gh-82052: Don't send partial UTF-8 sequences to the Windows API (GH-101103)
Don't send partial UTF-8 sequences to the Windows API
(cherry picked from commit f34176b77f222726d901595968a4b44456186da4)
Co-authored-by: Paul Moore <p.f.moore@gmail.com>
diff --git a/Misc/NEWS.d/next/Windows/2023-01-17-18-17-58.gh-issue-82052.mWyysT.rst b/Misc/NEWS.d/next/Windows/2023-01-17-18-17-58.gh-issue-82052.mWyysT.rst
new file mode 100644
index 0000000..4f7ab20
--- /dev/null
+++ b/Misc/NEWS.d/next/Windows/2023-01-17-18-17-58.gh-issue-82052.mWyysT.rst
@@ -0,0 +1 @@
+Fixed an issue where writing more than 32K of Unicode output to the console screen in one go can result in mojibake.
diff --git a/Modules/_io/winconsoleio.c b/Modules/_io/winconsoleio.c
index 5c1a6dd..c8f3481 100644
--- a/Modules/_io/winconsoleio.c
+++ b/Modules/_io/winconsoleio.c
@@ -954,7 +954,7 @@ _io__WindowsConsoleIO_write_impl(winconsoleio *self, Py_buffer *b)
{
BOOL res = TRUE;
wchar_t *wbuf;
- DWORD len, wlen, n = 0;
+ DWORD len, wlen, orig_len, n = 0;
HANDLE handle;
if (self->fd == -1)
@@ -984,6 +984,21 @@ _io__WindowsConsoleIO_write_impl(winconsoleio *self, Py_buffer *b)
have to reduce and recalculate. */
while (wlen > 32766 / sizeof(wchar_t)) {
len /= 2;
+ orig_len = len;
+ /* Reduce the length until we hit the final byte of a UTF-8 sequence
+ * (top bit is unset). Fix for github issue 82052.
+ */
+ while (len > 0 && (((char *)b->buf)[len-1] & 0x80) != 0)
+ --len;
+ /* If we hit a length of 0, something has gone wrong. This shouldn't
+ * be possible, as valid UTF-8 can have at most 3 non-final bytes
+ * before a final one, and our buffer is way longer than that.
+ * But to be on the safe side, if we hit this issue we just restore
+ * the original length and let the console API sort it out.
+ */
+ if (len == 0) {
+ len = orig_len;
+ }
wlen = MultiByteToWideChar(CP_UTF8, 0, b->buf, len, NULL, 0);
}
Py_END_ALLOW_THREADS
