| 1 |
#include <Python.h> |
| 2 |
#include <errcode.h> |
| 3 |
|
| 4 |
#include "pycore_pyerrors.h" // _PyErr_ProgramDecodedTextObject() |
| 5 |
#include "lexer/state.h" |
| 6 |
#include "lexer/lexer.h" |
| 7 |
#include "pegen.h" |
| 8 |
|
| 9 |
// TOKENIZER ERRORS |
| 10 |
|
| 11 |
void |
| 12 |
_PyPegen_raise_tokenizer_init_error(PyObject *filename) |
| 13 |
{ |
| 14 |
if (!(PyErr_ExceptionMatches(PyExc_LookupError) |
| 15 |
|| PyErr_ExceptionMatches(PyExc_SyntaxError) |
| 16 |
|| PyErr_ExceptionMatches(PyExc_ValueError) |
| 17 |
|| PyErr_ExceptionMatches(PyExc_UnicodeDecodeError))) { |
| 18 |
return; |
| 19 |
} |
| 20 |
PyObject *errstr = NULL; |
| 21 |
PyObject *tuple = NULL; |
| 22 |
PyObject *type; |
| 23 |
PyObject *value; |
| 24 |
PyObject *tback; |
| 25 |
PyErr_Fetch(&type, &value, &tback); |
| 26 |
errstr = PyObject_Str(value); |
| 27 |
if (!errstr) { |
| 28 |
goto error; |
| 29 |
} |
| 30 |
|
| 31 |
PyObject *tmp = Py_BuildValue("(OiiO)", filename, 0, -1, Py_None); |
| 32 |
if (!tmp) { |
| 33 |
goto error; |
| 34 |
} |
| 35 |
|
| 36 |
tuple = PyTuple_Pack(2, errstr, tmp); |
| 37 |
Py_DECREF(tmp); |
| 38 |
if (!tuple) { |
| 39 |
goto error; |
| 40 |
} |
| 41 |
PyErr_SetObject(PyExc_SyntaxError, tuple); |
| 42 |
|
| 43 |
error: |
| 44 |
Py_XDECREF(type); |
| 45 |
Py_XDECREF(value); |
| 46 |
Py_XDECREF(tback); |
| 47 |
Py_XDECREF(errstr); |
| 48 |
Py_XDECREF(tuple); |
| 49 |
} |
| 50 |
|
| 51 |
static inline void |
| 52 |
raise_unclosed_parentheses_error(Parser *p) { |
| 53 |
int error_lineno = p->tok->parenlinenostack[p->tok->level-1]; |
| 54 |
int error_col = p->tok->parencolstack[p->tok->level-1]; |
| 55 |
RAISE_ERROR_KNOWN_LOCATION(p, PyExc_SyntaxError, |
| 56 |
error_lineno, error_col, error_lineno, -1, |
| 57 |
"'%c' was never closed", |
| 58 |
p->tok->parenstack[p->tok->level-1]); |
| 59 |
} |
| 60 |
|
| 61 |
int |
| 62 |
_Pypegen_tokenizer_error(Parser *p) |
| 63 |
{ |
| 64 |
if (PyErr_Occurred()) { |
| 65 |
return -1; |
| 66 |
} |
| 67 |
|
| 68 |
const char *msg = NULL; |
| 69 |
PyObject* errtype = PyExc_SyntaxError; |
| 70 |
Py_ssize_t col_offset = -1; |
| 71 |
p->error_indicator = 1; |
| 72 |
switch (p->tok->done) { |
| 73 |
case E_TOKEN: |
| 74 |
msg = "invalid token"; |
| 75 |
break; |
| 76 |
case E_EOF: |
| 77 |
if (p->tok->level) { |
| 78 |
raise_unclosed_parentheses_error(p); |
| 79 |
} else { |
| 80 |
RAISE_SYNTAX_ERROR("unexpected EOF while parsing"); |
| 81 |
} |
| 82 |
return -1; |
| 83 |
case E_DEDENT: |
| 84 |
RAISE_INDENTATION_ERROR("unindent does not match any outer indentation level"); |
| 85 |
return -1; |
| 86 |
case E_INTR: |
| 87 |
if (!PyErr_Occurred()) { |
| 88 |
PyErr_SetNone(PyExc_KeyboardInterrupt); |
| 89 |
} |
| 90 |
return -1; |
| 91 |
case E_NOMEM: |
| 92 |
PyErr_NoMemory(); |
| 93 |
return -1; |
| 94 |
case E_TABSPACE: |
| 95 |
errtype = PyExc_TabError; |
| 96 |
msg = "inconsistent use of tabs and spaces in indentation"; |
| 97 |
break; |
| 98 |
case E_TOODEEP: |
| 99 |
errtype = PyExc_IndentationError; |
| 100 |
msg = "too many levels of indentation"; |
| 101 |
break; |
| 102 |
case E_LINECONT: { |
| 103 |
col_offset = p->tok->cur - p->tok->buf - 1; |
| 104 |
msg = "unexpected character after line continuation character"; |
| 105 |
break; |
| 106 |
} |
| 107 |
