bpo-42246: Partial implementation of PEP 626. (GH-23113)

* Implement new line number table format, as defined in PEP 626.

Include/cpython/code.h

@@ -38,7 +38,7 @@ struct PyCodeObject {
     Py_ssize_t *co_cell2arg;    /* Maps cell vars which are arguments. */
     PyObject *co_filename;      /* unicode (where it was loaded from) */
     PyObject *co_name;          /* unicode (name, for reference) */
-    PyObject *co_lnotab;        /* string (encoding addr<->lineno mapping) See
+    PyObject *co_linetable;     /* string (encoding addr<->lineno mapping) See
                                    Objects/lnotab_notes.txt for details. */
     void *co_zombieframe;       /* for optimization only (see frameobject.c) */
     PyObject *co_weakreflist;   /* to support weakrefs to code objects */
@@ -135,16 +135,18 @@ PyCode_NewEmpty(const char *filename, const char *funcname, int firstlineno);
 PyAPI_FUNC(int) PyCode_Addr2Line(PyCodeObject *, int);
 
 /* for internal use only */
-typedef struct _addr_pair {
+typedef struct _line_offsets {
-        int ap_lower;
+    int ar_start;
-        int ap_upper;
+    int ar_end;
-} PyAddrPair;
+    int ar_line;
+    int ar_computed_line;
+    char *lo_next;
+} PyCodeAddressRange;
 
 /* Update *bounds to describe the first and one-past-the-last instructions in the
    same line as lasti.  Return the number of that line.
 */
-PyAPI_FUNC(int) _PyCode_CheckLineNumber(PyCodeObject* co,
+PyAPI_FUNC(int) _PyCode_CheckLineNumber(int lasti, PyCodeAddressRange *bounds);
-                                        int lasti, PyAddrPair *bounds);
 
 /* Create a comparable key used to compare constants taking in account the
  * object type. It is used to make sure types are not coerced (e.g., float and
@@ -163,3 +165,15 @@ PyAPI_FUNC(int) _PyCode_GetExtra(PyObject *code, Py_ssize_t index,
                                  void **extra);
 PyAPI_FUNC(int) _PyCode_SetExtra(PyObject *code, Py_ssize_t index,
                                  void *extra);
+
+/** API for initializing the line number table. */
+int _PyCode_InitAddressRange(PyCodeObject* co, PyCodeAddressRange *bounds);
+
+/** Out of process API for initializing the line number table. */
+void PyLineTable_InitAddressRange(char *linetable, int firstlineno, PyCodeAddressRange *range);
+
+/** API for traversing the line number table. */
+int PyLineTable_NextAddressRange(PyCodeAddressRange *range);
+int PyLineTable_PreviousAddressRange(PyCodeAddressRange *range);
+
+

Lib/dis.py

@@ -449,32 +449,15 @@ def findlabels(code):
 def findlinestarts(code):
     """Find the offsets in a byte code which are start of lines in the source.
 
-    Generate pairs (offset, lineno) as described in Python/compile.c.
+    Generate pairs (offset, lineno)
-
     """
-    byte_increments = code.co_lnotab[0::2]
+    lastline = None
-    line_increments = code.co_lnotab[1::2]
+    for start, end, line in code.co_lines():
-    bytecode_len = len(code.co_code)
+        if line is not None and line != lastline:
+            lastline = line
+            yield start, line
+    return
 
-    lastlineno = None
-    lineno = code.co_firstlineno
-    addr = 0
-    for byte_incr, line_incr in zip(byte_increments, line_increments):
-        if byte_incr:
-            if lineno != lastlineno:
-                yield (addr, lineno)
-                lastlineno = lineno
-            addr += byte_incr
-            if addr >= bytecode_len:
-                # The rest of the lnotab byte offsets are past the end of
-                # the bytecode, so the lines were optimized away.
-                return
-        if line_incr >= 0x80:
-            # line_increments is an array of 8-bit signed integers
-            line_incr -= 0x100
-        lineno += line_incr
-    if lineno != lastlineno:
-        yield (addr, lineno)
 
 class Bytecode:
     """The bytecode operations of a piece of code

Lib/test/test_code.py

@@ -258,7 +258,7 @@ class CodeTest(unittest.TestCase):
             ("co_cellvars", ("cellvar",)),
             ("co_filename", "newfilename"),
             ("co_name", "newname"),
-            ("co_lnotab", code2.co_lnotab),
+            ("co_linetable", code2.co_linetable),
         ):
             with self.subTest(attr=attr, value=value):
                 new_code = code.replace(**{attr: value})

Lib/test/test_compile.py

@@ -155,8 +155,8 @@ if 1:
     def test_leading_newlines(self):
         s256 = "".join(["\n"] * 256 + ["spam"])
         co = compile(s256, 'fn', 'exec')
-        self.assertEqual(co.co_firstlineno, 257)
+        self.assertEqual(co.co_firstlineno, 1)
-        self.assertEqual(co.co_lnotab, bytes())
+        self.assertEqual(list(co.co_lines()), [(0, 4, 257), (4, 8, None)])
 
     def test_literals_with_leading_zeroes(self):
         for arg in ["077787", "0xj", "0x.", "0e",  "090000000000000",

Lib/test/test_opcodes.py

@@ -27,7 +27,7 @@ class OpcodeTest(unittest.TestCase):
             with open(ann_module.__file__) as f:
                 txt = f.read()
             co = compile(txt, ann_module.__file__, 'exec')
-            self.assertEqual(co.co_firstlineno, 3)
+            self.assertEqual(co.co_firstlineno, 1)
         except OSError:
             pass
 

Lib/test/test_pdb.py

@@ -1645,9 +1645,10 @@ def bœr():
             'debug doesnotexist',
             'c',
         ])
-        stdout, _ = self.run_pdb_script('', commands + '\n')
+        stdout, _ = self.run_pdb_script('pass', commands + '\n')
 
         self.assertEqual(stdout.splitlines()[1:], [
+            '-> pass',
             '(Pdb) *** SyntaxError: unexpected EOF while parsing',
 
             '(Pdb) ENTERING RECURSIVE DEBUGGER',

Lib/test/test_sys_settrace.py

@@ -220,8 +220,7 @@ ireturn_example.events = [(0, 'call'),
                           (2, 'line'),
                           (3, 'line'),
                           (4, 'line'),
-                          (6, 'line'),
+                          (4, 'return')]
-                          (6, 'return')]
 
 # Tight loop with while(1) example (SF #765624)
 def tightloop_example():
@@ -602,6 +601,17 @@ class TraceTestCase(unittest.TestCase):
         self.compare_events(doit_async.__code__.co_firstlineno,
                             tracer.events, events)
 
+    def test_21_repeated_pass(self):
+        def func():
+            pass
+            pass
+
+        self.run_and_compare(func,
+            [(0, 'call'),
+             (1, 'line'),
+             (2, 'line'),
+             (2, 'return')])
+
     def test_loop_in_try_except(self):
         # https://bugs.python.org/issue41670
 
@@ -766,7 +776,7 @@ class JumpTracer:
         if (self.firstLine is None and frame.f_code == self.code and
                 event == 'line'):
             self.firstLine = frame.f_lineno - 1
-        if (event == self.event and self.firstLine and
+        if (event == self.event and self.firstLine is not None and
                 frame.f_lineno == self.firstLine + self.jumpFrom):
             f = frame
             while f is not None and f.f_code != self.code:
@@ -1540,7 +1550,7 @@ output.append(4)
 """, "<fake module>", "exec")
         class fake_function:
             __code__ = code
-        tracer = JumpTracer(fake_function, 2, 0)
+        tracer = JumpTracer(fake_function, 4, 1)
         sys.settrace(tracer.trace)
         namespace = {"output": []}
         exec(code, namespace)

Misc/NEWS.d/next/Core and Builtins/2020-11-02-15-48-17.bpo-42246.3CNQEX.rst

@@ -0,0 +1,2 @@
+Improved accuracy of line tracing events and f_lineno attribute of Frame
+objects. See PEP 626 for details.

