unofficial mirror of emacs-devel@gnu.org 
 help / color / mirror / code / Atom feed
* bug#60974: 30.0.50; byte-compile-preprocess mutates self evaluating forms in expanded macro bodies
       [not found] <b979003119097f23f6b13d5bc4f561596cdb9704.camel@gmail.com>
@ 2023-01-20 21:35 ` Vibhav Pant
  2023-01-21  5:43   ` Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors
  0 siblings, 1 reply; 10+ messages in thread
From: Vibhav Pant @ 2023-01-20 21:35 UTC (permalink / raw)
  To: 60974, monnier; +Cc: emacs-devel


[-- Attachment #1.1: Type: text/plain, Size: 160 bytes --]

The attached patch should fix this, thoughts?

Best,
Vibhav

-- 
Vibhav Pant
vibhavp@gmail.com
GPG: 7ED1 D48C 513C A024 BE3A  785F E3FB 28CB 6AB5 9598

[-- Attachment #1.2: Type: text/x-patch, Size: 744 bytes --]

diff --git a/lisp/emacs-lisp/bytecomp.el b/lisp/emacs-lisp/bytecomp.el
index aa9521e5a65..847965e6af6 100644
--- a/lisp/emacs-lisp/bytecomp.el
+++ b/lisp/emacs-lisp/bytecomp.el
@@ -2581,7 +2581,8 @@ byte-compile-flush-pending
 
 (defun byte-compile-preprocess (form &optional _for-effect)
   (let ((print-symbols-bare t))         ; Possibly redundant binding.
-    (setq form (macroexpand-all form byte-compile-macro-environment)))
+    (setq form (copy-tree
+                (macroexpand-all form byte-compile-macro-environment))))
   ;; FIXME: We should run byte-optimize-form here, but it currently does not
   ;; recurse through all the code, so we'd have to fix this first.
   ;; Maybe a good fix would be to merge byte-optimize-form into

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

^ permalink raw reply related	[flat|nested] 10+ messages in thread

* bug#60974: 30.0.50; byte-compile-preprocess mutates self evaluating forms in expanded macro bodies
  2023-01-20 21:35 ` bug#60974: 30.0.50; byte-compile-preprocess mutates self evaluating forms in expanded macro bodies Vibhav Pant
@ 2023-01-21  5:43   ` Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors
  2023-01-27 12:44     ` Vibhav Pant
  0 siblings, 1 reply; 10+ messages in thread
From: Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors @ 2023-01-21  5:43 UTC (permalink / raw)
  To: Vibhav Pant; +Cc: 60974, emacs-devel

> The attached patch should fix this, thoughts?

It's not really an option:
- it's expensive
- it breaks code when it doesn't form a tree, e.g.

      (list '#1=(a b #1#) 'c 'd)

Instead, we need to find out where in the code we perform the
side effect and change just that part.


        Stefan






^ permalink raw reply	[flat|nested] 10+ messages in thread

* bug#60974: 30.0.50; byte-compile-preprocess mutates self evaluating forms in expanded macro bodies
  2023-01-21  5:43   ` Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors
@ 2023-01-27 12:44     ` Vibhav Pant
  2023-01-28 23:10       ` Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors
  0 siblings, 1 reply; 10+ messages in thread
From: Vibhav Pant @ 2023-01-27 12:44 UTC (permalink / raw)
  To: Stefan Monnier; +Cc: 60974, emacs-devel


[-- Attachment #1.1: Type: text/plain, Size: 723 bytes --]

On Sat, 2023-01-21 at 00:43 -0500, Stefan Monnier wrote:
> > The attached patch should fix this, thoughts?
> 
> It's not really an option:
> - it's expensive
> - it breaks code when it doesn't form a tree, e.g.
> 
>       (list '#1=(a b #1#) 'c 'd)
> 
> Instead, we need to find out where in the code we perform the
> side effect and change just that part.
> 
> 
>         Stefan
> 

Ah, right. Theother way I could think of a fix is setq-ing `form` to a
shallow copy of the original form, with only the place(s) changed. This
patch tries to do that by using `pcase-let` to destructure forms.
 

