[gpshead]

TextIOWrapper.__next__() stops updating the decoder snapshot that tell() uses, but flush() re-enabled tell() unconditionally, so tell() returned a wrong position (and failed an assertion in debug builds when next() was the first read). flush() now re-enables tell() only when there is neither a snapshot nor pending decoded text, which keeps tell() working after a write() or a read() to EOF that follows next(). seek(), which relied on flush() for this, now re-enables it itself, in both the C and Python implementations. The position cannot be recovered without the snapshot, so tell() raises the same OSError it raises without the flush() call; letting next() keep tell() working is gh-81217.

Fixes #161055

• Issue: TextIOWrapper.tell() returns a wrong position after next() and flush() #161055