mirror of
https://github.com/python/cpython.git
synced 2025-08-22 17:55:18 +00:00
gh-110721: Use the traceback module for PyErr_Display() and fallback to the C implementation (#110702)
This commit is contained in:
parent
8c6c14b91b
commit
e7331365b4
7 changed files with 146 additions and 837 deletions
|
@ -397,27 +397,9 @@ _Py_WriteIndent(int indent, PyObject *f)
|
|||
return 0;
|
||||
}
|
||||
|
||||
/* Writes indent spaces, followed by the margin if it is not `\0`.
|
||||
Returns 0 on success and non-zero on failure.
|
||||
*/
|
||||
int
|
||||
_Py_WriteIndentedMargin(int indent, const char *margin, PyObject *f)
|
||||
{
|
||||
if (_Py_WriteIndent(indent, f) < 0) {
|
||||
return -1;
|
||||
}
|
||||
if (margin) {
|
||||
if (PyFile_WriteString(margin, f) < 0) {
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
return 0;
|
||||
}
|
||||
|
||||
static int
|
||||
display_source_line_with_margin(PyObject *f, PyObject *filename, int lineno, int indent,
|
||||
int margin_indent, const char *margin,
|
||||
int *truncation, PyObject **line)
|
||||
display_source_line(PyObject *f, PyObject *filename, int lineno, int indent,
|
||||
int *truncation, PyObject **line)
|
||||
{
|
||||
int fd;
|
||||
int i;
|
||||
|
@ -545,10 +527,6 @@ display_source_line_with_margin(PyObject *f, PyObject *filename, int lineno, int
|
|||
*truncation = i - indent;
|
||||
}
|
||||
|
||||
if (_Py_WriteIndentedMargin(margin_indent, margin, f) < 0) {
|
||||
goto error;
|
||||
}
|
||||
|
||||
/* Write some spaces before the line */
|
||||
if (_Py_WriteIndent(indent, f) < 0) {
|
||||
goto error;
|
||||
|
@ -574,161 +552,11 @@ int
|
|||
_Py_DisplaySourceLine(PyObject *f, PyObject *filename, int lineno, int indent,
|
||||
int *truncation, PyObject **line)
|
||||
{
|
||||
return display_source_line_with_margin(f, filename, lineno, indent, 0,
|
||||
NULL, truncation, line);
|
||||
return display_source_line(f, filename, lineno, indent, truncation, line);
|
||||
}
|
||||
|
||||
/* AST based Traceback Specialization
|
||||
*
|
||||
* When displaying a new traceback line, for certain syntactical constructs
|
||||
* (e.g a subscript, an arithmetic operation) we try to create a representation
|
||||
* that separates the primary source of error from the rest.
|
||||
*
|
||||
* Example specialization of BinOp nodes:
|
||||
* Traceback (most recent call last):
|
||||
* File "/home/isidentical/cpython/cpython/t.py", line 10, in <module>
|
||||
* add_values(1, 2, 'x', 3, 4)
|
||||
* File "/home/isidentical/cpython/cpython/t.py", line 2, in add_values
|
||||
* return a + b + c + d + e
|
||||
* ~~~~~~^~~
|
||||
* TypeError: 'NoneType' object is not subscriptable
|
||||
*/
|
||||
|
||||
#define IS_WHITESPACE(c) (((c) == ' ') || ((c) == '\t') || ((c) == '\f'))
|
||||
|
||||
static int
|
||||
extract_anchors_from_expr(const char *segment_str, expr_ty expr, Py_ssize_t *left_anchor, Py_ssize_t *right_anchor,
|
||||
char** primary_error_char, char** secondary_error_char)
|
||||
{
|
||||
switch (expr->kind) {
|
||||
case BinOp_kind: {
|
||||
expr_ty left = expr->v.BinOp.left;
|
||||
expr_ty right = expr->v.BinOp.right;
|
||||
for (int i = left->end_col_offset; i < right->col_offset; i++) {
|
||||
if (IS_WHITESPACE(segment_str[i])) {
|
||||
continue;
|
||||
}
|
||||
|
||||
*left_anchor = i;
|
||||
*right_anchor = i + 1;
|
||||
|
||||
// Check whether if this a two-character operator (e.g //)
