bpo-39219: Fix SyntaxError attributes in the tokenizer. (GH-17828)

serhiy-storchaka · web-flow · commit 0cc6b5e559b8 · 2020-02-12T12:17:00.000+02:00
* Always set the text attribute.
* Correct the offset attribute for non-ascii sources.
diff --git a/Lib/test/test_exceptions.py b/Lib/test/test_exceptions.py
@@ -179,17 +179,25 @@ def ckmsg(src, msg, exception=SyntaxError):
         ckmsg(s, "inconsistent use of tabs and spaces in indentation", TabError)
 
     def testSyntaxErrorOffset(self):
-        def check(src, lineno, offset):
+        def check(src, lineno, offset, encoding='utf-8'):
             with self.assertRaises(SyntaxError) as cm:
                 compile(src, '<fragment>', 'exec')
             self.assertEqual(cm.exception.lineno, lineno)
             self.assertEqual(cm.exception.offset, offset)
+            if cm.exception.text is not None:
+                if not isinstance(src, str):
+                    src = src.decode(encoding, 'replace')
+                line = src.split('\n')[lineno-1]
+                self.assertEqual(cm.exception.text.rstrip('\n'), line)
 
         check('def fact(x):\n\treturn x!\n', 2, 10)
         check('1 +\n', 1, 4)
         check('def spam():\n  print(1)\n print(2)', 3, 10)
         check('Python = "Python" +', 1, 20)
         check('Python = "\u1e54\xfd\u0163\u0125\xf2\xf1" +', 1, 20)
+        check(b'# -*- coding: cp1251 -*-\nPython = "\xcf\xb3\xf2\xee\xed" +',
+              2, 19, encoding='cp1251')
+        check(b'Python = "\xcf\xb3\xf2\xee\xed" +', 1, 18)
         check('x = "a', 1, 7)
         check('lambda x: x = 2', 1, 1)
 
@@ -205,6 +213,10 @@ def check(src, lineno, offset):
         check('0010 + 2', 1, 4)
         check('x = 32e-+4', 1, 8)
         check('x = 0o9', 1, 6)
+        check('\u03b1 = 0xI', 1, 6)
+        check(b'\xce\xb1 = 0xI', 1, 6)
+        check(b'# -*- coding: iso8859-7 -*-\n\xe1 = 0xI', 2, 6,
+              encoding='iso8859-7')
 
         # Errors thrown by symtable.c
         check('x = [(yield i) for i in range(3)]', 1, 5)
diff --git a/Misc/NEWS.d/next/Core and Builtins/2020-01-05-13-36-08.bpo-39219.uHtKd4.rst b/Misc/NEWS.d/next/Core and Builtins/2020-01-05-13-36-08.bpo-39219.uHtKd4.rst
@@ -0,0 +1,2 @@
+Syntax errors raised in the tokenizer now always set correct "text" and
+"offset" attributes.
diff --git a/Parser/tokenizer.c b/Parser/tokenizer.c
@@ -1,6 +1,7 @@
 
 /* Tokenizer implementation */
 
+#define PY_SSIZE_T_CLEAN
 #include "Python.h"
 
 #include <ctype.h>
@@ -1034,17 +1035,44 @@ tok_backup(struct tok_state *tok, int c)
 static int
 syntaxerror(struct tok_state *tok, const char *format, ...)
 {
+    PyObject *errmsg, *errtext, *args;
     va_list vargs;
 #ifdef HAVE_STDARG_PROTOTYPES
     va_start(vargs, format);
 #else
     va_start(vargs);
 #endif
-    PyErr_FormatV(PyExc_SyntaxError, format, vargs);
+    errmsg = PyUnicode_FromFormatV(format, vargs);
     va_end(vargs);
-    PyErr_SyntaxLocationObject(tok->filename,
-                               tok->lineno,
-                               (int)(tok->cur - tok->line_start));
+    if (!errmsg) {
+        goto error;
+    }
+
+    errtext = PyUnicode_DecodeUTF8(tok->line_start, tok->cur - tok->line_start,
+                                   "replace");
+    if (!errtext) {
+        goto error;
+    }
+    int offset = (int)PyUnicode_GET_LENGTH(errtext);
+    Py_ssize_t line_len = strcspn(tok->line_start, "\n");
+    if (line_len != tok->cur - tok->line_start) {
+        Py_DECREF(errtext);
+        errtext = PyUnicode_DecodeUTF8(tok->line_start, line_len,
+                                       "replace");
+    }
+    if (!errtext) {
+        goto error;
+    }
+
+    args = Py_BuildValue("(O(OiiN))", errmsg,
+                         tok->filename, tok->lineno, offset, errtext);
+    if (args) {
+        PyErr_SetObject(PyExc_SyntaxError, args);
+        Py_DECREF(args);
+    }
+
+error:
+    Py_XDECREF(errmsg);
     tok->done = E_ERROR;
     return ERRORTOKEN;
 }

Original file line number	Diff line number	Diff line change
`@@ -0,0 +1,2 @@`
	`1`	`+Syntax errors raised in the tokenizer now always set correct "text" and`
	`2`	`+"offset" attributes.`