case E_COLUMNOVERFLOW: |
| 108 |
PyErr_SetString(PyExc_OverflowError, |
| 109 |
"Parser column offset overflow - source line is too big"); |
| 110 |
return -1; |
| 111 |
default: |
| 112 |
msg = "unknown parsing error"; |
| 113 |
} |
| 114 |
|
| 115 |
RAISE_ERROR_KNOWN_LOCATION(p, errtype, p->tok->lineno, |
| 116 |
col_offset >= 0 ? col_offset : 0, |
| 117 |
p->tok->lineno, -1, msg); |
| 118 |
return -1; |
| 119 |
} |
| 120 |
|
| 121 |
int |
| 122 |
_Pypegen_raise_decode_error(Parser *p) |
| 123 |
{ |
| 124 |
assert(PyErr_Occurred()); |
| 125 |
const char *errtype = NULL; |
| 126 |
if (PyErr_ExceptionMatches(PyExc_UnicodeError)) { |
| 127 |
errtype = "unicode error"; |
| 128 |
} |
| 129 |
else if (PyErr_ExceptionMatches(PyExc_ValueError)) { |
| 130 |
errtype = "value error"; |
| 131 |
} |
| 132 |
if (errtype) { |
| 133 |
PyObject *type; |
| 134 |
PyObject *value; |
| 135 |
PyObject *tback; |
| 136 |
PyObject *errstr; |
| 137 |
PyErr_Fetch(&type, &value, &tback); |
| 138 |
errstr = PyObject_Str(value); |
| 139 |
if (errstr) { |
| 140 |
RAISE_SYNTAX_ERROR("(%s) %U", errtype, errstr); |
| 141 |
Py_DECREF(errstr); |
| 142 |
} |
| 143 |
else { |
| 144 |
PyErr_Clear(); |
| 145 |
RAISE_SYNTAX_ERROR("(%s) unknown error", errtype); |
| 146 |
} |
| 147 |
Py_XDECREF(type); |
| 148 |
Py_XDECREF(value); |
| 149 |
Py_XDECREF(tback); |
| 150 |
} |
| 151 |
|
| 152 |
return -1; |
| 153 |
} |
| 154 |
|
| 155 |
static int |
| 156 |
_PyPegen_tokenize_full_source_to_check_for_errors(Parser *p) { |
| 157 |
// Tokenize the whole input to see if there are any tokenization |
| 158 |
// errors such as mistmatching parentheses. These will get priority |
| 159 |
// over generic syntax errors only if the line number of the error is |
| 160 |
// before the one that we had for the generic error. |
| 161 |
|
| 162 |
// We don't want to tokenize to the end for interactive input |
| 163 |
if (p->tok->prompt != NULL) { |
| 164 |
return 0; |
| 165 |
} |
| 166 |
|
| 167 |
PyObject *type, *value, *traceback; |
| 168 |
PyErr_Fetch(&type, &value, &traceback); |
| 169 |
|
| 170 |
Token *current_token = p->known_err_token != NULL ? p->known_err_token : p->tokens[p->fill - 1]; |
| 171 |
Py_ssize_t current_err_line = current_token->lineno; |
| 172 |
|
| 173 |
int ret = 0; |
| 174 |
struct token new_token; |
| 175 |
_PyToken_Init(&new_token); |
| 176 |
|
| 177 |
for (;;) { |
| 178 |
switch (_PyTokenizer_Get(p->tok, &new_token)) { |
| 179 |
case ERRORTOKEN: |
| 180 |
if (PyErr_Occurred()) { |
| 181 |
ret = -1; |
| 182 |
goto exit; |
| 183 |
} |
| 184 |
if (p->tok->level != 0) { |
| 185 |
int error_lineno = p->tok->parenlinenostack[p->tok->level-1]; |
| 186 |
if (current_err_line > error_lineno) { |
| 187 |
raise_unclosed_parentheses_error(p); |
| 188 |
ret = -1; |
| 189 |
goto exit; |
| 190 |
} |
| 191 |
} |
| 192 |
break; |
| 193 |
case ENDMARKER: |
| 194 |
break; |
| 195 |
default: |
| 196 |
continue; |
| 197 |
} |
| 198 |
break; |
| 199 |
} |
| 200 |
|
| 201 |
|
| 202 |
exit: |
| 203 |
_PyToken_Free(&new_token); |
| 204 |
// If we're in an f-string, we want the syntax error in the expression part |
| 205 |
// to propagate, so that tokenizer errors (like expecting '}') that happen afterwards |
| 206 |
// do not swallow it. |
| 207 |
