summaryrefslogtreecommitdiffstats
path: root/Parser
diff options
context:
space:
mode:
authorPablo Galindo Salgado <Pablogsal@gmail.com>2021-11-20 14:36:07 (GMT)
committerGitHub <noreply@github.com>2021-11-20 14:36:07 (GMT)
commitfdcc46d9554094994f78bedf6dc9220e5d5ee668 (patch)
tree4f8dbf3fc78bf76311d63c84a5c2384c9fd089d9 /Parser
parent6d430ef5ab62158a200b94dff31b89524a9576bb (diff)
downloadcpython-fdcc46d9554094994f78bedf6dc9220e5d5ee668.zip
cpython-fdcc46d9554094994f78bedf6dc9220e5d5ee668.tar.gz
cpython-fdcc46d9554094994f78bedf6dc9220e5d5ee668.tar.bz2
bpo-45848: Allow the parser to get error lines from encoded files (GH-29646)
Diffstat (limited to 'Parser')
-rw-r--r--Parser/pegen.c15
1 files changed, 8 insertions, 7 deletions
diff --git a/Parser/pegen.c b/Parser/pegen.c
index 09c1a19..b3fdae4 100644
--- a/Parser/pegen.c
+++ b/Parser/pegen.c
@@ -482,14 +482,12 @@ _PyPegen_raise_error_known_location(Parser *p, PyObject *errtype,
goto error;
}
- // PyErr_ProgramTextObject assumes that the text is utf-8 so we cannot call it with a file
- // with an arbitrary encoding or otherwise we could get some badly decoded text.
- int uses_utf8_codec = (!p->tok->encoding || strcmp(p->tok->encoding, "utf-8") == 0);
if (p->tok->fp_interactive) {
error_line = get_error_line(p, lineno);
}
- else if (uses_utf8_codec && p->start_rule == Py_file_input) {
- error_line = PyErr_ProgramTextObject(p->tok->filename, (int) lineno);
+ else if (p->start_rule == Py_file_input) {
+ error_line = _PyErr_ProgramDecodedTextObject(p->tok->filename,
+ (int) lineno, p->tok->encoding);
}
if (!error_line) {
@@ -500,15 +498,18 @@ _PyPegen_raise_error_known_location(Parser *p, PyObject *errtype,
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 || !uses_utf8_codec);
+ assert(p->tok->fp == NULL || p->tok->fp == stdin || p->tok->done == E_EOF);
if (p->tok->lineno <= lineno && p->tok->inp > p->tok->buf) {
Py_ssize_t size = p->tok->inp - p->tok->buf;
error_line = PyUnicode_DecodeUTF8(p->tok->buf, size, "replace");
}
- else {
+ else if (p->tok->fp == NULL || p->tok->fp == stdin) {
error_line = get_error_line(p, lineno);
}
+ else {
+ error_line = PyUnicode_FromStringAndSize("", 0);
+ }
if (!error_line) {
goto error;
}