|
||||
if (i + 1 < right->col_offset && !IS_WHITESPACE(segment_str[i + 1])) {
|
||||
++*right_anchor;
|
||||
}
|
||||
|
||||
// Keep going if the current char is not ')'
|
||||
if (i+1 < right->col_offset && (segment_str[i] == ')')) {
|
||||
continue;
|
||||
}
|
||||
|
||||
// Set the error characters
|
||||
*primary_error_char = "~";
|
||||
*secondary_error_char = "^";
|
||||
break;
|
||||
}
|
||||
return 1;
|
||||
}
|
||||
case Subscript_kind: {
|
||||
*left_anchor = expr->v.Subscript.value->end_col_offset;
|
||||
*right_anchor = expr->v.Subscript.slice->end_col_offset + 1;
|
||||
Py_ssize_t str_len = strlen(segment_str);
|
||||
|
||||
// Move right_anchor and left_anchor forward to the first non-whitespace character that is not ']' and '['
|
||||
while (*left_anchor < str_len && (IS_WHITESPACE(segment_str[*left_anchor]) || segment_str[*left_anchor] != '[')) {
|
||||
++*left_anchor;
|
||||
}
|
||||
while (*right_anchor < str_len && (IS_WHITESPACE(segment_str[*right_anchor]) || segment_str[*right_anchor] != ']')) {
|
||||
++*right_anchor;
|
||||
}
|
||||
if (*right_anchor < str_len){
|
||||
*right_anchor += 1;
|
||||
}
|
||||
|
||||
// Set the error characters
|
||||
*primary_error_char = "~";
|
||||
*secondary_error_char = "^";
|
||||
return 1;
|
||||
}
|
||||
default:
|
||||
return 0;
|
||||
}
|
||||
}
|
||||
|
||||
static int
|
||||
extract_anchors_from_stmt(const char *segment_str, stmt_ty statement, Py_ssize_t *left_anchor, Py_ssize_t *right_anchor,
|
||||
char** primary_error_char, char** secondary_error_char)
|
||||
{
|
||||
switch (statement->kind) {
|
||||
case Expr_kind: {
|
||||
return extract_anchors_from_expr(segment_str, statement->v.Expr.value, left_anchor, right_anchor,
|
||||
primary_error_char, secondary_error_char);
|
||||
}
|
||||
default:
|
||||
return 0;
|
||||
}
|
||||
}
|
||||
|
||||
static int
|
||||
extract_anchors_from_line(PyObject *filename, PyObject *line,
|
||||
Py_ssize_t start_offset, Py_ssize_t end_offset,
|
||||
Py_ssize_t *left_anchor, Py_ssize_t *right_anchor,
|
||||
char** primary_error_char, char** secondary_error_char)
|
||||
{
|
||||
int res = -1;
|
||||
PyArena *arena = NULL;
|
||||
PyObject *segment = PyUnicode_Substring(line, start_offset, end_offset);
|
||||
if (!segment) {
|
||||
goto done;
|
||||
}
|
||||
|
||||
const char *segment_str = PyUnicode_AsUTF8(segment);
|
||||
if (!segment_str) {
|
||||
goto done;
|
||||
}
|
||||
|
||||
arena = _PyArena_New();
|
||||
if (!arena) {
|
||||
goto done;
|
||||
}
|
||||
|
||||
PyCompilerFlags flags = _PyCompilerFlags_INIT;
|
||||
|
||||
mod_ty module = _PyParser_ASTFromString(segment_str, filename, Py_file_input,
|
||||
&flags, arena);
|
||||
if (!module) {
|
||||
goto done;
|
||||
}
|
||||
if (!_PyAST_Optimize(module, arena, _Py_GetConfig()->optimization_level, 0)) {
|
||||
goto done;
|
||||
}
|
||||
|
||||
assert(module->kind == Module_kind);
|
||||
if (asdl_seq_LEN(module->v.Module.body) == 1) {
|
||||
stmt_ty statement = asdl_seq_GET(module->v.Module.body, 0);
|
||||
res = extract_anchors_from_stmt(segment_str, statement, left_anchor, right_anchor,
|
||||
primary_error_char, secondary_error_char);
|
||||
} else {
|
||||
res = 0;
|
||||
}
|
||||
|
||||
done:
|
||||
if (res > 0) {
|
||||
// Normalize the AST offsets to byte offsets and adjust them with the
|
||||
// start of the actual line (instead of the source code segment).
