From d2171d2ba414def2ecf27b694ea27c2e9fde0fcf Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Martin=20v=2E=20L=C3=B6wis?= Date: Thu, 6 Nov 2003 20:47:57 +0000 Subject: Overallocate target buffer for normalization more early. Fixes #834676. Backported to 2.3. --- Lib/test/test_normalization.py | 3 +++ Modules/unicodedata.c | 12 +++++++----- 2 files changed, 10 insertions(+), 5 deletions(-) diff --git a/Lib/test/test_normalization.py b/Lib/test/test_normalization.py index 046dca6..0cbc2b4 100644 --- a/Lib/test/test_normalization.py +++ b/Lib/test/test_normalization.py @@ -84,5 +84,8 @@ def test_main(): continue assert X == NFC(X) == NFD(X) == NFKC(X) == NFKD(X), c + # Check for bug 834676 + normalize('NFC',u'\ud55c\uae00') + if __name__ == "__main__": test_main() diff --git a/Modules/unicodedata.c b/Modules/unicodedata.c index d266ad7..311db29 100644 --- a/Modules/unicodedata.c +++ b/Modules/unicodedata.c @@ -311,12 +311,14 @@ nfd_nfkd(PyObject *input, int k) stack[stackptr++] = *i++; while(stackptr) { Py_UNICODE code = stack[--stackptr]; - if (!space) { - space = PyString_GET_SIZE(result) + 10; - if (PyUnicode_Resize(&result, space) == -1) + /* Hangul Decomposition adds three characters in + a single step, so we need atleast that much room. */ + if (space < 3) { + int newsize = PyString_GET_SIZE(result) + 10; + space += 10; + if (PyUnicode_Resize(&result, newsize) == -1) return NULL; - o = PyUnicode_AS_UNICODE(result) + space - 10; - space = 10; + o = PyUnicode_AS_UNICODE(result) + newsize - space; } /* Hangul Decomposition. */ if (SBase <= code && code < (SBase+SCount)) { -- cgit v0.12