gh-94360: Fix a tokenizer crash when reading encoded files with syntax errors from stdin (#94386)

pablogsal · ambv · web-flow · commit 36fcde61ba48 · 2022-07-05T17:39:21.000+01:00
* gh-94360: Fix a tokenizer crash when reading encoded files with syntax errors from stdin Signed-off-by: Pablo Galindo <pablogsal@gmail.com> * nitty nit Co-authored-by: Łukasz Langa <lukasz@langa.pl>
diff --git a/Misc/NEWS.d/next/Core and Builtins/2022-06-28-14-20-36.gh-issue-94360.DiEnen.rst b/Misc/NEWS.d/next/Core and Builtins/2022-06-28-14-20-36.gh-issue-94360.DiEnen.rst
@@ -0,0 +1,2 @@
+Fixed a tokenizer crash when reading encoded files with syntax errors from
+``stdin`` with non utf-8 encoded text. Patch by Pablo Galindo
diff --git a/Parser/pegen_errors.c b/Parser/pegen_errors.c
@@ -259,15 +259,15 @@ get_error_line_from_tokenizer_buffers(Parser *p, Py_ssize_t lineno)
     const char* buf_end = p->tok->fp_interactive ? p->tok->interactive_src_end : p->tok->inp;
 
     for (int i = 0; i < relative_lineno - 1; i++) {
-        char *new_line = strchr(cur_line, '\n') + 1;
+        char *new_line = strchr(cur_line, '\n');
         // The assert is here for debug builds but the conditional that
         // follows is there so in release builds we do not crash at the cost
         // to report a potentially wrong line.
-        assert(new_line != NULL && new_line <= buf_end);
-        if (new_line == NULL || new_line > buf_end) {
+        assert(new_line != NULL && new_line + 1 < buf_end);
+        if (new_line == NULL || new_line + 1 > buf_end) {
             break;
         }
-        cur_line = new_line;
+        cur_line = new_line + 1;
     }
 
     char *next_newline;
diff --git a/Parser/tokenizer.c b/Parser/tokenizer.c
@@ -308,6 +308,10 @@ tok_concatenate_interactive_new_line(struct tok_state *tok, const char *line) {
 
     Py_ssize_t current_size = tok->interactive_src_end - tok->interactive_src_start;
     Py_ssize_t line_size = strlen(line);
+    char last_char = line[line_size > 0 ? line_size - 1 : line_size];
+    if (last_char != '\n') {
+        line_size += 1;
+    }
     char* new_str = tok->interactive_src_start;
 
     new_str = PyMem_Realloc(new_str, current_size + line_size + 1);
@@ -321,7 +325,11 @@ tok_concatenate_interactive_new_line(struct tok_state *tok, const char *line) {
         return -1;
     }
     strcpy(new_str + current_size, line);
-
+    if (last_char != '\n') {
+        /* Last line does not end in \n, fake one */
+        new_str[current_size + line_size - 1] = '\n';
+        new_str[current_size + line_size] = '\0';
+    }
     tok->interactive_src_start = new_str;
     tok->interactive_src_end = new_str + current_size + line_size;
     return 0;

Original file line number	Diff line number	Diff line change
`@@ -0,0 +1,2 @@`
	`1`	`+Fixed a tokenizer crash when reading encoded files with syntax errors from`
	`2`	+``stdin`` with non utf-8 encoded text. Patch by Pablo Galindo