bug#6920: 23.2; ESC passed out of order to post-read-conversion

[Stefan Kangas]

Ryan Johnson <ryanjohn@ece.cmu.edu> writes:

>  This problem shows up running in terminal mode on a cygwin xterm,
> apparently because it sends terminal escape sequences one character at a
> time (where linux usually sends the whole thing at once). To reproduce,
> run the following three elisp commands:
>
> (define-coding-system 'utf-8-echo
>   "Echoes all input to the message area"
>   :coding-type 'utf-8
>   :mnemonic ?U
>   :ascii-compatible-p t
>   :charset-list '(unicode)
>   :post-read-conversion 'echo-conversion)
> (defun echo-conversion (len)
>   (let* ((p (point))
>      (e (+ p len))
>      (str (buffer-substring p e)))
>     (message "%s" str)
>     len))
> (set-keyboard-coding-system 'utf-8-echo)
>
> Then start using the left arrow key to navigate. The escape sequence is
> '^[OD' and every few arrow presses garbled bits of the "OD" portion will
> appear in the buffer and the point doesn't move properly. A
> representative sample of the message buffer looks like this:
>
> ^[
> O
> D
> ^[
> O
> D
> O
> D
> ^[ [2 times]
> O
> D
>
> The lossage buffer (below) shows how emacs becomes very confused as a
> result, reporting 'ESC O D O D D O D D ESC' at one point. Defining a
> similar coding system based on iso-8859-1 instead of utf-8 causes emacs
> to crash. Unfortunately, cygwin-gdb doesn't trap the signal so I can't
> produce a stack trace, but sometimes it gets stuck in an endless loop
> inside malloc, which suggests some sort of memory corruption is the
> culprit. As far as I know, 'ESC O D' is valid in both coding systems
> so that shouldn't be the problem.
>
> System info follows...
>
> In GNU Emacs 23.2.1 (i686-pc-cygwin)
>  of 2010-08-12 on host
> configured using `configure  '--prefix=/home/Ryan/apps/emacs-23.2'
> '--without-xpm' '--without-png' '--without-gif''

This was reported 9 years ago but unfortunately never got a reply at the
time.  Can you still reproduce it on a modern version of Emacs, for
example the latest version 26.3?

Best regards,
Stefan Kangas