Objects/clinic/codeobject.c.h

@@ -5,7 +5,7 @@ preserve
 PyDoc_STRVAR(code_new__doc__,
 "code(argcount, posonlyargcount, kwonlyargcount, nlocals, stacksize,\n"
 "     flags, codestring, constants, names, varnames, filename, name,\n"
-"     firstlineno, lnotab, freevars=(), cellvars=(), /)\n"
+"     firstlineno, linetable, freevars=(), cellvars=(), /)\n"
 "--\n"
 "\n"
 "Create a code object.  Not for the faint of heart.");
@@ -15,7 +15,7 @@ code_new_impl(PyTypeObject *type, int argcount, int posonlyargcount,
               int kwonlyargcount, int nlocals, int stacksize, int flags,
               PyObject *code, PyObject *consts, PyObject *names,
               PyObject *varnames, PyObject *filename, PyObject *name,
-              int firstlineno, PyObject *lnotab, PyObject *freevars,
+              int firstlineno, PyObject *linetable, PyObject *freevars,
               PyObject *cellvars);
 
 static PyObject *
@@ -35,7 +35,7 @@ code_new(PyTypeObject *type, PyObject *args, PyObject *kwargs)
     PyObject *filename;
     PyObject *name;
     int firstlineno;
-    PyObject *lnotab;
+    PyObject *linetable;
     PyObject *freevars = NULL;
     PyObject *cellvars = NULL;
 
@@ -114,7 +114,7 @@ code_new(PyTypeObject *type, PyObject *args, PyObject *kwargs)
         _PyArg_BadArgument("code", "argument 14", "bytes", PyTuple_GET_ITEM(args, 13));
         goto exit;
     }
-    lnotab = PyTuple_GET_ITEM(args, 13);
+    linetable = PyTuple_GET_ITEM(args, 13);
     if (PyTuple_GET_SIZE(args) < 15) {
         goto skip_optional;
     }
@@ -132,7 +132,7 @@ code_new(PyTypeObject *type, PyObject *args, PyObject *kwargs)
     }
     cellvars = PyTuple_GET_ITEM(args, 15);
 skip_optional:
-    return_value = code_new_impl(type, argcount, posonlyargcount, kwonlyargcount, nlocals, stacksize, flags, code, consts, names, varnames, filename, name, firstlineno, lnotab, freevars, cellvars);
+    return_value = code_new_impl(type, argcount, posonlyargcount, kwonlyargcount, nlocals, stacksize, flags, code, consts, names, varnames, filename, name, firstlineno, linetable, freevars, cellvars);
 
 exit:
     return return_value;
@@ -144,7 +144,7 @@ PyDoc_STRVAR(code_replace__doc__,
 "        co_flags=-1, co_firstlineno=-1, co_code=None, co_consts=None,\n"
 "        co_names=None, co_varnames=None, co_freevars=None,\n"
 "        co_cellvars=None, co_filename=None, co_name=None,\n"
-"        co_lnotab=None)\n"
+"        co_linetable=None)\n"
 "--\n"
 "\n"
 "Return a copy of the code object with new values for the specified fields.");
@@ -160,13 +160,13 @@ code_replace_impl(PyCodeObject *self, int co_argcount,
                   PyObject *co_consts, PyObject *co_names,
                   PyObject *co_varnames, PyObject *co_freevars,
                   PyObject *co_cellvars, PyObject *co_filename,
-                  PyObject *co_name, PyBytesObject *co_lnotab);
+                  PyObject *co_name, PyBytesObject *co_linetable);
 
 static PyObject *
 code_replace(PyCodeObject *self, PyObject *const *args, Py_ssize_t nargs, PyObject *kwnames)
 {
     PyObject *return_value = NULL;
-    static const char * const _keywords[] = {"co_argcount", "co_posonlyargcount", "co_kwonlyargcount", "co_nlocals", "co_stacksize", "co_flags", "co_firstlineno", "co_code", "co_consts", "co_names", "co_varnames", "co_freevars", "co_cellvars", "co_filename", "co_name", "co_lnotab", NULL};
+    static const char * const _keywords[] = {"co_argcount", "co_posonlyargcount", "co_kwonlyargcount", "co_nlocals", "co_stacksize", "co_flags", "co_firstlineno", "co_code", "co_consts", "co_names", "co_varnames", "co_freevars", "co_cellvars", "co_filename", "co_name", "co_linetable", NULL};
     static _PyArg_Parser _parser = {NULL, _keywords, "replace", 0};
     PyObject *argsbuf[16];
     Py_ssize_t noptargs = nargs + (kwnames ? PyTuple_GET_SIZE(kwnames) : 0) - 0;
@@ -185,7 +185,7 @@ code_replace(PyCodeObject *self, PyObject *const *args, Py_ssize_t nargs, PyObje
     PyObject *co_cellvars = self->co_cellvars;
     PyObject *co_filename = self->co_filename;
     PyObject *co_name = self->co_name;
-    PyBytesObject *co_lnotab = (PyBytesObject *)self->co_lnotab;
+    PyBytesObject *co_linetable = (PyBytesObject *)self->co_linetable;
 
     args = _PyArg_UnpackKeywords(args, nargs, NULL, kwnames, &_parser, 0, 0, 0, argsbuf);
     if (!args) {
@@ -344,14 +344,14 @@ code_replace(PyCodeObject *self, PyObject *const *args, Py_ssize_t nargs, PyObje
         }
     }
     if (!PyBytes_Check(args[15])) {
-        _PyArg_BadArgument("replace", "argument 'co_lnotab'", "bytes", args[15]);
+        _PyArg_BadArgument("replace", "argument 'co_linetable'", "bytes", args[15]);
         goto exit;
     }
-    co_lnotab = (PyBytesObject *)args[15];
+    co_linetable = (PyBytesObject *)args[15];
 skip_optional_kwonly:
-    return_value = code_replace_impl(self, co_argcount, co_posonlyargcount, co_kwonlyargcount, co_nlocals, co_stacksize, co_flags, co_firstlineno, co_code, co_consts, co_names, co_varnames, co_freevars, co_cellvars, co_filename, co_name, co_lnotab);
+    return_value = code_replace_impl(self, co_argcount, co_posonlyargcount, co_kwonlyargcount, co_nlocals, co_stacksize, co_flags, co_firstlineno, co_code, co_consts, co_names, co_varnames, co_freevars, co_cellvars, co_filename, co_name, co_linetable);
 
 exit:
     return return_value;
 }
-/*[clinic end generated code: output=18c31941ec09e9ca input=a9049054013a1b77]*/
+/*[clinic end generated code: output=e3091c7baaaaa420 input=a9049054013a1b77]*/

Objects/codeobject.c

@@ -119,7 +119,7 @@ PyCode_NewWithPosOnlyArgs(int argcount, int posonlyargcount, int kwonlyargcount,
                           PyObject *code, PyObject *consts, PyObject *names,
                           PyObject *varnames, PyObject *freevars, PyObject *cellvars,
                           PyObject *filename, PyObject *name, int firstlineno,
-                          PyObject *lnotab)
+                          PyObject *linetable)
 {
     PyCodeObject *co;
     Py_ssize_t *cell2arg = NULL;
@@ -137,7 +137,7 @@ PyCode_NewWithPosOnlyArgs(int argcount, int posonlyargcount, int kwonlyargcount,
         cellvars == NULL || !PyTuple_Check(cellvars) ||
         name == NULL || !PyUnicode_Check(name) ||
         filename == NULL || !PyUnicode_Check(filename) ||
-        lnotab == NULL || !PyBytes_Check(lnotab)) {
+        linetable == NULL || !PyBytes_Check(linetable)) {
         PyErr_BadInternalCall();
         return NULL;
     }
@@ -258,8 +258,8 @@ PyCode_NewWithPosOnlyArgs(int argcount, int posonlyargcount, int kwonlyargcount,
     Py_INCREF(name);
     co->co_name = name;
     co->co_firstlineno = firstlineno;
-    Py_INCREF(lnotab);
+    Py_INCREF(linetable);
-    co->co_lnotab = lnotab;
+    co->co_linetable = linetable;
     co->co_zombieframe = NULL;
     co->co_weakreflist = NULL;
     co->co_extra = NULL;
@@ -277,12 +277,12 @@ PyCode_New(int argcount, int kwonlyargcount,
            PyObject *code, PyObject *consts, PyObject *names,
            PyObject *varnames, PyObject *freevars, PyObject *cellvars,
            PyObject *filename, PyObject *name, int firstlineno,
-           PyObject *lnotab)
+           PyObject *linetable)
 {
     return PyCode_NewWithPosOnlyArgs(argcount, 0, kwonlyargcount, nlocals,
                                      stacksize, flags, code, consts, names,
                                      varnames, freevars, cellvars, filename,
-                                     name, firstlineno, lnotab);
+                                     name, firstlineno, linetable);
 }
 