-- 
Vibhav Pant
vibhavp@gmail.com
GPG: 7ED1 D48C 513C A024 BE3A  785F E3FB 28CB 6AB5 9598

[-- Attachment #1.2: Type: text/x-patch, Size: 2412 bytes --]

diff --git a/lisp/emacs-lisp/cconv.el b/lisp/emacs-lisp/cconv.el
index e715bd90a00..f6160a13579 100644
--- a/lisp/emacs-lisp/cconv.el
+++ b/lisp/emacs-lisp/cconv.el
@@ -477,20 +477,37 @@ cconv-convert
                                         branch))
                               cond-forms)))
 
-    (`(function (lambda ,args . ,body) . ,_)
+    (`(function (lambda ,args . ,body) . ,rest)
      (let* ((docstring (if (eq :documentation (car-safe (car body)))
                            (cconv-convert (cadr (pop body)) env extend)))
             (bf (if (stringp (car body)) (cdr body) body))
             (if (when (eq 'interactive (car-safe (car bf)))
                   (gethash form cconv--interactive-form-funs)))
             (cif (when if (cconv-convert if env extend)))
-            (_ (pcase cif
-                 (`#'(lambda () ,form) (setf (cadr (car bf)) form) (setq cif nil))
-                 ('nil nil)
-                 ;; The interactive form needs special treatment, so the form
-                 ;; inside the `interactive' won't be used any further.
-                 (_ (setf (cadr (car bf)) nil))))
-            (cf (cconv--convert-function args body env form docstring)))
+            (cf nil))
+       (pcase cif
+         (`#'(lambda () ,form)
+          (pcase-let ((`((,f1 . (,_ . ,f2)) . ,f3) bf))
+            (setq bf `((,f1 . (,form . ,f2)) . ,f3)))
+          (setq cif nil))
+         ('nil (setq bf nil))
+         ;; The interactive form needs special treatment, so the form
+         ;; inside the `interactive' won't be used any further.
+         (_ (pcase-let ((`((,f1 . (,_ . ,f2)) . ,f3) bf))
+              (setq bf `((,f1 . (nil . ,f2)) . ,f3)))))
+       (when bf
+         ;; If we modified bf, re-build body and form as
+         ;; copies with the modified bits.
+         (setq body (if (stringp (car body))
+                        (cons (car body) bf)
+                      bf)
+               form `(function (lambda ,args . ,body) . ,rest))
+         ;; Also, remove the current old entry on the alist, replacing
+         ;; it with the new one.
+         (let ((entry (pop cconv-freevars-alist)))
+           (push (cons body (cdr entry)) cconv-freevars-alist)))
+       (setq cf (cconv--convert-function args body env form docstring))
+
        (if (not cif)
            ;; Normal case, the interactive form needs no special treatment.
            cf

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

^ permalink raw reply related	[flat|nested] 10+ messages in thread

* bug#60974: 30.0.50; byte-compile-preprocess mutates self evaluating forms in expanded macro bodies
  2023-01-27 12:44     ` Vibhav Pant
@ 2023-01-28 23:10       ` Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors
  2023-01-31 13:52         ` Vibhav Pant
  0 siblings, 1 reply; 10+ messages in thread
From: Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors @ 2023-01-28 23:10 UTC (permalink / raw)
  To: Vibhav Pant; +Cc: 60974, emacs-devel

> Ah, right. Theother way I could think of a fix is setq-ing `form` to a
> shallow copy of the original form, with only the place(s) changed. This
> patch tries to do that by using `pcase-let` to destructure forms.

Hmm... yes, that's closer.  It's pretty ugly, tho.
Yet it's hard to make it less ugly here because the
`cconv-freevars-alist` really depends on `eq` equality.

Maybe we should pass `cif` to `cconv--convert-function` (or maybe even
move most of that `cconv--interactive-form-funs`-handling to that
function) and arrange for that function to do the
"non-side-effecting-equivalent-of" (setf (cadr (car bf)) form)?