if (PyErr_Occurred() && p->tok->tok_mode_stack_index <= 0) { |
| 208 |
Py_XDECREF(value); |
| 209 |
Py_XDECREF(type); |
| 210 |
Py_XDECREF(traceback); |
| 211 |
} else { |
| 212 |
PyErr_Restore(type, value, traceback); |
| 213 |
} |
| 214 |
return ret; |
| 215 |
} |
| 216 |
|
| 217 |
// PARSER ERRORS |
| 218 |
|
| 219 |
void * |
| 220 |
_PyPegen_raise_error(Parser *p, PyObject *errtype, int use_mark, const char *errmsg, ...) |
| 221 |
{ |
| 222 |
// Bail out if we already have an error set. |
| 223 |
if (p->error_indicator && PyErr_Occurred()) { |
| 224 |
return NULL; |
| 225 |
} |
| 226 |
if (p->fill == 0) { |
| 227 |
va_list va; |
| 228 |
va_start(va, errmsg); |
| 229 |
_PyPegen_raise_error_known_location(p, errtype, 0, 0, 0, -1, errmsg, va); |
| 230 |
va_end(va); |
| 231 |
return NULL; |
| 232 |
} |
| 233 |
if (use_mark && p->mark == p->fill && _PyPegen_fill_token(p) < 0) { |
| 234 |
p->error_indicator = 1; |
| 235 |
return NULL; |
| 236 |
} |
| 237 |
Token *t = p->known_err_token != NULL |
| 238 |
? p->known_err_token |
| 239 |
: p->tokens[use_mark ? p->mark : p->fill - 1]; |
| 240 |
Py_ssize_t col_offset; |
| 241 |
Py_ssize_t end_col_offset = -1; |
| 242 |
if (t->col_offset == -1) { |
| 243 |
if (p->tok->cur == p->tok->buf) { |
| 244 |
col_offset = 0; |
| 245 |
} else { |
| 246 |
const char* start = p->tok->buf ? p->tok->line_start : p->tok->buf; |
| 247 |
col_offset = Py_SAFE_DOWNCAST(p->tok->cur - start, intptr_t, int); |
| 248 |
} |
| 249 |
} else { |
| 250 |
col_offset = t->col_offset + 1; |
| 251 |
} |
| 252 |
|
| 253 |
if (t->end_col_offset != -1) { |
| 254 |
end_col_offset = t->end_col_offset + 1; |
| 255 |
} |
| 256 |
|
| 257 |
va_list va; |
| 258 |
va_start(va, errmsg); |
| 259 |
_PyPegen_raise_error_known_location(p, errtype, t->lineno, col_offset, t->end_lineno, end_col_offset, errmsg, va); |
| 260 |
va_end(va); |
| 261 |
|
| 262 |
return NULL; |
| 263 |
} |
| 264 |
|
| 265 |
static PyObject * |
| 266 |
get_error_line_from_tokenizer_buffers(Parser *p, Py_ssize_t lineno) |
| 267 |
{ |
| 268 |
/* If the file descriptor is interactive, the source lines of the current |
| 269 |
* (multi-line) statement are stored in p->tok->interactive_src_start. |
| 270 |
* If not, we're parsing from a string, which means that the whole source |
| 271 |
* is stored in p->tok->str. */ |
| 272 |
assert((p->tok->fp == NULL && p->tok->str != NULL) || p->tok->fp != NULL); |
| 273 |
|
| 274 |
char *cur_line = p->tok->fp_interactive ? p->tok->interactive_src_start : p->tok->str; |
| 275 |
if (cur_line == NULL) { |
| 276 |
assert(p->tok->fp_interactive); |
| 277 |
// We can reach this point if the tokenizer buffers for interactive source have not been |
| 278 |
// initialized because we failed to decode the original source with the given locale. |
| 279 |
return PyUnicode_FromStringAndSize("", 0); |
| 280 |
} |
| 281 |
|
| 282 |
Py_ssize_t relative_lineno = p->starting_lineno ? lineno - p->starting_lineno + 1 : lineno; |
| 283 |
const char* buf_end = p->tok->fp_interactive ? p->tok->interactive_src_end : p->tok->inp; |
| 284 |
|
| 285 |
if (buf_end < cur_line) { |
| 286 |
buf_end = cur_line + strlen(cur_line); |
| 287 |
} |
| 288 |
|
| 289 |
for (int i = 0; i < relative_lineno - 1; i++) { |
| 290 |
char *new_line = strchr(cur_line, '\n'); |
| 291 |
