🚀 go-pugleaf

Recently I submitted an enquiry to gnu.emacs.help, but I didn't get
back to reading the news soon enough to find any responses.  The
following code induces a strange side effect that I may be mistaking
for a subtle bug in Rmail.

Switching from using 19.19 to 19.25 has introduced a subtle bug into
code that worked on the older Emacs.  Executing the following code
suffices to reproduce the problem on an Emacs 19.25 invoked with the
-q option:

(setq rmail-delete-after-output t)

(setq rmail-show-message-hook 'rmail-show-message-hook-fun)

(defun rmail-show-message-hook-fun ()
  (rmail-clear-headers)
  (dcbe-rmail-redisplay-summary))

(defun dcbe-rmail-redisplay-summary ()
  "Readjust the summary window."
  (cond ((rmail-summary-displayed)
	 (save-excursion
	   (pop-to-buffer rmail-summary-buffer)
	   (recenter)))))

Now an attempt to use the "o" command from an Rmail summary buffer
yields the following *Backtrace* if debug-on-error is set

Signalling: (void-variable rmail-buffer)
  rmail-summary-delete-forward(nil)
  rmail-summary-output-to-rmail-file()
  call-interactively(rmail-summary-output-to-rmail-file)

I have examined the source code for the Rmail functions involved, but
I couldn't make out what's going wrong.  The code listed above works
well on 19.19.

I would be grateful for any response, but please respond by email.

Thanks,
Terry Stroup