bpo-42827: Fix crash on SyntaxError in multiline expressions (GH-24140)

When trying to extract the error line for the error message there
are two distinct cases:

1. The input comes from a file, which means that we can extract the
   error line by using `PyErr_ProgramTextObject` and which we already
   do.
2. The input does not come from a file, at which point we need to get
   the source code from the tokenizer:
   * If the tokenizer's current line number is the same with the line
     of the error, we get the line from `tok->buf` and we're ready.
   * Else, we can extract the error line from the source code in the
     following two ways:
     * If the input comes from a string we have all the input
       in `tok->str` and we can extract the error line from it.
     * If the input comes from stdin, i.e. the interactive prompt, we
       do not have access to the previous line. That's why a new
       field `tok->stdin_content` is added which holds the whole input for the
       current (multiline) statement or expression. We can then extract the
       error line from `tok->stdin_content` like we do in the string case above.

Co-authored-by: Pablo Galindo <Pablogsal@gmail.com>

3 files changed, 59 insertions, 2 deletions

diff --git a/Parser/pegen.c b/Parser/pegen.c
index 188fd282b76..a6f97929255 100644
--- a/Parser/pegen.c
+++ b/Parser/pegen.c
@@ -380,6 +380,27 @@ _PyPegen_raise_error(Parser *p, PyObject *errtype, const char *errmsg, ...)
     return NULL;
 }
 
+static PyObject *
+get_error_line(Parser *p, Py_ssize_t lineno)
+{
+    /* If p->tok->fp == NULL, then we're parsing from a string, which means that
+       the whole source is stored in p->tok->str. If not, then we're parsing
+       from the REPL, so the source lines of the current (multi-line) statement
+       are stored in p->tok->stdin_content */
+    assert(p->tok->fp == NULL || p->tok->fp == stdin);
+
+    char *cur_line = p->tok->fp == NULL ? p->tok->str : p->tok->stdin_content;
+    for (int i = 0; i < lineno - 1; i++) {
+        cur_line = strchr(cur_line, '\n') + 1;
+    }
+
+    char *next_newline;
+    if ((next_newline = strchr(cur_line, '\n')) == NULL) { // This is the last line
+        next_newline = cur_line + strlen(cur_line);
+    }
+    return PyUnicode_DecodeUTF8(cur_line, next_newline - cur_line, "replace");
+}
+
 void *
 _PyPegen_raise_error_known_location(Parser *p, PyObject *errtype,
                                     Py_ssize_t lineno, Py_ssize_t col_offset,
@@ -416,8 +437,22 @@ _PyPegen_raise_error_known_location(Parser *p, PyObject *errtype,
     }
 
     if (!error_line) {
-        Py_ssize_t size = p->tok->inp - p->tok->buf;
-        error_line = PyUnicode_DecodeUTF8(p->tok->buf, size, "replace");
+        /* PyErr_ProgramTextObject was not called or returned NULL. If it was not called,
+           then we need to find the error line from some other source, because
+           p->start_rule != Py_file_input. If it returned NULL, then it either unexpectedly
+           failed or we're parsing from a string or the REPL. There's a third edge case where
+           we're actually parsing from a file, which has an E_EOF SyntaxError and in that case
+           `PyErr_ProgramTextObject` fails because lineno points to last_file_line + 1, which
+           does not physically exist */
+        assert(p->tok->fp == NULL || p->tok->fp == stdin || p->tok->done == E_EOF);
+
+        if (p->tok->lineno == lineno) {
+            Py_ssize_t size = p->tok->inp - p->tok->buf;
+            error_line = PyUnicode_DecodeUTF8(p->tok->buf, size, "replace");
+        }
+        else {
+            error_line = get_error_line(p, lineno);
+        }
         if (!error_line) {
             goto error;
         }
diff --git a/Parser/tokenizer.c b/Parser/tokenizer.c
index 96539bd5565..62cd2966231 100644
--- a/Parser/tokenizer.c
+++ b/Parser/tokenizer.c
@@ -81,6 +81,7 @@ tok_new(void)
     tok->decoding_readline = NULL;
     tok->decoding_buffer = NULL;
     tok->type_comments = 0;
+    tok->stdin_content = NULL;
 
     tok->async_hacks = 0;
     tok->async_def = 0;
@@ -816,6 +817,8 @@ PyTokenizer_Free(struct tok_state *tok)
         PyMem_Free(tok->buf);
     if (tok->input)
         PyMem_Free(tok->input);
+    if (tok->stdin_content)
+        PyMem_Free(tok->stdin_content);
     PyMem_Free(tok);
 }
 
@@ -856,6 +859,24 @@ tok_nextc(struct tok_state *tok)
                 if (translated == NULL)
                     return EOF;
                 newtok = translated;
+                if (tok->stdin_content == NULL) {
+                    tok->stdin_content = PyMem_Malloc(strlen(translated) + 1);
+                    if (tok->stdin_content == NULL) {
+                        tok->done = E_NOMEM;
+                        return EOF;
+                    }
+                    sprintf(tok->stdin_content, "%s", translated);
+                }
+                else {
+                    char *new_str = PyMem_Malloc(strlen(tok->stdin_content) + strlen(translated) + 1);
+                    if (new_str == NULL) {
+                        tok->done = E_NOMEM;
+                        return EOF;
+                    }
+                    sprintf(new_str, "%s%s", tok->stdin_content, translated);
+                    PyMem_Free(tok->stdin_content);
+                    tok->stdin_content = new_str;
+                }
             }
             if (tok->encoding && newtok && *newtok) {
                 /* Recode to UTF-8 */
diff --git a/Parser/tokenizer.h b/Parser/tokenizer.h
index 5660ea38e94..b659f34796e 100644
--- a/Parser/tokenizer.h
+++ b/Parser/tokenizer.h
@@ -37,6 +37,7 @@ struct tok_state {
     int atbol;          /* Nonzero if at begin of new line */
     int pendin;         /* Pending indents (if > 0) or dedents (if < 0) */
     const char *prompt, *nextprompt;          /* For interactive prompting */
+    char *stdin_content;
     int lineno;         /* Current line number */
     int first_lineno;   /* First line of a single line or multi line string
                            expression (cf. issue 16806) */