 int
@@ -369,7 +369,7 @@ PyCode_NewEmpty(const char *filename, const char *funcname, int firstlineno)
                 filename_ob,                    /* filename */
                 funcname_ob,                    /* name */
                 firstlineno,                    /* firstlineno */
-                emptystring                     /* lnotab */
+                emptystring                     /* linetable */
                 );
 
 failed:
@@ -395,11 +395,89 @@ static PyMemberDef code_memberlist[] = {
     {"co_cellvars",     T_OBJECT,       OFF(co_cellvars),        READONLY},
     {"co_filename",     T_OBJECT,       OFF(co_filename),        READONLY},
     {"co_name",         T_OBJECT,       OFF(co_name),            READONLY},
-    {"co_firstlineno", T_INT,           OFF(co_firstlineno),     READONLY},
+    {"co_firstlineno",  T_INT,          OFF(co_firstlineno),     READONLY},
-    {"co_lnotab",       T_OBJECT,       OFF(co_lnotab),          READONLY},
+    {"co_linetable",    T_OBJECT,       OFF(co_linetable),       READONLY},
     {NULL}      /* Sentinel */
 };
 
+static int
+emit_pair(PyObject **bytes, int *offset, int a, int b)
+{
+    Py_ssize_t len = PyBytes_GET_SIZE(*bytes);
+    if (*offset + 2 >= len) {
+        if (_PyBytes_Resize(bytes, len * 2) < 0)
+            return 0;
+    }
+    unsigned char *lnotab = (unsigned char *)
+                    PyBytes_AS_STRING(*bytes) + *offset;
+    *lnotab++ = a;
+    *lnotab++ = b;
+    *offset += 2;
+    return 1;
+}
+
+static int
+emit_delta(PyObject **bytes, int bdelta, int ldelta, int *offset)
+{
+    while (bdelta > 255) {
+        if (!emit_pair(bytes, offset, 255, 0)) {
+            return 0;
+        }
+        bdelta -= 255;
+    }
+    while (ldelta > 127) {
+        if (!emit_pair(bytes, offset, bdelta, 127)) {
+            return 0;
+        }
+        bdelta = 0;
+        ldelta -= 127;
+    }
+    while (ldelta < -128) {
+        if (!emit_pair(bytes, offset, bdelta, -128)) {
+            return 0;
+        }
+        bdelta = 0;
+        ldelta += 128;
+    }
+    return emit_pair(bytes, offset, bdelta, ldelta);
+}
+
+static PyObject *
+code_getlnotab(PyCodeObject *code, void *closure)
+{
+    PyCodeAddressRange bounds;
+    PyObject *bytes;
+    int table_offset = 0;
+    int code_offset = 0;
+    int line = code->co_firstlineno;
+    bytes = PyBytes_FromStringAndSize(NULL, 64);
+    if (bytes == NULL) {
+        return NULL;
+    }
+    _PyCode_InitAddressRange(code, &bounds);
+    while (PyLineTable_NextAddressRange(&bounds)) {
+        if (bounds.ar_computed_line != line) {
+            int bdelta = bounds.ar_start - code_offset;
+            int ldelta = bounds.ar_computed_line - line;
+            if (!emit_delta(&bytes, bdelta, ldelta, &table_offset)) {
+                Py_DECREF(bytes);
+                return NULL;
+            }
+            code_offset = bounds.ar_start;
+            line = bounds.ar_computed_line;
+        }
+    }
+    _PyBytes_Resize(&bytes, table_offset);
+    return bytes;
+}
+
+
+static PyGetSetDef code_getsetlist[] = {
+    {"co_lnotab",    (getter)code_getlnotab, NULL, NULL},
+    {0}
+};
+
+
 /* Helper for code_new: return a shallow copy of a tuple that is
    guaranteed to contain exact strings, by converting string subclasses
    to exact strings and complaining if a non-string is found. */
@@ -459,7 +537,7 @@ code.__new__ as code_new
     filename: unicode
     name: unicode
     firstlineno: int
-    lnotab: object(subclass_of="&PyBytes_Type")
+    linetable: object(subclass_of="&PyBytes_Type")
     freevars: object(subclass_of="&PyTuple_Type", c_default="NULL") = ()
     cellvars: object(subclass_of="&PyTuple_Type", c_default="NULL") = ()
     /
@@ -472,9 +550,9 @@ code_new_impl(PyTypeObject *type, int argcount, int posonlyargcount,
               int kwonlyargcount, int nlocals, int stacksize, int flags,
               PyObject *code, PyObject *consts, PyObject *names,
               PyObject *varnames, PyObject *filename, PyObject *name,
-              int firstlineno, PyObject *lnotab, PyObject *freevars,
+              int firstlineno, PyObject *linetable, PyObject *freevars,
               PyObject *cellvars)
-/*[clinic end generated code: output=612aac5395830184 input=85e678ea4178f234]*/
+/*[clinic end generated code: output=42c1839b082ba293 input=0ec80da632b99f57]*/
 {
     PyObject *co = NULL;
     PyObject *ournames = NULL;
@@ -540,7 +618,7 @@ code_new_impl(PyTypeObject *type, int argcount, int posonlyargcount,
                                                code, consts, ournames,
                                                ourvarnames, ourfreevars,
                                                ourcellvars, filename,
-                                               name, firstlineno, lnotab);
+                                               name, firstlineno, linetable);
   cleanup:
     Py_XDECREF(ournames);
     Py_XDECREF(ourvarnames);
@@ -584,7 +662,7 @@ code_dealloc(PyCodeObject *co)
     Py_XDECREF(co->co_cellvars);
     Py_XDECREF(co->co_filename);
     Py_XDECREF(co->co_name);
-    Py_XDECREF(co->co_lnotab);
+    Py_XDECREF(co->co_linetable);
     if (co->co_cell2arg != NULL)
         PyMem_FREE(co->co_cell2arg);
     if (co->co_zombieframe != NULL)
@@ -636,7 +714,7 @@ code.replace
     co_cellvars: object(subclass_of="&PyTuple_Type", c_default="self->co_cellvars") = None
     co_filename: unicode(c_default="self->co_filename") = None
     co_name: unicode(c_default="self->co_name") = None
-    co_lnotab: PyBytesObject(c_default="(PyBytesObject *)self->co_lnotab") = None
+    co_linetable: PyBytesObject(c_default="(PyBytesObject *)self->co_linetable") = None
 
 Return a copy of the code object with new values for the specified fields.
 [clinic start generated code]*/
@@ -649,8 +727,8 @@ code_replace_impl(PyCodeObject *self, int co_argcount,
                   PyObject *co_consts, PyObject *co_names,
                   PyObject *co_varnames, PyObject *co_freevars,
                   PyObject *co_cellvars, PyObject *co_filename,
-                  PyObject *co_name, PyBytesObject *co_lnotab)
+                  PyObject *co_name, PyBytesObject *co_linetable)
-/*[clinic end generated code: output=25c8e303913bcace input=d9051bc8f24e6b28]*/
+/*[clinic end generated code: output=50d77e668d3b449b input=a5f997b173d7f636]*/
 {
 #define CHECK_INT_ARG(ARG) \
         if (ARG < 0) { \
@@ -680,7 +758,7 @@ code_replace_impl(PyCodeObject *self, int co_argcount,
         co_argcount, co_posonlyargcount, co_kwonlyargcount, co_nlocals,
         co_stacksize, co_flags, (PyObject*)co_code, co_consts, co_names,
         co_varnames, co_freevars, co_cellvars, co_filename, co_name,
-        co_firstlineno, (PyObject*)co_lnotab);
+        co_firstlineno, (PyObject*)co_linetable);
 }
 
 static PyObject *
@@ -933,10 +1011,189 @@ code_hash(PyCodeObject *co)
     return h;
 }
 