// The assert is here for debug builds but the conditional that |
| 292 |
// follows is there so in release builds we do not crash at the cost |
| 293 |
// to report a potentially wrong line. |
| 294 |
assert(new_line != NULL && new_line + 1 < buf_end); |
| 295 |
if (new_line == NULL || new_line + 1 > buf_end) { |
| 296 |
break; |
| 297 |
} |
| 298 |
cur_line = new_line + 1; |
| 299 |
} |
| 300 |
|
| 301 |
char *next_newline; |
| 302 |
if ((next_newline = strchr(cur_line, '\n')) == NULL) { // This is the last line |
| 303 |
next_newline = cur_line + strlen(cur_line); |
| 304 |
} |
| 305 |
return PyUnicode_DecodeUTF8(cur_line, next_newline - cur_line, "replace"); |
| 306 |
} |
| 307 |
|
| 308 |
void * |
| 309 |
_PyPegen_raise_error_known_location(Parser *p, PyObject *errtype, |
| 310 |
Py_ssize_t lineno, Py_ssize_t col_offset, |
| 311 |
Py_ssize_t end_lineno, Py_ssize_t end_col_offset, |
| 312 |
const char *errmsg, va_list va) |
| 313 |
{ |
| 314 |
// Bail out if we already have an error set. |
| 315 |
if (p->error_indicator && PyErr_Occurred()) { |
| 316 |
return NULL; |
| 317 |
} |
| 318 |
PyObject *value = NULL; |
| 319 |
PyObject *errstr = NULL; |
| 320 |
PyObject *error_line = NULL; |
| 321 |
PyObject *tmp = NULL; |
| 322 |
p->error_indicator = 1; |
| 323 |
|
| 324 |
if (end_lineno == CURRENT_POS) { |
| 325 |
end_lineno = p->tok->lineno; |
| 326 |
} |
| 327 |
if (end_col_offset == CURRENT_POS) { |
| 328 |
end_col_offset = p->tok->cur - p->tok->line_start; |
| 329 |
} |
| 330 |
|
| 331 |
errstr = PyUnicode_FromFormatV(errmsg, va); |
| 332 |
if (!errstr) { |
| 333 |
goto error; |
| 334 |
} |
| 335 |
|
| 336 |
if (p->tok->fp_interactive && p->tok->interactive_src_start != NULL) { |
| 337 |
error_line = get_error_line_from_tokenizer_buffers(p, lineno); |
| 338 |
} |
| 339 |
else if (p->start_rule == Py_file_input) { |
| 340 |
error_line = _PyErr_ProgramDecodedTextObject(p->tok->filename, |
| 341 |
(int) lineno, p->tok->encoding); |
| 342 |
} |
| 343 |
|
| 344 |
if (!error_line) { |
| 345 |
/* PyErr_ProgramTextObject was not called or returned NULL. If it was not called, |
| 346 |
then we need to find the error line from some other source, because |
| 347 |
p->start_rule != Py_file_input. If it returned NULL, then it either unexpectedly |
| 348 |
failed or we're parsing from a string or the REPL. There's a third edge case where |
| 349 |
we're actually parsing from a file, which has an E_EOF SyntaxError and in that case |
| 350 |
`PyErr_ProgramTextObject` fails because lineno points to last_file_line + 1, which |
| 351 |
does not physically exist */ |
| 352 |
assert(p->tok->fp == NULL || p->tok->fp == stdin || p->tok->done == E_EOF); |
| 353 |
|
| 354 |
if (p->tok->lineno <= lineno && p->tok->inp > p->tok->buf) { |
| 355 |
Py_ssize_t size = p->tok->inp - p->tok->line_start; |
| 356 |
error_line = PyUnicode_DecodeUTF8(p->tok->line_start, size, "replace"); |
| 357 |
} |
| 358 |
else if (p->tok->fp == NULL || p->tok->fp == stdin) { |
| 359 |
error_line = get_error_line_from_tokenizer_buffers(p, lineno); |
| 360 |
} |
| 361 |
else { |
| 362 |
error_line = PyUnicode_FromStringAndSize("", 0); |
| 363 |
} |
| 364 |
if (!error_line) { |
| 365 |
goto error; |
| 366 |
} |
| 367 |
} |
| 368 |
|
| 369 |
Py_ssize_t col_number = col_offset; |
