From 61799c325d0d76d8e807d2f74f03c6221529387c Mon Sep 17 00:00:00 2001 From: winez <130703526+mountain-beaver@users.noreply.github.com> Date: Sat, 14 Nov 2026 13:52:57 +0100 Subject: [PATCH] gh-136063: Fix quadratic complexity in email.header.decode_header() decode_header() popped items from the front of the ecre.split() result, deleted whitespace words from the middle of the word list one by one and built runs of encoded words with repeated bytes concatenation. All three are O(n) per step, making the function quadratic in the number of encoded words. Walk the split result by index, rebuild the word list in one pass and join each run once. --- Lib/email/header.py | 63 ++++++++++--------- Lib/test/test_email/test_email.py | 15 +++++ ...-11-14-13-45-00.gh-issue-136063.dHdrLn.rst | 2 + 3 files changed, 49 insertions(+), 31 deletions(-) create mode 100644 Misc/NEWS.d/next/Security/2026-11-14-13-45-00.gh-issue-136063.dHdrLn.rst diff --git a/Lib/email/header.py b/Lib/email/header.py index ea1f84ea964ca6..751c8feaba55b2 100644 --- a/Lib/email/header.py +++ b/Lib/email/header.py @@ -88,28 +88,29 @@ def decode_header(header): # two parts will be None. words = [] for line in header.splitlines(): + # ecre.split() returns [unencoded, charset, encoding, encoded, + # unencoded, ...]. Walk it by index: popping from the front would + # be quadratic in the number of encoded words. parts = ecre.split(line) - first = True - while parts: - unencoded = parts.pop(0) - if first: + for i in range(0, len(parts), 4): + unencoded = parts[i] + if i == 0: unencoded = unencoded.lstrip() - first = False if unencoded: words.append((unencoded, None, None)) - if parts: - charset = parts.pop(0).lower() - encoding = parts.pop(0).lower() - encoded = parts.pop(0) + if i + 1 < len(parts): + charset = parts[i + 1].lower() + encoding = parts[i + 2].lower() + encoded = parts[i + 3] words.append((encoded, encoding, charset)) - # Now loop over words and remove words that consist of whitespace - # between two encoded strings. - droplist = [] - for n, w in enumerate(words): - if n>1 and w[1] and words[n-2][1] and words[n-1][0].isspace(): - droplist.append(n-1) - for d in reversed(droplist): - del words[d] + # Now remove words that consist of whitespace between two encoded + # strings. Build a new list instead of deleting items one by one, + # which would be quadratic. + droplist = {n - 1 for n, w in enumerate(words) + if n > 1 and w[1] and words[n-2][1] + and words[n-1][0].isspace()} + if droplist: + words = [w for n, w in enumerate(words) if n not in droplist] # The next step is to decode each encoded word by applying the reverse # base64 or quopri transformation. decoded_words is now a list of the @@ -135,24 +136,24 @@ def decode_header(header): else: raise AssertionError('Unexpected encoding: ' + encoding) # Now convert all words to bytes and collapse consecutive runs of - # similarly encoded words. + # similarly encoded words. Each run is collected in a list and joined + # once, since repeated bytes concatenation would be quadratic. + def join_run(run, charset): + return (BSPACE if charset is None else b'').join(run) + collapsed = [] - last_word = last_charset = None + run = [] + last_charset = None for word, charset in decoded_words: if isinstance(word, str): word = bytes(word, 'raw-unicode-escape') - if last_word is None: - last_word = word - last_charset = charset - elif charset != last_charset: - collapsed.append((last_word, last_charset)) - last_word = word - last_charset = charset - elif last_charset is None: - last_word += BSPACE + word - else: - last_word += word - collapsed.append((last_word, last_charset)) + if run and charset != last_charset: + collapsed.append((join_run(run, last_charset), last_charset)) + run = [] + run.append(word) + last_charset = charset + collapsed.append((join_run(run, last_charset) if run else None, + last_charset)) return collapsed diff --git a/Lib/test/test_email/test_email.py b/Lib/test/test_email/test_email.py index 2a5794a5b66fc5..8edd301eca846a 100644 --- a/Lib/test/test_email/test_email.py +++ b/Lib/test/test_email/test_email.py @@ -2393,6 +2393,21 @@ def test_rfc2047_multiline(self): Re: =?mac-iceland?q?r=8Aksm=9Arg=8Cs?= baz foo bar =?mac-iceland?q?r=8Aksm?= =?mac-iceland?q?=9Arg=8Cs?=""") + def test_decode_header_linear_complexity(self): + # Ensure that decode_header() is fast for many encoded words. + # See https://github.com/python/cpython/issues/136063. + N = 100_000 + # Whitespace between adjacent encoded words is dropped and runs + # with the same charset are collapsed. + dh = decode_header(' '.join(['=?us-ascii?q?a?='] * N)) + self.assertEqual(dh, [(b'a' * N, 'us-ascii')]) + dh = decode_header(' '.join(['=?us-ascii?q?a?= =?utf-8?q?b?='] * N)) + self.assertEqual(len(dh), 2 * N) + self.assertEqual(dh[-2:], [(b'a', 'us-ascii'), (b'b', 'utf-8')]) + dh = decode_header(' x '.join(['=?us-ascii?q?a?='] * N)) + self.assertEqual(len(dh), 2 * N - 1) + self.assertEqual(dh[-2:], [(b' x ', None), (b'a', 'us-ascii')]) + def test_whitespace_keeper_unicode(self): eq = self.assertEqual s = '=?ISO-8859-1?Q?Andr=E9?= Pirard ' diff --git a/Misc/NEWS.d/next/Security/2026-11-14-13-45-00.gh-issue-136063.dHdrLn.rst b/Misc/NEWS.d/next/Security/2026-11-14-13-45-00.gh-issue-136063.dHdrLn.rst new file mode 100644 index 00000000000000..10a428051e50bf --- /dev/null +++ b/Misc/NEWS.d/next/Security/2026-11-14-13-45-00.gh-issue-136063.dHdrLn.rst @@ -0,0 +1,2 @@ +Fix quadratic complexity in :func:`email.header.decode_header` for header +values containing many :rfc:`2047` encoded words.