+typedef struct {
+    PyObject_HEAD
+    PyCodeObject *li_code;
+    PyCodeAddressRange li_line;
+    char *li_end;
+} lineiterator;
+
+
+static void
+lineiter_dealloc(lineiterator *li)
+{
+    Py_DECREF(li->li_code);
+    Py_TYPE(li)->tp_free(li);
+}
+
+static PyObject *
+lineiter_next(lineiterator *li)
+{
+    PyCodeAddressRange *bounds = &li->li_line;
+    if (!PyLineTable_NextAddressRange(bounds)) {
+        return NULL;
+    }
+    PyObject *start = NULL;
+    PyObject *end = NULL;
+    PyObject *line = NULL;
+    PyObject *result = PyTuple_New(3);
+    start = PyLong_FromLong(bounds->ar_start);
+    end = PyLong_FromLong(bounds->ar_end);
+    if (bounds->ar_line < 0) {
+        Py_INCREF(Py_None);
+        line = Py_None;
+    }
+    else {
+        line = PyLong_FromLong(bounds->ar_line);
+    }
+    if (result == NULL || start == NULL || end == NULL || line == NULL) {
+        goto error;
+    }
+    PyTuple_SET_ITEM(result, 0, start);
+    PyTuple_SET_ITEM(result, 1, end);
+    PyTuple_SET_ITEM(result, 2, line);
+    return result;
+error:
+    Py_XDECREF(start);
+    Py_XDECREF(end);
+    Py_XDECREF(line);
+    Py_XDECREF(result);
+    return result;
+}
+
+static PyTypeObject LineIterator = {
+    PyVarObject_HEAD_INIT(&PyType_Type, 0)
+    "line_iterator",                    /* tp_name */
+    sizeof(lineiterator),               /* tp_basicsize */
+    0,                                  /* tp_itemsize */
+    /* methods */
+    (destructor)lineiter_dealloc,       /* tp_dealloc */
+    0,                                  /* tp_vectorcall_offset */
+    0,                                  /* tp_getattr */
+    0,                                  /* tp_setattr */
+    0,                                  /* tp_as_async */
+    0,                                  /* tp_repr */
+    0,                                  /* tp_as_number */
+    0,                                  /* tp_as_sequence */
+    0,                                  /* tp_as_mapping */
+    0,                                  /* tp_hash */
+    0,                                  /* tp_call */
+    0,                                  /* tp_str */
+    0,                                  /* tp_getattro */
+    0,                                  /* tp_setattro */
+    0,                                  /* tp_as_buffer */
+    Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE,       /* tp_flags */
+    0,                                  /* tp_doc */
+    0,                                  /* tp_traverse */
+    0,                                  /* tp_clear */
+    0,                                  /* tp_richcompare */
+    0,                                  /* tp_weaklistoffset */
+    PyObject_SelfIter,                  /* tp_iter */
+    (iternextfunc)lineiter_next,        /* tp_iternext */
+    0,                                  /* tp_methods */
+    0,                                  /* tp_members */
+    0,                                  /* tp_getset */
+    0,                                  /* tp_base */
+    0,                                  /* tp_dict */
+    0,                                  /* tp_descr_get */
+    0,                                  /* tp_descr_set */
+    0,                                  /* tp_dictoffset */
+    0,                                  /* tp_init */
+    0,                                  /* tp_alloc */
+    0,                                  /* tp_new */
+    PyObject_Del,                       /* tp_free */
+};
+
+static PyObject *
+code_linesiterator(PyCodeObject *code, PyObject *Py_UNUSED(args))
+{
+    lineiterator *li = (lineiterator *)PyType_GenericAlloc(&LineIterator, 0);
+    if (li == NULL) {
+        return NULL;
+    }
+    Py_INCREF(code);
+    li->li_code = code;
+    _PyCode_InitAddressRange(code, &li->li_line);
+    return (PyObject *)li;
+}
+
+static void
+retreat(PyCodeAddressRange *bounds)
+{
+    int ldelta = ((signed char *)bounds->lo_next)[-1];
+    if (ldelta == -128) {
+        ldelta = 0;
+    }
+    bounds->ar_computed_line -= ldelta;
+    bounds->lo_next -= 2;
+    bounds->ar_end = bounds->ar_start;
+    bounds->ar_start -= ((unsigned char *)bounds->lo_next)[-2];
+    ldelta = ((signed char *)bounds->lo_next)[-1];
+    if (ldelta == -128) {
+        bounds->ar_line = -1;
+    }
+    else {
+        bounds->ar_line = bounds->ar_computed_line;
+    }
+}
+
+static void
+advance(PyCodeAddressRange *bounds)
+{
+    bounds->ar_start = bounds->ar_end;
+    int delta = ((unsigned char *)bounds->lo_next)[0];
+    assert (delta < 255);
+    bounds->ar_end += delta;
+    int ldelta = ((signed char *)bounds->lo_next)[1];
+    bounds->lo_next += 2;
+    if (ldelta == -128) {
+        bounds->ar_line = -1;
+    }
+    else {
+        bounds->ar_computed_line += ldelta;
+        bounds->ar_line = bounds->ar_computed_line;
+    }
+}
+
+static inline int
+at_end(PyCodeAddressRange *bounds) {
+    return ((unsigned char *)bounds->lo_next)[0] == 255;
+}
+
+int
+PyLineTable_PreviousAddressRange(PyCodeAddressRange *range)
+{
+    if (range->ar_start <= 0) {
+        return 0;
+    }
+    retreat(range);
+    while (range->ar_start == range->ar_end) {
+        assert(range->ar_start > 0);
+        retreat(range);
+    }
+    return 1;
+}
+
+int
+PyLineTable_NextAddressRange(PyCodeAddressRange *range)
+{
+    if (at_end(range)) {
+        return 0;
+    }
+    advance(range);
+    while (range->ar_start == range->ar_end) {
+        assert(!at_end(range));
+        advance(range);
+    }
+    return 1;
+}
+
+
 /* XXX code objects need to participate in GC? */
 
 static struct PyMethodDef code_methods[] = {
     {"__sizeof__", (PyCFunction)code_sizeof, METH_NOARGS},
+    {"co_lines", (PyCFunction)code_linesiterator, METH_NOARGS},
     CODE_REPLACE_METHODDEF
     {NULL, NULL}                /* sentinel */
 };
@@ -971,7 +1228,7 @@ PyTypeObject PyCode_Type = {
     0,                                  /* tp_iternext */
     code_methods,                       /* tp_methods */
     code_memberlist,                    /* tp_members */
-    0,                                  /* tp_getset */
+    code_getsetlist,                    /* tp_getset */
     0,                                  /* tp_base */
     0,                                  /* tp_dict */
     0,                                  /* tp_descr_get */
@@ -982,78 +1239,55 @@ PyTypeObject PyCode_Type = {
     code_new,                           /* tp_new */
 };
 
