bug#53294: 29.0.50; Indirect font changes incorrectly affecting original buffer

[Andrew Hyatt <ahyatt@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 2090 bytes --]

On Mon, Jan 17, 2022 at 04:47 PM Stefan Monnier 
<monnier@iro.umontreal.ca> wrote: 

>>> - In `face-remap.el`, refrain from modifying the 
>>> `face-remapping-alist` 
>>>   by side-effects (i.e. avoid `delq`, `setcdr`, and friends). 
>>>  - Add a `make-indirect-buffer-hook` and arrange for 
>>> `face-remap.el` to 
>>>   add a function there that does a deep enough copy of 
>>>   `face-remapping-alist`. 
>>>  - Remember that indirect buffers are an attractive nuisance 
>>> and should 
>>>   be deprecated (but note that I suspect the same bug affects 
>>>   `clone-buffer` because it doesn't make a deep enough copy of 
>>>   `face-remapping-alist` either). 
>> 
>> The last one tells me we are better with leaving this sleeping 
>> dog lie. 
> 
> Actually, looking back at the code, I see we already have the 
> needed hook (tho called from `clone-indirect-buffer`), so I 
> think the better option is to change `make-indirect-buffer` so 
> it runs `clone-indirect-buffer-hook` (when called with a non-nil 
> `clone` arg) instead of having `clone-indirect-buffer` run it. 
> And then in `face-remap.el` add a function to that hook which 
> copies `face-remapping-alist`. 
> 
> An intermediate option is to change the Org code to use 
> `clone-indirect-buffer` instead of `make-indirect-buffer` (but 
> that still requires the `face-remap.el` addition to the hook). 
> 
> [ Sorry Andrew for not noticing your earlier patch which does 
> something 
>   similar to my first suggestion above.  ] 
> 
> Andrew, would you be willing to write that patch?
>  
>         Stefan 

That sounds reasonable, I can do that. In fact, I just did this 
based on your suggestions, and it does work well. The only weird 
thing is that I had to pull `clone-indirect-buffer-hook' into the 
c code, because that's where `make-indirect-buffer' is. Let me 
know if that seems wrong.

I've attached the patch, please let me know if there is an issue. 
I have commit access, so I can just commit it myself after your OK 
(if so, I'll wait for a week or so to see if Eli has a comment as 
well before checkin).


[-- Attachment #2: text-scale-patch-v2 --]
[-- Type: application/octet-stream, Size: 2895 bytes --]

diff --git a/lisp/face-remap.el b/lisp/face-remap.el
index 00560f9d2e..fc0edb7119 100644
--- a/lisp/face-remap.el
+++ b/lisp/face-remap.el
@@ -70,6 +70,14 @@ internal-lisp-face-attributes
    :foreground :background :stipple :overline :strike-through :box
    :font :inherit :fontset :distant-foreground :extend :vector])
 
+(defun face-attrs--make-indirect-safe ()
+  "Do a deep copy of `face-remapping-alist' to keep the original buffer safe."
+  (setq-local face-remapping-alist
+              (copy-tree
+               (buffer-local-value 'face-remapping-alist (buffer-base-buffer)))))
+
+(add-hook 'clone-indirect-buffer-hook #'face-attrs--make-indirect-safe)
+
 (defun face-attrs-more-relative-p (attrs1 attrs2)
   "Return true if ATTRS1 contains a greater number of relative
 face-attributes than ATTRS2.  A face attribute is considered
diff --git a/lisp/simple.el b/lisp/simple.el
index cbcde9fb8d..3ef9cca249 100644
--- a/lisp/simple.el
+++ b/lisp/simple.el
@@ -9441,9 +9441,6 @@ function-key-map
 (defvar clone-buffer-hook nil
   "Normal hook to run in the new buffer at the end of `clone-buffer'.")
 
-(defvar clone-indirect-buffer-hook nil
-  "Normal hook to run in the new buffer at the end of `clone-indirect-buffer'.")
-
 (defun clone-process (process &optional newname)
   "Create a twin copy of PROCESS.
 If NEWNAME is nil, it defaults to PROCESS' name;
diff --git a/src/buffer.c b/src/buffer.c
index 10ac91915c..00646f8779 100644
--- a/src/buffer.c
+++ b/src/buffer.c
@@ -912,6 +912,10 @@ DEFUN ("make-indirect-buffer", Fmake_indirect_buffer, Smake_indirect_buffer,
       Fset (intern ("buffer-save-without-query"), Qnil);
       Fset (intern ("buffer-file-number"), Qnil);
       Fset (intern ("buffer-stale-function"), Qnil);
+      /* Cloned buffers need extra setup, to do things such as deep
+	 variable copies for list variables that might be mangled due
+	 to destructive operations in the indirect buffer. */
+      safe_run_hooks (Qclone_indirect_buffer_hook);
       set_buffer_internal_1 (old_b);
     }
 
@@ -5568,6 +5572,9 @@ syms_of_buffer (void)
 	pure_list (Qprotected_field, Qerror));
   Fput (Qprotected_field, Qerror_message,
 	build_pure_c_string ("Attempt to modify a protected field"));
+  
+  DEFSYM (Qclone_indirect_buffer_hook, "clone-indirect-buffer-hook");
+ 
 
   DEFVAR_PER_BUFFER ("tab-line-format",
 		     &BVAR (current_buffer, tab_line_format),
@@ -6392,6 +6399,10 @@ Functions (implicitly) running this hook are `get-buffer-create',
 This is the default.  If nil, auto-save file deletion is inhibited.  */);
   delete_auto_save_files = 1;
 
+  DEFVAR_LISP ("clone-indirect-buffer-hook", Vclone_indirect_buffer_hook,
+	       doc: /* Normal hook to run in the new buffer at the end of `clone-indirect-buffer'. */);
+  Vclone_indirect_buffer_hook = Qnil;
+
   defsubr (&Sbuffer_live_p);
   defsubr (&Sbuffer_list);
   defsubr (&Sget_buffer);