Re: bug#75313: 31.0.50; ERC 5.6.1-git: M-w cannot copy multiline text when kill-ring-deindent-mode is on

Po Lu,

Hi. Just wondering if you might look at this when you get a chance. No
rush, of course, and TIA.

"J.P." <[email protected]> writes:

> DU Zaichuan <[email protected]> writes:
>
>> To reproduce the bug in emacs -Q,
>> 1. (kill-ring-deindent-mode t)
>> 2. open erc and join any channel.
>> 3. try to copy with M-w someone's message or the channel notice.
>
> Thanks. A slightly distilled version, sans ERC:
>
>   1. M-x kill-ring-deindent-mode RET
>   2. M-: (insert (propertize "(progn\n  1\n  2\n  3)" 'read-only t)) RET
>   3. C-SPC
>   4. C-P ; error: Text is read-only
>   5. M-w ; error: Text is read-only
>
>> When using the set-mark-command to highlight the texts, message gives
>> "kill-ring-deindent-buffer-substring-function: Text is read-only".
>
> Indeed. It seems all `filter-buffer-substring-function' advice members,
> like `kill-ring-deindent-buffer-substring-function', run after every
> command when the region is active because `region-extract-function'
> needs to supply `gui-set-selection' with the region's text. I see two
> ways of dealing with this:
>
> 1. Local advice around `filter-buffer-substring-function'.
>
>    (defun erc--filter-buffer-substring (orig &rest args)
>      (if (eq this-command #'kill-ring-save)
>          (with-silent-modifications (apply orig args))
>        (apply orig args)))
>   
>    So, somewhere in ERC's major-mode setup, we'd do:
>
>    (add-function :around (local 'filter-buffer-substring-function)
>                  #'erc--filter-buffer-substring)
>
>    However, this won't do anything for "Text is read only" spam related
>    to other commands, like `previous-line' or `move-end-of-line' (when
>    the region is active). But it should at least restore the ability to
>    copy with M-w.
>
> 2. Address this somewhere in lisp/indent-aux.el.
>
>    Binding `inhibit-read-only' to t around the call to `indent-rigidly'
>    in `kill-ring-deindent-buffer-substring-function' seems to help in
>    cursory experiments. Not sure if that's the right move, though.
>    Perhaps Po Lu (Cc'd) has some insights.