| 370 |
Py_ssize_t end_col_number = end_col_offset; |
| 371 |
|
| 372 |
col_number = _PyPegen_byte_offset_to_character_offset(error_line, col_offset); |
| 373 |
if (col_number < 0) { |
| 374 |
goto error; |
| 375 |
} |
| 376 |
|
| 377 |
if (end_col_offset > 0) { |
| 378 |
end_col_number = _PyPegen_byte_offset_to_character_offset(error_line, end_col_offset); |
| 379 |
if (end_col_number < 0) { |
| 380 |
goto error; |
| 381 |
} |
| 382 |
} |
| 383 |
|
| 384 |
tmp = Py_BuildValue("(OnnNnn)", p->tok->filename, lineno, col_number, error_line, end_lineno, end_col_number); |
| 385 |
if (!tmp) { |
| 386 |
goto error; |
| 387 |
} |
| 388 |
value = PyTuple_Pack(2, errstr, tmp); |
| 389 |
Py_DECREF(tmp); |
| 390 |
if (!value) { |
| 391 |
goto error; |
| 392 |
} |
| 393 |
PyErr_SetObject(errtype, value); |
| 394 |
|
| 395 |
Py_DECREF(errstr); |
| 396 |
Py_DECREF(value); |
| 397 |
return NULL; |
| 398 |
|
| 399 |
error: |
| 400 |
Py_XDECREF(errstr); |
| 401 |
Py_XDECREF(error_line); |
| 402 |
return NULL; |
| 403 |
} |
| 404 |
|
| 405 |
void |
| 406 |
_Pypegen_set_syntax_error(Parser* p, Token* last_token) { |
| 407 |
// Existing syntax error |
| 408 |
if (PyErr_Occurred()) { |
| 409 |
// Prioritize tokenizer errors to custom syntax errors raised |
| 410 |
// on the second phase only if the errors come from the parser. |
| 411 |
int is_tok_ok = (p->tok->done == E_DONE || p->tok->done == E_OK); |
| 412 |
if (is_tok_ok && PyErr_ExceptionMatches(PyExc_SyntaxError)) { |
| 413 |
_PyPegen_tokenize_full_source_to_check_for_errors(p); |
| 414 |
} |
| 415 |
// Propagate the existing syntax error. |
| 416 |
return; |
| 417 |
} |
| 418 |
// Initialization error |
| 419 |
if (p->fill == 0) { |
| 420 |
RAISE_SYNTAX_ERROR("error at start before reading any input"); |
| 421 |
} |
| 422 |
// Parser encountered EOF (End of File) unexpectedtly |
| 423 |
if (last_token->type == ERRORTOKEN && p->tok->done == E_EOF) { |
| 424 |
if (p->tok->level) { |
| 425 |
raise_unclosed_parentheses_error(p); |
| 426 |
} else { |
| 427 |
RAISE_SYNTAX_ERROR("unexpected EOF while parsing"); |
| 428 |
} |
| 429 |
return; |
| 430 |
} |
| 431 |
// Indentation error in the tokenizer |
| 432 |
if (last_token->type == INDENT || last_token->type == DEDENT) { |
| 433 |
RAISE_INDENTATION_ERROR(last_token->type == INDENT ? "unexpected indent" : "unexpected unindent"); |
| 434 |
return; |
| 435 |
} |
| 436 |
// Unknown error (generic case) |
| 437 |
|
| 438 |
// Use the last token we found on the first pass to avoid reporting |
| 439 |
// incorrect locations for generic syntax errors just because we reached |
| 440 |
// further away when trying to find specific syntax errors in the second |
| 441 |
// pass. |
| 442 |
RAISE_SYNTAX_ERROR_KNOWN_LOCATION(last_token, "invalid syntax"); |
| 443 |
// _PyPegen_tokenize_full_source_to_check_for_errors will override the existing |
| 444 |
// generic SyntaxError we just raised if errors are found. |
| 445 |
_PyPegen_tokenize_full_source_to_check_for_errors(p); |
| 446 |
} |
| 447 |
|
| 448 |
void |
| 449 |
_Pypegen_stack_overflow(Parser *p) |
| 450 |
{ |
| 451 |
p->error_indicator = 1; |
| 452 |
PyErr_SetString(PyExc_MemoryError, |
| 453 |
"Parser stack overflowed - Python source too complex to parse"); |
| 454 |
} |