|
||||
assert(segment != NULL);
|
||||
assert(*left_anchor >= 0);
|
||||
assert(*right_anchor >= 0);
|
||||
*left_anchor = _PyPegen_byte_offset_to_character_offset(segment, *left_anchor) + start_offset;
|
||||
*right_anchor = _PyPegen_byte_offset_to_character_offset(segment, *right_anchor) + start_offset;
|
||||
}
|
||||
Py_XDECREF(segment);
|
||||
if (arena) {
|
||||
_PyArena_Free(arena);
|
||||
}
|
||||
return res;
|
||||
}
|
||||
|
||||
#define _TRACEBACK_SOURCE_LINE_INDENT 4
|
||||
|
||||
static inline int
|
||||
|
@ -742,42 +570,14 @@ ignore_source_errors(void) {
|
|||
return 0;
|
||||
}
|
||||
|
||||
static inline int
|
||||
print_error_location_carets(PyObject *f, int offset, Py_ssize_t start_offset, Py_ssize_t end_offset,
|
||||
Py_ssize_t right_start_offset, Py_ssize_t left_end_offset,
|
||||
const char *primary, const char *secondary) {
|
||||
int special_chars = (left_end_offset != -1 || right_start_offset != -1);
|
||||
const char *str;
|
||||
while (++offset <= end_offset) {
|
||||
if (offset <= start_offset) {
|
||||
str = " ";
|
||||
} else if (special_chars && left_end_offset < offset && offset <= right_start_offset) {
|
||||
str = secondary;
|
||||
} else {
|
||||
str = primary;
|
||||
}
|
||||
if (PyFile_WriteString(str, f) < 0) {
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
if (PyFile_WriteString("\n", f) < 0) {
|
||||
return -1;
|
||||
}
|
||||
return 0;
|
||||
}
|
||||
|
||||
static int
|
||||
tb_displayline(PyTracebackObject* tb, PyObject *f, PyObject *filename, int lineno,
|
||||
PyFrameObject *frame, PyObject *name, int margin_indent, const char *margin)
|
||||
PyFrameObject *frame, PyObject *name)
|
||||
{
|
||||
if (filename == NULL || name == NULL) {
|
||||
return -1;
|
||||
}
|
||||
|
||||
if (_Py_WriteIndentedMargin(margin_indent, margin, f) < 0) {
|
||||
return -1;
|
||||
}
|
||||
|
||||
PyObject *line = PyUnicode_FromFormat(" File \"%U\", line %d, in %U\n",
|
||||
filename, lineno, name);
|
||||
if (line == NULL) {
|
||||
|
@ -794,9 +594,9 @@ tb_displayline(PyTracebackObject* tb, PyObject *f, PyObject *filename, int linen
|
|||
|
||||
int truncation = _TRACEBACK_SOURCE_LINE_INDENT;
|
||||
PyObject* source_line = NULL;
|
||||
int rc = display_source_line_with_margin(
|
||||
int rc = display_source_line(
|
||||
f, filename, lineno, _TRACEBACK_SOURCE_LINE_INDENT,
|
||||
margin_indent, margin, &truncation, &source_line);
|
||||
&truncation, &source_line);
|
||||
if (rc != 0 || !source_line) {
|
||||
/* ignore errors since we can't report them, can we? */
|
||||
err = ignore_source_errors();
|
||||
|
@ -823,87 +623,19 @@ tb_displayline(PyTracebackObject* tb, PyObject *f, PyObject *filename, int linen