        Stefan






^ permalink raw reply	[flat|nested] 10+ messages in thread

* bug#60974: 30.0.50; byte-compile-preprocess mutates self evaluating forms in expanded macro bodies
  2023-01-28 23:10       ` Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors
@ 2023-01-31 13:52         ` Vibhav Pant
  2023-01-31 14:34           ` Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors
  0 siblings, 1 reply; 10+ messages in thread
From: Vibhav Pant @ 2023-01-31 13:52 UTC (permalink / raw)
  To: Stefan Monnier; +Cc: 60974, emacs-devel

[-- Attachment #1: Type: text/plain, Size: 1086 bytes --]

On Sat, 2023-01-28 at 18:10 -0500, Stefan Monnier wrote:
> > Ah, right. Theother way I could think of a fix is setq-ing `form`
> > to a
> > shallow copy of the original form, with only the place(s) changed.
> > This
> > patch tries to do that by using `pcase-let` to destructure forms.
> 
> Hmm... yes, that's closer.  It's pretty ugly, tho.
> Yet it's hard to make it less ugly here because the
> `cconv-freevars-alist` really depends on `eq` equality.
> 
> Maybe we should pass `cif` to `cconv--convert-function` (or maybe
> even
> move most of that `cconv--interactive-form-funs`-handling to that
> function) and arrange for that function to do the
> "non-side-effecting-equivalent-of" (setf (cadr (car bf)) form)?
> 
> 
>         Stefan
> 

I tried to shift the `cif` related bits to `cconc--convert-function`,
but my understanding of cconv.el isn't that great, so I wasn't
successful at doing that :(. Should I add a TODO to the change for the
future?

-- 
Vibhav Pant
vibhavp@gmail.com
GPG: 7ED1 D48C 513C A024 BE3A  785F E3FB 28CB 6AB5 9598

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

^ permalink raw reply	[flat|nested] 10+ messages in thread

* bug#60974: 30.0.50; byte-compile-preprocess mutates self evaluating forms in expanded macro bodies
  2023-01-31 13:52         ` Vibhav Pant
@ 2023-01-31 14:34           ` Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors
  2023-02-01  7:41             ` Vibhav Pant
  0 siblings, 1 reply; 10+ messages in thread
From: Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors @ 2023-01-31 14:34 UTC (permalink / raw)
  To: Vibhav Pant; +Cc: 60974, emacs-devel

>> > Ah, right. Theother way I could think of a fix is setq-ing `form`
>> > to a
>> > shallow copy of the original form, with only the place(s) changed.
>> > This
>> > patch tries to do that by using `pcase-let` to destructure forms.
>> 
>> Hmm... yes, that's closer.  It's pretty ugly, tho.
>> Yet it's hard to make it less ugly here because the
>> `cconv-freevars-alist` really depends on `eq` equality.
>> 
>> Maybe we should pass `cif` to `cconv--convert-function` (or maybe
>> even
>> move most of that `cconv--interactive-form-funs`-handling to that
>> function) and arrange for that function to do the
>> "non-side-effecting-equivalent-of" (setf (cadr (car bf)) form)?
>> 
>> 
>>         Stefan
>> 
>
> I tried to shift the `cif` related bits to `cconc--convert-function`,
> but my understanding of cconv.el isn't that great, so I wasn't
> successful at doing that :( Should I add a TODO to the change for the
> future?

Yes, please,


        Stefan






^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: bug#60974: 30.0.50; byte-compile-preprocess mutates self evaluating forms in expanded macro bodies
  2023-01-31 14:34           ` Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors
@ 2023-02-01  7:41             ` Vibhav Pant
  2023-02-01 14:33               ` Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors
  0 siblings, 1 reply; 10+ messages in thread
From: Vibhav Pant @ 2023-02-01  7:41 UTC (permalink / raw)
  To: Stefan Monnier; +Cc: 60974, emacs-devel


[-- Attachment #1.1: Type: text/plain, Size: 1295 bytes --]

On Tue, 2023-01-31 at 09:34 -0500, Stefan Monnier wrote:
> > > > Ah, right. Theother way I could think of a fix is setq-ing
> > > > `form`
> > > > to a
> > > > shallow copy of the original form, with only the place(s)
> > > > changed.
> > > > This
> > > > patch tries to do that by using `pcase-let` to destructure
> > > > forms.
> > > 
> > > Hmm... yes, that's closer.  It's pretty ugly, tho.
> > > Yet it's hard to make it less ugly here because the
> > > `cconv-freevars-alist` really depends on `eq` equality.
> > > 
> > > Maybe we should pass `cif` to `cconv--convert-function` (or maybe
> > > even
> > > move most of that `cconv--interactive-form-funs`-handling to that
> > > function) and arrange for that function to do the
> > > "non-side-effecting-equivalent-of" (setf (cadr (car bf)) form)?
> > > 
> > > 
> > >         Stefan
> > > 
> > 
> > I tried to shift the `cif` related bits to `cconc--convert-
> > function`,
> > but my understanding of cconv.el isn't that great, so I wasn't
> > successful at doing that :( Should I add a TODO to the change for
> > the
> > future?
> 
> Yes, please,
> 
> 
>         Stefan
> 

Done, thoughts?

