Bug report

Bug description:

StreamWriter.reset() of the stateful CJK codecs (hz, iso2022_jp*, iso2022_kr) usually writes nothing. MultibyteStreamWriter.reset() returns early when self->pending is NULL, but pending only holds an unfinished combining character, not the shift state, so the codec's encreset (which writes ESC ( B, SI or ~}) is skipped and the writer stays in double-byte mode.

import codecs, io
out = io.BytesIO()
w = codecs.getwriter('iso2022_jp')(out)
w.write('こんにちは。')
w.reset()
print(out.getvalue())  # b'\x1b$B$3$s$K$A$O!#'

Expected b'\x1b$B$3$s$K$A$O!#\x1b(B', as str.encode() gives. Anything written after reset() is read in JIS X 0208 mode, and the next CJK text written by the same writer has no opening escape. hz gives b'~{Dc:C' for '你好' instead of b'~{Dc:C~}'.

This is a regression from the gh-67436 crash fix (2015, 3.4.4/3.5), which added the early return to avoid passing a NULL pending to the encoder; before that, reset() always ran the codec reset. The same function also leaks the result of stream.write().

CPython versions tested on:

CPython main branch

Operating systems tested on:

Linux

Linked PRs

• gh-160185: Fix CJK StreamWriter.reset() not returning to ASCII #160186
• [3.15] gh-160185: Fix CJK StreamWriter.reset() not returning to ASCII (GH-160186) #161067
• [3.14] gh-160185: Fix CJK StreamWriter.reset() not returning to ASCII (GH-160186) #161068