|
|||
goto done;
|
||||
}
|
||||
|
||||
// When displaying errors, we will use the following generic structure:
|
||||
//
|
||||
// ERROR LINE ERROR LINE ERROR LINE ERROR LINE ERROR LINE ERROR LINE ERROR LINE
|
||||
// ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^~~~~~~~~~~~~~~~~~~~
|
||||
// | |-> left_end_offset | |-> end_offset
|
||||
// |-> start_offset |-> right_start_offset
|
||||
//
|
||||
// In general we will only have (start_offset, end_offset) but we can gather more information
|
||||
// by analyzing the AST of the text between *start_offset* and *end_offset*. If this succeeds
|
||||
// we could get *left_end_offset* and *right_start_offset* and some selection of characters for
|
||||
// the different ranges (primary_error_char and secondary_error_char). If we cannot obtain the
|
||||
// AST information or we cannot identify special ranges within it, then left_end_offset and
|
||||
// right_end_offset will be set to -1.
|
||||
//
|
||||
// To keep the column indicators pertinent, they are not shown when the primary character
|
||||
// spans the whole line.
|
||||
|
||||
// Convert the utf-8 byte offset to the actual character offset so we print the right number of carets.
|
||||
assert(source_line);
|
||||
Py_ssize_t start_offset = _PyPegen_byte_offset_to_character_offset(source_line, start_col_byte_offset);
|
||||
if (start_offset < 0) {
|
||||
err = ignore_source_errors() < 0;
|
||||
// If this is a multi-line expression, then we will highlight until
|
||||
// the last non-whitespace character.
|
||||
const char *source_line_str = PyUnicode_AsUTF8(source_line);
|
||||
if (!source_line_str) {
|
||||
goto done;
|
||||
}
|
||||
|
||||
Py_ssize_t end_offset = _PyPegen_byte_offset_to_character_offset(source_line, end_col_byte_offset);
|
||||
if (end_offset < 0) {
|
||||
err = ignore_source_errors() < 0;
|
||||
goto done;
|
||||
}
|
||||
|
||||
Py_ssize_t left_end_offset = -1;
|
||||
Py_ssize_t right_start_offset = -1;
|
||||
|
||||
char *primary_error_char = "^";
|
||||
char *secondary_error_char = primary_error_char;
|
||||
|
||||
if (start_line == end_line) {
|
||||
int res = extract_anchors_from_line(filename, source_line, start_offset, end_offset,
|
||||
&left_end_offset, &right_start_offset,
|
||||
&primary_error_char, &secondary_error_char);
|
||||
if (res < 0 && ignore_source_errors() < 0) {
|
||||
goto done;
|
||||
Py_ssize_t i = source_line_len;
|
||||
while (--i >= 0) {
|
||||
if (!IS_WHITESPACE(source_line_str[i])) {
|
||||
break;
|
||||
}
|
||||
}
|
||||
else {
|
||||
// If this is a multi-line expression, then we will highlight until
|
||||
// the last non-whitespace character.