-/* Use co_lnotab to compute the line number from a bytecode index, addrq.  See
+/* Use co_linetable to compute the line number from a bytecode index, addrq.  See
    lnotab_notes.txt for the details of the lnotab representation.
 */
 
 int
 PyCode_Addr2Line(PyCodeObject *co, int addrq)
 {
-    Py_ssize_t size = PyBytes_Size(co->co_lnotab) / 2;
+    if (addrq == -1) {
-    unsigned char *p = (unsigned char*)PyBytes_AsString(co->co_lnotab);
+        return co->co_firstlineno;
-    int line = co->co_firstlineno;
-    int addr = 0;
-    while (--size >= 0) {
-        addr += *p++;
-        if (addr > addrq)
-            break;
-        line += (signed char)*p;
-        p++;
     }
-    return line;
+    assert(addrq >= 0 && addrq < PyBytes_GET_SIZE(co->co_code));
+    PyCodeAddressRange bounds;
+    _PyCode_InitAddressRange(co, &bounds);
+    return _PyCode_CheckLineNumber(addrq, &bounds);
+}
+
+void
+PyLineTable_InitAddressRange(char *linetable, int firstlineno, PyCodeAddressRange *range)
+{
+    range->lo_next = linetable;
+    range->ar_start = -1;
+    range->ar_end = 0;
+    range->ar_computed_line = range->ar_line = firstlineno;
+}
+
+int
+_PyCode_InitAddressRange(PyCodeObject* co, PyCodeAddressRange *bounds)
+{
+    char *linetable = PyBytes_AS_STRING(co->co_linetable);
+    PyLineTable_InitAddressRange(linetable, co->co_firstlineno, bounds);
+    return bounds->ar_line;
 }
 
 /* Update *bounds to describe the first and one-past-the-last instructions in
-   the same line as lasti.  Return the number of that line. */
+   the same line as lasti.  Return the number of that line, or -1 if lasti is out of bounds. */
 int
-_PyCode_CheckLineNumber(PyCodeObject* co, int lasti, PyAddrPair *bounds)
+_PyCode_CheckLineNumber(int lasti, PyCodeAddressRange *bounds)
 {
-    Py_ssize_t size;
+    while (bounds->ar_end <= lasti) {
-    int addr, line;
+        if (!PyLineTable_NextAddressRange(bounds)) {
-    unsigned char* p;
+            return -1;
-
-    p = (unsigned char*)PyBytes_AS_STRING(co->co_lnotab);
-    size = PyBytes_GET_SIZE(co->co_lnotab) / 2;
-
-    addr = 0;
-    line = co->co_firstlineno;
-    assert(line > 0);
-
-    /* possible optimization: if f->f_lasti == instr_ub
-       (likely to be a common case) then we already know
-       instr_lb -- if we stored the matching value of p
-       somewhere we could skip the first while loop. */
-
-    /* See lnotab_notes.txt for the description of
-       co_lnotab.  A point to remember: increments to p
-       come in (addr, line) pairs. */
-
-    bounds->ap_lower = 0;
-    while (size > 0) {
-        if (addr + *p > lasti)
-            break;
-        addr += *p++;
-        if ((signed char)*p)
-            bounds->ap_lower = addr;
-        line += (signed char)*p;
-        p++;
-        --size;
-    }
-
-    if (size > 0) {
-        while (--size >= 0) {
-            addr += *p++;
-            if ((signed char)*p)
-                break;
-            p++;
         }
-        bounds->ap_upper = addr;
     }
-    else {
+    while (bounds->ar_start > lasti) {
-        bounds->ap_upper = INT_MAX;
+        if (!PyLineTable_PreviousAddressRange(bounds)) {
+            return -1;
+        }
     }
-
+    return bounds->ar_line;
-    return line;
 }
 
 

Objects/frameobject.c

@@ -249,36 +249,22 @@ explain_incompatible_block_stack(int64_t to_stack)
 static int *
 marklines(PyCodeObject *code, int len)
 {
+    PyCodeAddressRange bounds;
+    _PyCode_InitAddressRange(code, &bounds);
+    assert (bounds.ar_end == 0);
+
     int *linestarts = PyMem_New(int, len);
     if (linestarts == NULL) {
         return NULL;
     }
-    Py_ssize_t size = PyBytes_GET_SIZE(code->co_lnotab) / 2;
+    for (int i = 0; i < len; i++) {
-    unsigned char *p = (unsigned char*)PyBytes_AS_STRING(code->co_lnotab);
+        linestarts[i] = -1;
-    int line = code->co_firstlineno;
-    int addr = 0;
-    int index = 0;
-    while (--size >= 0) {
-        addr += *p++;
-        if (index*2 < addr) {
-            linestarts[index++] = line;
-        }
-        while (index*2 < addr) {
-            linestarts[index++] = -1;
-            if (index >= len) {
-                break;
-            }
-        }
-        line += (signed char)*p;
-        p++;
     }
-    if (index < len) {
+
-        linestarts[index++] = line;
+    while (PyLineTable_NextAddressRange(&bounds)) {
+        assert(bounds.ar_start/2 < len);
+        linestarts[bounds.ar_start/2] = bounds.ar_line;
     }
-    while (index < len) {
-        linestarts[index++] = -1;
-    }
-    assert(index == len);
     return linestarts;
 }
 
@@ -925,7 +911,7 @@ _PyFrame_New_NoTrack(PyThreadState *tstate, PyCodeObject *code,
     }
 
     f->f_lasti = -1;
-    f->f_lineno = code->co_firstlineno;
+    f->f_lineno = 0;
     f->f_iblock = 0;
     f->f_state = FRAME_CREATED;
     f->f_gen = NULL;

Objects/lnotab_notes.txt

@@ -1,11 +1,103 @@
-All about co_lnotab, the line number table.
+Description of the internal format of the line number table
+
+Conceptually, the line number table consists of a sequence of triples:
+    start-offset (inclusive), end-offset (exclusive), line-number.
+
+Note that note all byte codes have a line number so we need handle `None` for the line-number.
+
+However, storing the above sequence directly would be very inefficient as we would need 12 bytes per entry.
+
+First of all, we can note that the end of one entry is the same as the start of the next, so we can overlap entries.
+Secondly we also note that we don't really need arbitrary access to the sequence, so we can store deltas.
+
+We just need to store (end - start, line delta) pairs. The start offset of the first entry is always zero.
+
+Thirdly, most deltas are small, so we can use a single byte for each value, as long we allow several entries for the same line.
+
+Consider the following table
+     Start    End     Line
+      0       6       1
+      6       50      2
+      50      350     7
+      350     360     No line number
+      360     376     8
+      376     380     208
+
+Stripping the redundant ends gives:
+
+   End-Start  Line-delta
+      6         +1
+      44        +1
+      300       +5
+      10        No line number
+      16        +1
+      4         +200
+
+
+Note that the end - start value is always positive.
+
+Finally in order, to fit into a single byte we need to convert start deltas to the range 0 <= delta <= 254,
+and line deltas to the range -127  <= delta <= 127.
+A line delta of -128 is used to indicate no line number.
+A start delta of 255 is used as a sentinel to mark the end of the table.
+Also note that a delta of zero indicates that there are no bytecodes in the given range,
+which means can use an invalidate line number for that range.
+
+Final form:
+
+   Start delta   Line delta
+    6               +1
+    44              +1
+    254             +5
+    46              0
+    10              -128 (No line number, treated as a delta of zero)
+    16              +1
+    0               +127 (line 135, but the range is empty as no bytecodes are at line 135)
+    4               +73
+    255 (end mark)  ---
+
+Iterating over the table.
+-------------------------
+
+For the `co_lines` attribute we want to emit the full form, omitting the (350, 360, No line number) and empty entries.
+
+The code is as follows:
+
+def co_lines(code):
+    line = code.co_firstlineno
+    end = 0
+    table_iter = iter(code.internal_line_table):
+    for sdelta, ldelta in table_iter:
+        if sdelta == 255:
+            break
+        if ldelta == 0: # No change to line number, just accumulate changes to end
+            end += odelta
+            continue
+        start = end
+        end = start + sdelta
+        if ldelta == -128: # No valid line number -- skip entry
+            continue
+        line += ldelta
+        if end == start: # Empty range, omit.
+            continue
+        yield start, end, line
+
+
+
+
+The historical co_lnotab format
+-------------------------------
+
+prior to 3.10 code objects stored a field named co_lnotab.
+This was an array of unsigned bytes disguised as a Python bytes object.
+
+The old co_lnotab did not account for the presence of bytecodes without a line number,
+nor was it well suited to tracing as a number of workarounds were required.
+
+The old format can still be accessed via `code.co_lnotab`, which is lazily computed from the new format.
+
+Below is the description of the old co_lnotab format:
 
-Code objects store a field named co_lnotab.  This is an array of unsigned bytes
-disguised as a Python bytes object.  It is used to map bytecode offsets to
-source code line #s for tracebacks and to identify line number boundaries for
-line tracing. Because of internals of the peephole optimizer, it's possible
-for lnotab to contain bytecode offsets that are no longer valid (for example
-if the optimizer removed the last line in a function).
 
