From e3d5f92a9f9bbbe9fd6af90bc6105260a9e70229 Mon Sep 17 00:00:00 2001 From: Sebastião Guerra Date: Sat, 3 Oct 2026 22:00:40 -0300 Subject: Reject lone surrogates converting TQString to Python MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Decode TQString UTF-8 with Python's strict decoder. Valid surrogate pairs become one supplementary Python code point, while lone surrogate encodings fail with UnicodeDecodeError. Python-to-TQString conversion is already in master. Signed-off-by: Sebastião Guerra --- sip/tqt/tqstring.sip | 43 ++++++------------------------------------- 1 file changed, 6 insertions(+), 37 deletions(-) diff --git a/sip/tqt/tqstring.sip b/sip/tqt/tqstring.sip index 108706c..162913a 100644 --- a/sip/tqt/tqstring.sip +++ b/sip/tqt/tqstring.sip @@ -750,44 +750,13 @@ extern PyObject *PyTQt_tqt_PyObject_FromTQString(const TQString *tqstr); // Convert a TQString to a Python Unicode object. PyObject *PyTQt_tqt_PyObject_FromTQString(const TQString *tqstr) { - PyObject *obj; + TQCString utf8 = tqstr->utf8(); - // We have to work out exactly which kind to use. We assume ASCII while we - // are checking so that we only go through the string once in the most - // common case. Note that we can't use PyUnicode_FromKindAndData() because - // it doesn't handle surrogates in UCS2 strings. - - int tqt_len = tqstr->length(); - Py_UCS4 maxchar = 0x007f; - - for (int tqt_i = 0; tqt_i < tqt_len && maxchar < 0xffff; ++tqt_i) - { - Py_UCS4 uch = tqstr->at(tqt_i).unicode(); - - if (uch > 0x00ff) - { - maxchar = 0xffff; - } - else if (uch > 0x007f) - { - maxchar = 0x00ff; - } - } - - // Create the correctly sized object. - if (!(obj = PyUnicode_New(tqt_len, maxchar))) - return nullptr; - - int kind = PyUnicode_KIND(obj); - void *data = PyUnicode_DATA(obj); - - for (int tqt_i = 0; tqt_i < tqt_len; ++tqt_i) - { - Py_UCS4 uch = tqstr->at(tqt_i).unicode(); - PyUnicode_WRITE(kind, data, tqt_i, uch); - } - - return obj; + // TQt encodes valid surrogate pairs as normal UTF-8, which Python decodes + // as one code point. Lone surrogates produce invalid UTF-8 and are rejected + // by the strict decoder. TQCString::length() stops at embedded NULs, so use + // size() - 1 to preserve the full payload while excluding the trailing NUL. + return PyUnicode_DecodeUTF8(utf8.data(), utf8.size() - 1, nullptr); } %End -- cgit v1.2.3