From 6b7fcff96999dcd420823ada7c097ac2b2daa25d Mon Sep 17 00:00:00 2001 From: Serhiy Storchaka Date: Sat, 22 Jun 2019 12:49:46 +0300 Subject: [PATCH 1/2] bpo-24214: Fixed the UTF-8 and UTF-16 incremental decoders. * The UTF-8 incremental decoders fails now fast if encounter a sequence that can't be handled by the error handler. * The UTF-16 incremental decoders with the surrogatepass error handler decodes now a lone low surrogate with final=False. --- Lib/test/test_codecs.py | 7 +++++++ .../2019-06-22-12-45-20.bpo-24214.hIiHeD.rst | 2 ++ Objects/stringlib/codecs.h | 4 ++-- Objects/unicodeobject.c | 8 +++++--- 4 files changed, 16 insertions(+), 5 deletions(-) create mode 100644 Misc/NEWS.d/next/Core and Builtins/2019-06-22-12-45-20.bpo-24214.hIiHeD.rst diff --git a/Lib/test/test_codecs.py b/Lib/test/test_codecs.py index d98f178384abf1..b375eb95b23b18 100644 --- a/Lib/test/test_codecs.py +++ b/Lib/test/test_codecs.py @@ -429,11 +429,18 @@ def test_lone_surrogates(self): def test_incremental_surrogatepass(self): # Test incremental decoder for surrogatepass handler: # see issue #24214 + # High surrogate data = '\uD901'.encode(self.encoding, 'surrogatepass') for i in range(1, len(data)): dec = codecs.getincrementaldecoder(self.encoding)('surrogatepass') self.assertEqual(dec.decode(data[:i]), '') self.assertEqual(dec.decode(data[i:], True), '\uD901') + # Low surrogate + data = '\uDC02'.encode(self.encoding, 'surrogatepass') + for i in range(1, len(data)): + dec = codecs.getincrementaldecoder(self.encoding)('surrogatepass') + self.assertEqual(dec.decode(data[:i]), '') + self.assertEqual(dec.decode(data[i:]), '\uDC02') class UTF32Test(ReadTest, unittest.TestCase): diff --git a/Misc/NEWS.d/next/Core and Builtins/2019-06-22-12-45-20.bpo-24214.hIiHeD.rst b/Misc/NEWS.d/next/Core and Builtins/2019-06-22-12-45-20.bpo-24214.hIiHeD.rst new file mode 100644 index 00000000000000..2d70ce05deaedf --- /dev/null +++ b/Misc/NEWS.d/next/Core and Builtins/2019-06-22-12-45-20.bpo-24214.hIiHeD.rst @@ -0,0 +1,2 @@ +Improved support of the surrogatepass error handler in the UTF-8 and UTF-16 +incremental decoders. diff --git a/Objects/stringlib/codecs.h b/Objects/stringlib/codecs.h index 8645bc26cff8ca..5fcc0aa30b8290 100644 --- a/Objects/stringlib/codecs.h +++ b/Objects/stringlib/codecs.h @@ -573,10 +573,10 @@ STRINGLIB(utf16_decode)(const unsigned char **inptr, const unsigned char *e, } /* UTF-16 code pair: */ - if (q >= e) - goto UnexpectedEnd; if (!Py_UNICODE_IS_HIGH_SURROGATE(ch)) goto IllegalEncoding; + if (q >= e) + goto UnexpectedEnd; ch2 = (q[ihi] << 8) | q[ilo]; q += 2; if (!Py_UNICODE_IS_LOW_SURROGATE(ch2)) diff --git a/Objects/unicodeobject.c b/Objects/unicodeobject.c index 4f8362590595cd..09990ca96c2b59 100644 --- a/Objects/unicodeobject.c +++ b/Objects/unicodeobject.c @@ -4937,11 +4937,13 @@ unicode_decode_utf8(const