 The array is conceptually a compressed list of
     (bytecode offset increment, line number increment)

Python/ceval.c

@@ -66,8 +66,8 @@ static void call_exc_trace(Py_tracefunc, PyObject *,
                            PyThreadState *, PyFrameObject *);
 static int maybe_call_line_trace(Py_tracefunc, PyObject *,
                                  PyThreadState *, PyFrameObject *,
-                                 int *, int *, int *);
+                                 PyCodeAddressRange *, int *);
-static void maybe_dtrace_line(PyFrameObject *, int *, int *, int *);
+static void maybe_dtrace_line(PyFrameObject *, PyCodeAddressRange *, int *);
 static void dtrace_function_entry(PyFrameObject *);
 static void dtrace_function_return(PyFrameObject *);
 
@@ -976,7 +976,6 @@ _PyEval_EvalFrameDefault(PyThreadState *tstate, PyFrameObject *f, int throwflag)
        is true when the line being executed has changed.  The
        initial values are such as to make this false the first
        time it is tested. */
-    int instr_ub = -1, instr_lb = 0, instr_prev = -1;
 
     const _Py_CODEUNIT *first_instr;
     PyObject *names;
@@ -1390,6 +1389,10 @@ _PyEval_EvalFrameDefault(PyThreadState *tstate, PyFrameObject *f, int throwflag)
         dtrace_function_entry(f);
 
     co = f->f_code;
+    PyCodeAddressRange bounds;
+    _PyCode_InitAddressRange(co, &bounds);
+    int instr_prev = -1;
+
     names = co->co_names;
     consts = co->co_consts;
     fastlocals = f->f_localsplus;
@@ -1514,7 +1517,7 @@ main_loop:
         f->f_lasti = INSTR_OFFSET();
 
         if (PyDTrace_LINE_ENABLED())
-            maybe_dtrace_line(f, &instr_lb, &instr_ub, &instr_prev);
+            maybe_dtrace_line(f, &bounds, &instr_prev);
 
         /* line-by-line tracing support */
 
@@ -1528,7 +1531,7 @@ main_loop:
             err = maybe_call_line_trace(tstate->c_tracefunc,
                                         tstate->c_traceobj,
                                         tstate, f,
-                                        &instr_lb, &instr_ub, &instr_prev);
+                                        &bounds, &instr_prev);
             /* Reload possibly changed frame fields */
             JUMPTO(f->f_lasti);
             stack_pointer = f->f_valuestack+f->f_stackdepth;
@@ -4039,14 +4042,7 @@ exception_unwind:
                 PUSH(exc);
                 JUMPTO(handler);
                 if (_Py_TracingPossible(ceval2)) {
-                    int needs_new_execution_window = (f->f_lasti < instr_lb || f->f_lasti >= instr_ub);
+                    instr_prev = INT_MAX;
-                    int needs_line_update = (f->f_lasti == instr_lb || f->f_lasti < instr_prev);
-                    /* Make sure that we trace line after exception if we are in a new execution
-                     * window or we don't need a line update and we are not in the first instruction
-                     * of the line. */
-                    if (needs_new_execution_window || (!needs_line_update && instr_lb > 0)) {
-                        instr_prev = INT_MAX;
-                    }
                 }
                 /* Resume normal execution */
                 f->f_state = FRAME_EXECUTING;
@@ -4960,7 +4956,7 @@ _PyEval_CallTracing(PyObject *func, PyObject *args)
 static int
 maybe_call_line_trace(Py_tracefunc func, PyObject *obj,
                       PyThreadState *tstate, PyFrameObject *frame,
-                      int *instr_lb, int *instr_ub, int *instr_prev)
+                      PyCodeAddressRange *bounds, int *instr_prev)
 {
     int result = 0;
     int line = frame->f_lineno;
@@ -4968,21 +4964,17 @@ maybe_call_line_trace(Py_tracefunc func, PyObject *obj,
     /* If the last instruction executed isn't in the current
        instruction window, reset the window.
     */
-    if (frame->f_lasti < *instr_lb || frame->f_lasti >= *instr_ub) {
+    line = _PyCode_CheckLineNumber(frame->f_lasti, bounds);
-        PyAddrPair bounds;
-        line = _PyCode_CheckLineNumber(frame->f_code, frame->f_lasti,
-                                       &bounds);
-        *instr_lb = bounds.ap_lower;
-        *instr_ub = bounds.ap_upper;
-    }
     /* If the last instruction falls at the start of a line or if it
        represents a jump backwards, update the frame's line number and
        then call the trace function if we're tracing source lines.
     */
-    if ((frame->f_lasti == *instr_lb || frame->f_lasti < *instr_prev)) {
+    if ((line != frame->f_lineno || frame->f_lasti < *instr_prev)) {
-        frame->f_lineno = line;
+        if (line != -1) {
-        if (frame->f_trace_lines) {
+            frame->f_lineno = line;
-            result = call_trace(func, obj, tstate, frame, PyTrace_LINE, Py_None);
+            if (frame->f_trace_lines) {
+                result = call_trace(func, obj, tstate, frame, PyTrace_LINE, Py_None);
+            }
         }
     }
     /* Always emit an opcode event if we're tracing all opcodes. */
@@ -5907,33 +5899,28 @@ dtrace_function_return(PyFrameObject *f)
 /* DTrace equivalent of maybe_call_line_trace. */
 static void
 maybe_dtrace_line(PyFrameObject *frame,
-                  int *instr_lb, int *instr_ub, int *instr_prev)
+                  PyCodeAddressRange *bounds, int *instr_prev)
 {
-    int line = frame->f_lineno;
     const char *co_filename, *co_name;
 
     /* If the last instruction executed isn't in the current
        instruction window, reset the window.
     */
-    if (frame->f_lasti < *instr_lb || frame->f_lasti >= *instr_ub) {
+    int line = _PyCode_CheckLineNumber(frame->f_lasti, bounds);
-        PyAddrPair bounds;
-        line = _PyCode_CheckLineNumber(frame->f_code, frame->f_lasti,
-                                       &bounds);
-        *instr_lb = bounds.ap_lower;
-        *instr_ub = bounds.ap_upper;
-    }
     /* If the last instruction falls at the start of a line or if
        it represents a jump backwards, update the frame's line
        number and call the trace function. */
-    if (frame->f_lasti == *instr_lb || frame->f_lasti < *instr_prev) {
+    if (line != frame->f_lineno || frame->f_lasti < *instr_prev) {
-        frame->f_lineno = line;
+        if (line != -1) {
-        co_filename = PyUnicode_AsUTF8(frame->f_code->co_filename);
+            frame->f_lineno = line;
-        if (!co_filename)
+            co_filename = PyUnicode_AsUTF8(frame->f_code->co_filename);
-            co_filename = "?";
+            if (!co_filename)
-        co_name = PyUnicode_AsUTF8(frame->f_code->co_name);
+                co_filename = "?";
-        if (!co_name)
+            co_name = PyUnicode_AsUTF8(frame->f_code->co_name);
-            co_name = "?";
+            if (!co_name)
-        PyDTrace_LINE(co_filename, co_name, line);
+                co_name = "?";
+            PyDTrace_LINE(co_filename, co_name, line);
+        }
     }
     *instr_prev = frame->f_lasti;
 }

