tv->vval.v_dict = (((DictionaryObject *)(obj))->dict);
 	++tv->vval.v_dict->dv_refcount;
     }
-    else if (PyType_IsSubtype(obj->ob_type, &ListType))
+    else if (PyType_IsSubtype(obj->ob_type, ListTypePtr))
     {
 	tv->v_type = VAR_LIST;
 	tv->vval.v_list = (((ListObject *)(obj))->list);
 	++tv->vval.v_list->lv_refcount;
     }
-    else if (PyType_IsSubtype(obj->ob_type, &FunctionType))
+    else if (PyType_IsSubtype(obj->ob_type, FunctionTypePtr))
     {
 	FunctionObject *func = (FunctionObject *) obj;
 	if (func->self != NULL || func->argv != NULL)
@@ -6365,9 +6654,9 @@ _ConvertFromPyObject(PyObject *obj, typval_T *tv, PyObject *lookup_dict)
     }
     else
     {
-	PyErr_FORMAT(PyExc_TypeError,
+	PyErr_FORMAT_TYPE(
 		N_("unable to convert %s to a Vim structure"),
-		Py_TYPE_NAME(obj));
+		obj);
 	return -1;
     }
     return 0;
@@ -6445,11 +6734,17 @@ ConvertToPyObject(typval_T *tv)
     return NULL;
 }
 
+DEFINE_PY_TYPE_OBJECT(CurrentType);
+
 typedef struct
 {
     PyObject_HEAD
 } CurrentObject;
-static PyTypeObject CurrentType;
+
+static CurrentObject TheCurrent =
+{
+    PyObject_HEAD_INIT_TYPE(CurrentType)
+};
 
     static void
 init_structs(void)
@@ -6466,7 +6761,11 @@ init_structs(void)
     OutputType.tp_alloc = call_PyType_GenericAlloc;
     OutputType.tp_new = call_PyType_GenericNew;
     OutputType.tp_free = call_PyObject_Free;
+# ifndef USE_LIMITED_API
+    // The std printer type is only exposed in full API. It is not essential
+    // anyway and so in limited API we don't set it.
     OutputType.tp_base = &PyStdPrinter_Type;
+# endif
 #else
     OutputType.tp_getattr = (getattrfunc)OutputGetattr;
     OutputType.tp_setattr = (setattrfunc)OutputSetattr;
@@ -6487,7 +6786,7 @@ init_structs(void)
 
     CLEAR_FIELD(BufferType);
     BufferType.tp_name = "vim.buffer";
-    BufferType.tp_basicsize = sizeof(BufferType);
+    BufferType.tp_basicsize = sizeof(BufferObject);
     BufferType.tp_dealloc = (destructor)BufferDestructor;
     BufferType.tp_repr = (reprfunc)BufferRepr;
     BufferType.tp_as_sequence = &BufferAsSeq;
@@ -6550,11 +6849,11 @@ init_structs(void)
     BufMapType.tp_as_mapping = &BufMapAsMapping;
     BufMapType.tp_flags = Py_TPFLAGS_DEFAULT;
     BufMapType.tp_iter = BufMapIter;
-    BufferType.tp_doc = "vim buffer list";
+    BufMapType.tp_doc = "vim buffer list";
 
     CLEAR_FIELD(WinListType);
     WinListType.tp_name = "vim.windowlist";
-    WinListType.tp_basicsize = sizeof(WinListType);
+    WinListType.tp_basicsize = sizeof(WinListObject);
     WinListType.tp_as_sequence = &WinListAsSeq;
     WinListType.tp_flags = Py_TPFLAGS_DEFAULT;
     WinListType.tp_doc = "vim window list";
@@ -6562,7 +6861,7 @@ init_structs(void)
 
     CLEAR_FIELD(TabListType);
     TabListType.tp_name = "vim.tabpagelist";
-    TabListType.tp_basicsize = sizeof(TabListType);
+    TabListType.tp_basicsize = sizeof(TabListObject);
     TabListType.tp_as_sequence = &TabListAsSeq;
     TabListType.tp_flags = Py_TPFLAGS_DEFAULT;
     TabListType.tp_doc = "vim tab page list";
@@ -6690,10 +6989,6 @@ init_structs(void)
 #endif
 }
 
-#define PYTYPE_READY(type) \
-    if (PyType_Ready(&(type))) \
-	return -1;
-
     static int
 init_types(void)
 {
@@ -6714,9 +7009,46 @@ init_types(void)
 #if PY_VERSION_HEX < 0x030700f0
     PYTYPE_READY(LoaderType);
 #endif
+
+#ifdef USE_LIMITED_API
+    // We need to finish initializing all the static objects because the types
+    // are only just allocated on the heap now.
+    // Each PyObject_HEAD_INIT_TYPE should correspond to a
+    // PyObject_FINISH_INIT_TYPE below.
+    PyObject_FINISH_INIT_TYPE(Output, OutputType);
+    PyObject_FINISH_INIT_TYPE(Error, OutputType);
+    PyObject_FINISH_INIT_TYPE(TheBufferMap, BufMapType);
+    PyObject_FINISH_INIT_TYPE(TheWindowList, WinListType);
+    PyObject_FINISH_INIT_TYPE(TheCurrent, CurrentType);
+    PyObject_FINISH_INIT_TYPE(TheTabPageList, TabListType);
+#endif
     return 0;
 }
 
+#ifdef USE_LIMITED_API
+    static void
+shutdown_types(void)
+{
+    PYTYPE_CLEANUP(IterType);
+    PYTYPE_CLEANUP(BufferType);
+    PYTYPE_CLEANUP(RangeType);
+    PYTYPE_CLEANUP(WindowType);
+    PYTYPE_CLEANUP(TabPageType);
+    PYTYPE_CLEANUP(BufMapType);
+    PYTYPE_CLEANUP(WinListType);
+    PYTYPE_CLEANUP(TabListType);
+    PYTYPE_CLEANUP(CurrentType);
+    PYTYPE_CLEANUP(DictionaryType);
+    PYTYPE_CLEANUP(ListType);
+    PYTYPE_CLEANUP(FunctionType);
+    PYTYPE_CLEANUP(OptionsType);
+    PYTYPE_CLEANUP(OutputType);
+# if PY_VERSION_HEX < 0x030700f0
+    PYTYPE_CLEANUP(LoaderType);
+# endif
+}
+#endif
+
     static int
 init_sys_path(void)
 {
@@ -6789,27 +7121,6 @@ init_sys_path(void)
     return 0;
 }
 
-static BufMapObject TheBufferMap =
-{
-    PyObject_HEAD_INIT(&BufMapType)
-};
-
-static WinListObject TheWindowList =
-{
-    PyObject_HEAD_INIT(&WinListType)
-    NULL
-};
-
-static CurrentObject TheCurrent =
-{
-    PyObject_HEAD_INIT(&CurrentType)
-};
-
-static TabListObject TheTabPageList =
-{
-    PyObject_HEAD_INIT(&TabListType)
-};
-
 static struct numeric_constant {
     char	*name;
     int