bug#50946: insert-file-contents can corrupt buffers. [Was: bug#50946: Emacs-28: Inadequate coding in hack-elisp-shorthands]

[João Távora]

João Távora <joaotavora@gmail.com> writes:

> I tried with a hook but it doesn't work: it breaks a shorthands test,
> a most basic one.  Maybe you can understand where the problem is?

As usual, digging a bit after posting the broken patch brought me to an
apparent solution.  The following patch seems to work, slightly
complexifies load-with-code-conversion, but simplifies shorthands.el
very much (and should solve the problems).  But please have a look.

João

commit 9bc76f971597d42ba0c8b4655501223d03181b3b
Author: João Távora <joaotavora@gmail.com>
Date:   Sun Oct 3 16:05:40 2021 +0100

    Simplify hack-read-symbol-shorthands again (bug#50946)
    
    * lisp/loadup.el (load-source-file-function): Don't set twice.
    
    * lisp/shorthands.el (hack-read-symbol-shorthands): Simplify.
    (load-with-shorthands-and-code-conversion): Remove.
    
    * lisp/international/mule.el (load-with-code-conversion): Call
    load-with-code-coversion-hook.  Set up shorthands.

diff --git a/lisp/international/mule.el b/lisp/international/mule.el
index 2a855b5673..b612ea5c77 100644
--- a/lisp/international/mule.el
+++ b/lisp/international/mule.el
@@ -294,6 +294,9 @@ define-charset
 
     (apply 'define-charset-internal name (mapcar 'cdr attrs))))
 
+(defvar load-with-code-conversion-hook nil
+  "Hook run in `load-with-code-conversion'.")
+
 (defun load-with-code-conversion (fullname file &optional noerror nomessage)
   "Execute a file of Lisp code named FILE whose absolute name is FULLNAME.
 The file contents are decoded before evaluation if necessary.
@@ -319,7 +322,8 @@ load-with-code-conversion
          (let ((load-true-file-name fullname)
                 (load-file-name fullname)
                 (set-auto-coding-for-load t)
-               (inhibit-file-name-operation nil))
+               (inhibit-file-name-operation nil)
+                shorthands)
            (with-current-buffer buffer
               ;; So that we don't get completely screwed if the
               ;; file is encoded in some complicated character set,
@@ -328,6 +332,8 @@ load-with-code-conversion
              ;; Don't let deactivate-mark remain set.
              (let (deactivate-mark)
                (insert-file-contents fullname))
+              (run-hooks 'load-with-code-conversion-hook)
+              (setq shorthands read-symbol-shorthands)
              ;; If the loaded file was inserted with no-conversion or
              ;; raw-text coding system, make the buffer unibyte.
              ;; Otherwise, eval-buffer might try to interpret random
@@ -338,11 +344,13 @@ load-with-code-conversion
                  (set-buffer-multibyte nil))
              ;; Make `kill-buffer' quiet.
              (set-buffer-modified-p nil))
-           ;; Have the original buffer current while we eval.
-           (eval-buffer buffer nil
-                        ;; This is compatible with what `load' does.
-                         (if dump-mode file fullname)
-                        nil t))
+           ;; Have the original buffer current while we eval,
+            ;; but consider shorthands of the eval'ed one.
+           (let ((read-symbol-shorthands shorthands))
+              (eval-buffer buffer nil
+                          ;; This is compatible with what `load' does.
+                           (if dump-mode file fullname)
+                          nil t)))
        (let (kill-buffer-hook kill-buffer-query-functions)
          (kill-buffer buffer)))
       (do-after-load-evaluation fullname)
diff --git a/lisp/loadup.el b/lisp/loadup.el
index 3fb6b81328..3a55d2c805 100644
--- a/lisp/loadup.el
+++ b/lisp/loadup.el
@@ -355,7 +355,6 @@
 (load "paren")
 
 (load "shorthands")
-(setq load-source-file-function #'load-with-shorthands-and-code-conversion)
 
 (load "emacs-lisp/eldoc")
 (load "cus-start") ;Late to reduce customize-rogue (needs loaddefs.el anyway)
diff --git a/lisp/shorthands.el b/lisp/shorthands.el
index c31ef3d216..ecf04ac587 100644
--- a/lisp/shorthands.el
+++ b/lisp/shorthands.el
@@ -28,35 +28,17 @@
 (require 'files)
 (eval-when-compile (require 'cl-lib))
 
-(defun hack-read-symbol-shorthands (fullname)
-  "Return value of `read-symbol-shorthands' file-local variable in FULLNAME.
-FULLNAME is the absolute file name of an Elisp .el file which
-potentially specifies a file-local value for
-`read-symbol-shorthands'.  The Elisp code in FULLNAME isn't read
-or evaluated in any way, except for extraction of the
-buffer-local value of `read-symbol-shorthands'."
-  (let* ((size (nth 7 (file-attributes fullname)))
-         (from (max 0 (- size 3000)))
-         (to size))
-    (with-temp-buffer
-      (while (and (< (buffer-size) 3000) (>= from 0))
-        (insert-file-contents fullname nil from to)
-        (setq to from
-              from (cond
-                    ((= from 0) -1)
-                    (t (max 0 (- from 100))))))
-      ;; FIXME: relies on the `hack-local-variables--find-variables'
-      ;; detail of files.el.  That function should be exported,
-      ;; possibly be refactored into two parts, since we're only
-      ;; interested in basic "Local Variables" parsing.
-      (alist-get 'read-symbol-shorthands 
(hack-local-variables--find-variables)))))
-
-(defun load-with-shorthands-and-code-conversion (fullname file noerror 
nomessage)
-  "Like `load-with-code-conversion', but also consider Elisp shorthands.
-This function uses shorthands defined in the file FULLNAME's local
-value of `read-symbol-shorthands', when it processes that file's Elisp code."
-  (let ((read-symbol-shorthands (hack-read-symbol-shorthands fullname)))
-    (load-with-code-conversion fullname file noerror nomessage)))
+(add-hook 'load-with-code-conversion-hook #'hack-read-symbol-shorthands)
+
+(defun hack-read-symbol-shorthands ()
+  "Set `read-symbol-shorthands' from Local Variables section."
+  ;; FIXME: relies on the `hack-local-variables--find-variables'
+  ;; detail of files.el.  That function should be exported,
+  ;; possibly be refactored into two parts, since we're only
+  ;; interested in basic "Local Variables" parsing.
+  (setq-local read-symbol-shorthands
+              (alist-get 'read-symbol-shorthands
+                         (hack-local-variables--find-variables))))
 

 ;; FIXME: move this all to progmodes/elisp-mode.el?  OTOH it'd make