Python/compile.c

@@ -1827,7 +1827,7 @@ compiler_mod(struct compiler *c, mod_ty mod)
             return NULL;
     }
     /* Use 0 for firstlineno initially, will fixup in assemble(). */
-    if (!compiler_enter_scope(c, module, COMPILER_SCOPE_MODULE, mod, 0))
+    if (!compiler_enter_scope(c, module, COMPILER_SCOPE_MODULE, mod, 1))
         return NULL;
     switch (mod->kind) {
     case Module_kind:
@@ -2271,7 +2271,9 @@ compiler_function(struct compiler *c, stmt_ty s, int is_async)
     c->u->u_argcount = asdl_seq_LEN(args->args);
     c->u->u_posonlyargcount = asdl_seq_LEN(args->posonlyargs);
     c->u->u_kwonlyargcount = asdl_seq_LEN(args->kwonlyargs);
-    VISIT_SEQ_IN_SCOPE(c, stmt, body);
+    for (i = docstring ? 1 : 0; i < asdl_seq_LEN(body); i++) {
+        VISIT(c, stmt, (stmt_ty)asdl_seq_GET(body, i));
+    }
     co = assemble(c, 1);
     qualname = c->u->u_qualname;
     Py_INCREF(qualname);
@@ -2808,6 +2810,8 @@ compiler_async_for(struct compiler *c, stmt_ty s)
 
     /* Except block for __anext__ */
     compiler_use_next_block(c, except);
+
+    c->u->u_lineno = -1;
     ADDOP(c, END_ASYNC_FOR);
 
     /* `else` block */
@@ -3115,7 +3119,8 @@ compiler_try_except(struct compiler *c, stmt_ty s)
             compiler_pop_fblock(c, HANDLER_CLEANUP, cleanup_body);
             ADDOP(c, POP_BLOCK);
             ADDOP(c, POP_EXCEPT);
-            /* name = None; del name */
+            /* name = None; del name; # Mark as artificial */
+            c->u->u_lineno = -1;
             ADDOP_LOAD_CONST(c, Py_None);
             compiler_nameop(c, handler->v.ExceptHandler.name, Store);
             compiler_nameop(c, handler->v.ExceptHandler.name, Del);
@@ -3124,7 +3129,8 @@ compiler_try_except(struct compiler *c, stmt_ty s)
             /* except: */
             compiler_use_next_block(c, cleanup_end);
 
-            /* name = None; del name */
+            /* name = None; del name; # Mark as artificial */
+            c->u->u_lineno = -1;
             ADDOP_LOAD_CONST(c, Py_None);
             compiler_nameop(c, handler->v.ExceptHandler.name, Store);
             compiler_nameop(c, handler->v.ExceptHandler.name, Del);
@@ -3359,6 +3365,7 @@ compiler_visit_stmt_expr(struct compiler *c, expr_ty value)
 
     if (value->kind == Constant_kind) {
         /* ignore constant statement */
+        ADDOP(c, NOP);
         return 1;
     }
 
@@ -3431,6 +3438,7 @@ compiler_visit_stmt(struct compiler *c, stmt_ty s)
     case Expr_kind:
         return compiler_visit_stmt_expr(c, s->v.Expr.value);
     case Pass_kind:
+        ADDOP(c, NOP);
         break;
     case Break_kind:
         return compiler_break(c);
@@ -5429,8 +5437,9 @@ struct assembler {
     basicblock **a_reverse_postorder; /* list of blocks in dfs postorder */
     PyObject *a_lnotab;    /* string containing lnotab */
     int a_lnotab_off;      /* offset into lnotab */
-    int a_lineno;              /* last lineno of emitted instruction */
+    int a_prevlineno;     /* lineno of last emitted line in line table */
-    int a_lineno_off;      /* bytecode offset of last lineno */
+    int a_lineno;          /* lineno of last emitted instruction */
+    int a_lineno_start;    /* bytecode start offset of current lineno */
 };
 
 static void
@@ -5533,7 +5542,7 @@ static int
 assemble_init(struct assembler *a, int nblocks, int firstlineno)
 {
     memset(a, 0, sizeof(struct assembler));
-    a->a_lineno = firstlineno;
+    a->a_prevlineno = a->a_lineno = firstlineno;
     a->a_bytecode = PyBytes_FromStringAndSize(NULL, DEFAULT_CODE_SIZE);
     if (!a->a_bytecode)
         return 0;
@@ -5573,114 +5582,82 @@ blocksize(basicblock *b)
     return size;
 }
 
-/* Appends a pair to the end of the line number table, a_lnotab, representing
-   the instruction's bytecode offset and line number.  See
-   Objects/lnotab_notes.txt for the description of the line number table. */
-
 static int
-assemble_lnotab(struct assembler *a, struct instr *i)
+assemble_emit_linetable_pair(struct assembler *a, int bdelta, int ldelta)
 {
-    int d_bytecode, d_lineno;
+    Py_ssize_t len = PyBytes_GET_SIZE(a->a_lnotab);
-    Py_ssize_t len;
-    unsigned char *lnotab;
-
-    d_lineno = i->i_lineno - a->a_lineno;
-    if (d_lineno == 0) {
-        return 1;
-    }
-
-    d_bytecode = (a->a_offset - a->a_lineno_off) * sizeof(_Py_CODEUNIT);
-    assert(d_bytecode >= 0);
-
-    if (d_bytecode > 255) {
-        int j, nbytes, ncodes = d_bytecode / 255;
-        nbytes = a->a_lnotab_off + 2 * ncodes;
-        len = PyBytes_GET_SIZE(a->a_lnotab);
-        if (nbytes >= len) {
-            if ((len <= INT_MAX / 2) && (len * 2 < nbytes))
-                len = nbytes;
-            else if (len <= INT_MAX / 2)
-                len *= 2;
-            else {
-                PyErr_NoMemory();
-                return 0;
-            }
-            if (_PyBytes_Resize(&a->a_lnotab, len) < 0)
-                return 0;
-        }
-        lnotab = (unsigned char *)
-                   PyBytes_AS_STRING(a->a_lnotab) + a->a_lnotab_off;
-        for (j = 0; j < ncodes; j++) {
-            *lnotab++ = 255;
-            *lnotab++ = 0;
-        }
-        d_bytecode -= ncodes * 255;
-        a->a_lnotab_off += ncodes * 2;
-    }
-    assert(0 <= d_bytecode && d_bytecode <= 255);
-
-    if (d_lineno < -128 || 127 < d_lineno) {
-        int j, nbytes, ncodes, k;
-        if (d_lineno < 0) {
-            k = -128;
-            /* use division on positive numbers */
-            ncodes = (-d_lineno) / 128;
-        }
-        else {
-            k = 127;
-            ncodes = d_lineno / 127;
-        }
-        d_lineno -= ncodes * k;
-        assert(ncodes >= 1);
-        nbytes = a->a_lnotab_off + 2 * ncodes;
-        len = PyBytes_GET_SIZE(a->a_lnotab);
-        if (nbytes >= len) {
-            if ((len <= INT_MAX / 2) && len * 2 < nbytes)
-                len = nbytes;
-            else if (len <= INT_MAX / 2)
-                len *= 2;
-            else {
-                PyErr_NoMemory();
-                return 0;
-            }
-            if (_PyBytes_Resize(&a->a_lnotab, len) < 0)
-                return 0;
-        }
-        lnotab = (unsigned char *)
-                   PyBytes_AS_STRING(a->a_lnotab) + a->a_lnotab_off;
-        *lnotab++ = d_bytecode;
-        *lnotab++ = k;
-        d_bytecode = 0;
-        for (j = 1; j < ncodes; j++) {
-            *lnotab++ = 0;
-            *lnotab++ = k;
-        }
-        a->a_lnotab_off += ncodes * 2;
-    }
-    assert(-128 <= d_lineno && d_lineno <= 127);
-
-    len = PyBytes_GET_SIZE(a->a_lnotab);
     if (a->a_lnotab_off + 2 >= len) {
         if (_PyBytes_Resize(&a->a_lnotab, len * 2) < 0)
             return 0;
     }
-    lnotab = (unsigned char *)
+    unsigned char *lnotab = (unsigned char *)
                     PyBytes_AS_STRING(a->a_lnotab) + a->a_lnotab_off;
 
     a->a_lnotab_off += 2;
-    if (d_bytecode) {
+    *lnotab++ = bdelta;
-        *lnotab++ = d_bytecode;
+    *lnotab++ = ldelta;
-        *lnotab++ = d_lineno;
-    }
-    else {      /* First line of a block; def stmt, etc. */
-        *lnotab++ = 0;
-        *lnotab++ = d_lineno;
-    }
-    a->a_lineno = i->i_lineno;
-    a->a_lineno_off = a->a_offset;
     return 1;
 }
 