char *s, Py_ssize_t size, endinpos = startinpos + 1; break; case 2: - case 3: - case 4: - if (s == end || consumed) { + if (consumed && (unsigned char)*s == 0xED && end - s < 3) { + /* Truncated surrogate code in range D800-DFFF */ goto End; } + /* fall through */ + case 3: + case 4: errmsg = "invalid continuation byte"; startinpos = s - starts; endinpos = startinpos + ch - 1; From 46e236506ba9081be945a4b7a7974e653e4f1836 Mon Sep 17 00:00:00 2001 From: Serhiy Storchaka Date: Sat, 22 Jun 2019 20:18:42 +0300 Subject: [PATCH 2/2] Add tests. --- Lib/test/test_codecs.py | 17 +++++++++++++++++ Objects/stringlib/codecs.h | 2 +- Objects/unicodeobject.c | 4 +++- 3 files changed, 21 insertions(+), 2 deletions(-) diff --git a/Lib/test/test_codecs.py b/Lib/test/test_codecs.py index b375eb95b23b18..4317dfceb0399e 100644 --- a/Lib/test/test_codecs.py +++ b/Lib/test/test_codecs.py @@ -881,6 +881,23 @@ def test_surrogatepass_handler(self): with self.assertRaises(UnicodeDecodeError): b"abc\xed\xa0z".decode(self.encoding, "surrogatepass") + def test_incremental_errors(self): + # Test that the incremental decoder can fail with final=False. + # See issue #24214 + cases = [b'\x80', b'\xBF', b'\xC0', b'\xC1', b'\xF5', b'\xF6', b'\xFF'] + for prefix in (b'\xC2', b'\xDF', b'\xE0', b'\xE0\xA0', b'\xEF', + b'\xEF\xBF', b'\xF0', b'\xF0\x90', b'\xF0\x90\x80', + b'\xF4', b'\xF4\x8F', b'\xF4\x8F\xBF'): + for suffix in b'\x7F', b'\xC0': + cases.append(prefix + suffix) + cases.extend((b'\xE0\x80', b'\xE0\x9F', b'\xED\xA0\x80', + b'\xED\xBF\xBF', b'\xF0\x80', b'\xF0\x8F', b'\xF4\x90')) + + for data in cases: + with self.subTest(data=data): + dec = codecs.getincrementaldecoder(self.encoding)() + self.assertRaises(UnicodeDecodeError, dec.decode, data) + class UTF7Test(ReadTest, unittest.TestCase): encoding = "utf-7" diff --git a/Objects/stringlib/codecs.h b/Objects/stringlib/codecs.h index 5fcc0aa30b8290..d6f2b98f2b30a3 100644 --- a/Objects/stringlib/codecs.h +++ b/Objects/stringlib/codecs.h @@ -207,7 +207,7 @@ STRINGLIB(utf8_decode)(const char **inptr, const char *end, goto InvalidContinuation1; } else if (ch == 0xF4 && ch2 >= 0x90) { /* invalid sequence - \xF4\x90\x80\80- -- 110000- overflow */ + \xF4\x90\x80\x80- -- 110000- overflow */ goto InvalidContinuation1; } if (!IS_CONTINUATION_BYTE(ch3)) { diff --git a/Objects/unicodeobject.c b/Objects/unicodeobject.c index 09990ca96c2b59..61cf39b10a4e6d 100644 --- a/Objects/unicodeobject.c +++ b/Objects/unicodeobject.c @@ -4937,7 +4937,9 @@ unicode_decode_utf8(const char *s, Py_ssize_t size, endinpos = startinpos + 1; break; case 2: - if (consumed && (unsigned char)*s == 0xED && end - s < 3) { + if (consumed && (unsigned char)s[0] == 0xED && end - s == 2 + && (unsigned char)s[1] >= 0xA0 && (unsigned char)s[1] <= 0xBF) + { /* Truncated surrogate code in range D800-DFFF */ goto End; }