|
||||
const char *source_line_str = PyUnicode_AsUTF8(source_line);
|
||||
if (!source_line_str) {
|
||||
goto done;
|
||||
}
|
||||
|
||||
Py_ssize_t i = source_line_len;
|
||||
while (--i >= 0) {
|
||||
if (!IS_WHITESPACE(source_line_str[i])) {
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
end_offset = i + 1;
|
||||
}
|
||||
|
||||
// Elide indicators if primary char spans the frame line
|
||||
Py_ssize_t stripped_line_len = source_line_len - truncation - _TRACEBACK_SOURCE_LINE_INDENT;
|
||||
bool has_secondary_ranges = (left_end_offset != -1 || right_start_offset != -1);
|
||||
if (end_offset - start_offset == stripped_line_len && !has_secondary_ranges) {
|
||||
goto done;
|
||||
}
|
||||
|
||||
if (_Py_WriteIndentedMargin(margin_indent, margin, f) < 0) {
|
||||
err = -1;
|
||||
goto done;
|
||||
}
|
||||
|
||||
if (print_error_location_carets(f, truncation, start_offset, end_offset,
|
||||
right_start_offset, left_end_offset,
|
||||
primary_error_char, secondary_error_char) < 0) {
|
||||
err = -1;
|
||||
goto done;
|
||||
}
|
||||
|
||||
done:
|
||||
Py_XDECREF(source_line);
|
||||
|
@ -930,8 +662,7 @@ tb_print_line_repeated(PyObject *f, long cnt)
|
|||
}
|
||||
|
||||
static int
|
||||
tb_printinternal(PyTracebackObject *tb, PyObject *f, long limit,
|
||||
int indent, const char *margin)
|
||||
tb_printinternal(PyTracebackObject *tb, PyObject *f, long limit)
|
||||
{
|
||||
PyCodeObject *code = NULL;
|
||||
Py_ssize_t depth = 0;
|
||||
|
@ -967,7 +698,7 @@ tb_printinternal(PyTracebackObject *tb, PyObject *f, long limit,
|
|||
cnt++;
|
||||
if (cnt <= TB_RECURSIVE_CUTOFF) {
|
||||
if (tb_displayline(tb, f, code->co_filename, tb->tb_lineno,
|
||||
tb->tb_frame, code->co_name, indent, margin) < 0) {
|
||||
tb->tb_frame, code->co_name) < 0) {
|
||||
goto error;
|
||||
}
|
||||
|
||||
|
@ -992,8 +723,7 @@ error:
|
|||
#define PyTraceBack_LIMIT 1000
|
||||
|
||||
int
|
||||
_PyTraceBack_Print_Indented(PyObject *v, int indent, const char *margin,
|
||||
const char *header_margin, const char *header, PyObject *f)
|
||||
_PyTraceBack_Print(PyObject *v, const char *header, PyObject *f)
|
||||
{
|
||||
PyObject *limitv;
|
||||
long limit = PyTraceBack_LIMIT;
|
||||
|
@ -1016,15 +746,12 @@ _PyTraceBack_Print_Indented(PyObject *v, int indent, const char *margin,
|
|||
return 0;
|
||||
}
|
||||
}
|
||||
if (_Py_WriteIndentedMargin(indent, header_margin, f) < 0) {
|
||||
return -1;
|
||||
}
|
||||
|
||||
if (PyFile_WriteString(header, f) < 0) {
|
||||
return -1;
|
||||
}
|
||||
|
||||
if (tb_printinternal((PyTracebackObject *)v, f, limit, indent, margin) < 0) {
|
||||
if (tb_printinternal((PyTracebackObject *)v, f, limit) < 0) {
|
||||
return -1;
|
||||
}
|
||||
|
||||
|
@ -1034,12 +761,8 @@ _PyTraceBack_Print_Indented(PyObject *v, int indent, const char *margin,
|
|||
int
|
||||
PyTraceBack_Print(PyObject *v, PyObject *f)
|
||||
{
|
||||
int indent = 0;
|
||||
const char *margin = NULL;
|
||||
const char *header_margin = NULL;
|
||||
const char *header = EXCEPTION_TB_HEADER;
|
||||
|
||||
return _PyTraceBack_Print_Indented(v, indent, margin, header_margin, header, f);
|
||||
return _PyTraceBack_Print(v, header, f);
|
||||
}
|
||||
|
||||
/* Format an integer in range [0; 0xffffffff] to decimal and write it
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue