bug#52435: 29.0.50; C-y seems to contain an old contents

[Peter Dyballa]

Yes, I can reproduce it with -Q.

Launch: /opt/local/bin/emacs-29.0.50 -Q. Or equivalent.

Inside GNU Emacs 29.0.50: M-x shell RET.

In *shell* buffer just invoke ls -l RET for some contents. Or cat something 
with a few lines contents. Move the mouse cursor and point somewhere into this 
output, press C-SPACE to start marking a region. Move the mouse cursor some 
lines away (above or below the start of region). M-| or M-x 
shell-command-on-region RET. I chose to use "awk '{print $NF}'".

Now change to buffer *Shell Command Output* with C-x o. I think this is 
decisive. Just clicking into *Shell Command Output* buffer seems to do the 
right things. You are at the beginning of buffer and can immediately 
replace-string ^J against SPACE CHARACTER, go to beginning of line (C-a) and 
kill it with C-k. Now return to *shell* buffer with C-x o again. The region is 
still marked, gets high-lighted again. With Esc-> go to the command line prompt 
and C-y to insert the line you killed in *Shell Command Output* buffer – but 
something else (not from The Move) is inserted.


But: This is also the behaviour in GNU Emacsen 26.1, 27.1, 28.0.50 – I just 
used C-x o for the first time in this everyday procedure? transient-mark-mode 
is t, assuming it would clear the marked region when leaving it or the buffer.

The documentation in GNU Emacs 29.0.50 has become vague (on function 
transient-mark-mode):

        The mark is "deactivated" after certain non-motion
        commands, including those that change the text in the buffer, and
        during shift or mouse selection by any unshifted cursor motion
        command (see Info node ‘Shift Selection’ for more details).

The documentation in GNU Emacs 26.1 seems to be clearer:

        The mark is "deactivated" by changing the buffer,
        and after certain other operations that set the mark but whose
        main purpose is something else--for example, incremental search,
        M-<, and M->.

And it states that the region is cleared when changing to another buffer.

But OK, maybe I have to learn this different behaviour…


(After marking and applying shell-command-on-region I go to *compilation* 
buffer and then exchange it with *Shell Command Output* so that I am 
immediately at compilation again. Then I kill the now useless *Shell Command 
Output* buffer. During this period with new GNU Emacs 29.0.50 I exchanged 
*shell* with *Shell Command Output* and fell back into it afterwards, and the 
mark was still active and became high-lighted again – which puzzled me. When I 
then invoked compile again, the yank contents was a different one than killed. 
The first time this happened I quit Emacs, assuming it or I would have a 
problem with the new macOS version Monterey. The next time in new instance I 
just started my procedure again, and now for the third occurrence I tested 
Esc-y and more.)

--
Greetings

  Pete

With Capitalism man exploits man. With communism it's the exact opposite.