+/* Appends a range to the end of the line number table. See
+ *  Objects/lnotab_notes.txt for the description of the line number table. */
+
+static int
+assemble_line_range(struct assembler *a)
+{
+    int ldelta, bdelta;
+    bdelta =  (a->a_offset - a->a_lineno_start) * 2;
+    if (bdelta == 0) {
+        return 1;
+    }
+    if (a->a_lineno < 0) {
+        ldelta = -128;
+    }
+    else {
+        ldelta = a->a_lineno - a->a_prevlineno;
+        a->a_prevlineno = a->a_lineno;
+        while (ldelta > 127) {
+            if (!assemble_emit_linetable_pair(a, 0, 127)) {
+                return 0;
+            }
+            ldelta -= 127;
+        }
+        while (ldelta < -127) {
+            if (!assemble_emit_linetable_pair(a, 0, -127)) {
+                return 0;
+            }
+            ldelta += 127;
+        }
+    }
+    assert(-128 <= ldelta && ldelta < 128);
+    while (bdelta > 254) {
+        if (!assemble_emit_linetable_pair(a, 254, ldelta)) {
+            return 0;
+        }
+        ldelta = a->a_lineno < 0 ? -128 : 0;
+        bdelta -= 254;
+    }
+    if (!assemble_emit_linetable_pair(a, bdelta, ldelta)) {
+        return 0;
+    }
+    a->a_lineno_start = a->a_offset;
+    return 1;
+}
+
+static int
+assemble_lnotab(struct assembler *a, struct instr *i)
+{
+    if (i->i_lineno == a->a_lineno) {
+        return 1;
+    }
+    if (!assemble_line_range(a)) {
+        return 0;
+    }
+    a->a_lineno = i->i_lineno;
+    return 1;
+}
+
+
 /* assemble_emit()
    Extend the bytecode with a new instruction.
    Update lnotab if necessary.
@@ -5998,7 +5975,7 @@ assemble(struct compiler *c, int addNone)
        block ends with a jump or return b_next shouldn't set.
      */
     if (!c->u->u_curblock->b_return) {
-        NEXT_BLOCK(c);
+        c->u->u_lineno = -1;
         if (addNone)
             ADDOP_LOAD_CONST(c, Py_None);
         ADDOP(c, RETURN_VALUE);
@@ -6015,7 +5992,7 @@ assemble(struct compiler *c, int addNone)
     if (!c->u->u_firstlineno) {
         if (entryblock && entryblock->b_instr && entryblock->b_instr->i_lineno)
             c->u->u_firstlineno = entryblock->b_instr->i_lineno;
-        else
+       else
             c->u->u_firstlineno = 1;
     }
     if (!assemble_init(&a, nblocks, c->u->u_firstlineno))
@@ -6040,6 +6017,13 @@ assemble(struct compiler *c, int addNone)
             if (!assemble_emit(&a, &b->b_instr[j]))
                 goto error;
     }
+    if (!assemble_line_range(&a)) {
+        return 0;
+    }
+    /* Emit sentinel at end of line number table */
+    if (!assemble_emit_linetable_pair(&a, 255, -128)) {
+        goto error;
+    }
 
     if (_PyBytes_Resize(&a.a_lnotab, a.a_lnotab_off) < 0)
         goto error;
@@ -6280,22 +6264,42 @@ static void
 clean_basic_block(basicblock *bb) {
     /* Remove NOPs and any code following a return or re-raise. */
     int dest = 0;
+    int prev_lineno = -1;
     for (int src = 0; src < bb->b_iused; src++) {
+        int lineno = bb->b_instr[src].i_lineno;
         switch(bb->b_instr[src].i_opcode) {
-            case NOP:
-                /* skip */
-                break;
             case RETURN_VALUE:
             case RERAISE:
                 bb->b_next = NULL;
                 bb->b_instr[dest] = bb->b_instr[src];
                 dest++;
                 goto end;
+            case NOP:
+            {
+                /* Eliminate no-op if it doesn't have a line number, or
+                 * if the next instruction has same line number or no line number, or
+                 * if the previous instruction had the same line number. */
+                if (lineno < 0) {
+                    break;
+                }
+                if (prev_lineno == lineno) {
+                    break;
+                }
+                if (src < bb->b_iused - 1) {
+                    int next_lineno = bb->b_instr[src+1].i_lineno;
+                    if (next_lineno < 0 || next_lineno == lineno) {
+                        bb->b_instr[src+1].i_lineno = lineno;
+                        break;
+                    }
+                }
+            }
+            /* fallthrough */
             default:
                 if (dest != src) {
                     bb->b_instr[dest] = bb->b_instr[src];
                 }
                 dest++;
+                prev_lineno = lineno;
                 break;
         }
     }

Python/marshal.c

@@ -524,7 +524,7 @@ w_complex_object(PyObject *v, char flag, WFILE *p)
         w_object(co->co_filename, p);
         w_object(co->co_name, p);
         w_long(co->co_firstlineno, p);
-        w_object(co->co_lnotab, p);
+        w_object(co->co_linetable, p);
     }
     else if (PyObject_CheckBuffer(v)) {
         /* Write unknown bytes-like objects as a bytes object */
@@ -1312,7 +1312,7 @@ r_object(RFILE *p)
             PyObject *filename = NULL;
             PyObject *name = NULL;
             int firstlineno;
-            PyObject *lnotab = NULL;
+            PyObject *linetable = NULL;
 
             idx = r_ref_reserve(flag, p);
             if (idx < 0)
@@ -1367,8 +1367,8 @@ r_object(RFILE *p)
             firstlineno = (int)r_long(p);
             if (firstlineno == -1 && PyErr_Occurred())
                 break;
-            lnotab = r_object(p);
+            linetable = r_object(p);
-            if (lnotab == NULL)
+            if (linetable == NULL)
                 goto code_error;
 
             if (PySys_Audit("code.__new__", "OOOiiiiii",
@@ -1382,7 +1382,7 @@ r_object(RFILE *p)
                             nlocals, stacksize, flags,
                             code, consts, names, varnames,
                             freevars, cellvars, filename, name,
-                            firstlineno, lnotab);
+                            firstlineno, linetable);
             v = r_ref_insert(v, idx, flag, p);
 
           code_error:
@@ -1394,7 +1394,7 @@ r_object(RFILE *p)
             Py_XDECREF(cellvars);
             Py_XDECREF(filename);
             Py_XDECREF(name);
-            Py_XDECREF(lnotab);
+            Py_XDECREF(linetable);
         }
         retval = v;
         break;

Tools/gdb/libpython.py

@@ -648,19 +648,28 @@ class PyCodeObjectPtr(PyObjectPtr):
         Analogous to PyCode_Addr2Line; translated from pseudocode in
         Objects/lnotab_notes.txt
         '''
-        co_lnotab = self.pyop_field('co_lnotab').proxyval(set())
+        co_linetable = self.pyop_field('co_linetable').proxyval(set())
 
         # Initialize lineno to co_firstlineno as per PyCode_Addr2Line
         # not 0, as lnotab_notes.txt has it:
         lineno = int_from_int(self.field('co_firstlineno'))
 
+        if addrq < 0:
+            return lineno
         addr = 0
-        for addr_incr, line_incr in zip(co_lnotab[::2], co_lnotab[1::2]):
+        for addr_incr, line_incr in zip(co_linetable[::2], co_linetable[1::2]):
+            if addr_incr == 255:
+                break
             addr += ord(addr_incr)
+            line_delta = ord(line_incr)
+            if line_delta == 128:
+                line_delta = 0
+            elif line_delta > 128:
+                line_delta -= 256
+            lineno += line_delta
             if addr > addrq:
                 return lineno
-            lineno += ord(line_incr)
+        assert False, "Unreachable"
-        return lineno
 
 
 class PyDictObjectPtr(PyObjectPtr):