bug#56210: 29.0.50; Keyboard macros do not trigger after-change-functions

[Eli Zaretskii]

> Date: Sat, 25 Jun 2022 16:34:11 -0400
> Cc: larsi@gnus.org, 56210@debbugs.gnu.org
> From: Richard Hansen <rhansen@rhansen.org>
> 
> On 2022-06-25 14:56, Eli Zaretskii wrote:
> > Maybe we should step back and ask what do you think the below do?
> > 
> >    (execute-kbd-macro (kbd "x"))
> > 
> > What is, in your mental model, the effect of the above in Emacs?
> 
> I would expect that to be roughly equivalent to typing "x" on the physical 
> keyboard, which would insert an event into some sort of event queue.  (I'm 
> not familiar enough with Emacs to know whether that queue is per-buffer, 
> per-window, per-frame, per-terminal, per-process, or something else.)  Under 
> normal circumstances, the event would be dispatched to the buffer associated 
> with the `selected-window' of the `selected-frame', and insert an "x".

That is correct.

> I think my understanding of "current buffer" is incorrect.  My current mental 
> model of `with-current-buffer' is that it behaves as if the window's 
> associated buffer is briefly switched to the specified buffer (like "C-x b" 
> does) except it doesn't actually go to the expense of rendering the other 
> buffer in the window (or update the toolbar, etc.).  Looking at the 
> documentation of `set-window-buffer', that mental model appears to be wrong.

Yes, with-temp-buffer doesn't associate the temporary buffer with the
selected window.  And execute-kbd-macro starts a command loop, which
always makes sure the selected-window's buffer is current.

> But if my mental model is wrong, why does my original code insert "x" into 
> the temporary buffer? It is not due to something specific to 
> `ert-with-test-buffer', because if I evaluate the following in *scratch* then 
> I see the same behavior that Lars saw with `with-temp-buffer' (the "x" is 
> inserted into *scratch*):
> 
>      (ert-with-test-buffer () (execute-kbd-macro (kbd "x")))
> 
> Maybe `ert-run-tests-batch-and-exit' is setting it up to more closely 
> simulate a user interacting with Emacs?  Or maybe the behavior is caused by 
> batch mode?

Could be, you will have to investigate what ERT does in those cases.
Its macro complexity is above my pay grade.

> Regardless, given that the temporary buffer is modified in my original 
> example, shouldn't `after-change-functions' run?

Is the temporary buffer indeed modified, and is the hook set in that
buffer?