-- 
Vibhav Pant
vibhavp@gmail.com
GPG: 7ED1 D48C 513C A024 BE3A  785F E3FB 28CB 6AB5 9598

[-- Attachment #1.2: 60974-3.patch --]
[-- Type: text/x-patch, Size: 2591 bytes --]

diff --git a/lisp/emacs-lisp/cconv.el b/lisp/emacs-lisp/cconv.el
index e715bd90a00..5a6ae798c93 100644
--- a/lisp/emacs-lisp/cconv.el
+++ b/lisp/emacs-lisp/cconv.el
@@ -477,20 +477,40 @@ cconv-convert
                                         branch))
                               cond-forms)))
 
-    (`(function (lambda ,args . ,body) . ,_)
+    (`(function (lambda ,args . ,body) . ,rest)
      (let* ((docstring (if (eq :documentation (car-safe (car body)))
                            (cconv-convert (cadr (pop body)) env extend)))
             (bf (if (stringp (car body)) (cdr body) body))
             (if (when (eq 'interactive (car-safe (car bf)))
                   (gethash form cconv--interactive-form-funs)))
             (cif (when if (cconv-convert if env extend)))
-            (_ (pcase cif
-                 (`#'(lambda () ,form) (setf (cadr (car bf)) form) (setq cif nil))
-                 ('nil nil)
-                 ;; The interactive form needs special treatment, so the form
-                 ;; inside the `interactive' won't be used any further.
-                 (_ (setf (cadr (car bf)) nil))))
-            (cf (cconv--convert-function args body env form docstring)))
+            (cf nil))
+       ;; TODO: Because we need to non-destructively modify body, this code
+       ;; is particularly ugly.  This should ideally be moved to
+       ;; cconv--convert-function.
+       (pcase cif
+         (`#'(lambda () ,form)
+          (pcase-let ((`((,f1 . (,_ . ,f2)) . ,f3) bf))
+            (setq bf `((,f1 . (,form . ,f2)) . ,f3)))
+          (setq cif nil))
+         ('nil (setq bf nil))
+         ;; The interactive form needs special treatment, so the form
+         ;; inside the `interactive' won't be used any further.
+         (_ (pcase-let ((`((,f1 . (,_ . ,f2)) . ,f3) bf))
+              (setq bf `((,f1 . (nil . ,f2)) . ,f3)))))
+       (when bf
+         ;; If we modified bf, re-build body and form as
+         ;; copies with the modified bits.
+         (setq body (if (stringp (car body))
+                        (cons (car body) bf)
+                      bf)
+               form `(function (lambda ,args . ,body) . ,rest))
+         ;; Also, remove the current old entry on the alist, replacing
+         ;; it with the new one.
+         (let ((entry (pop cconv-freevars-alist)))
+           (push (cons body (cdr entry)) cconv-freevars-alist)))
+       (setq cf (cconv--convert-function args body env form docstring))
+
        (if (not cif)
            ;; Normal case, the interactive form needs no special treatment.
            cf

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

^ permalink raw reply related	[flat|nested] 10+ messages in thread

* bug#60974: 30.0.50; byte-compile-preprocess mutates self evaluating forms in expanded macro bodies
  2023-02-01  7:41             ` Vibhav Pant
@ 2023-02-01 14:33               ` Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors
  2023-02-02 12:35                 ` Vibhav Pant
  0 siblings, 1 reply; 10+ messages in thread
From: Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors @ 2023-02-01 14:33 UTC (permalink / raw)
  To: Vibhav Pant; +Cc: 60974, emacs-devel

>> Yes, please,
> Done, thoughts?

I think we can go with that for now.  I'll try to fix the todo at some
point, but I'm short on time for now,


        Stefan






^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: bug#60974: 30.0.50; byte-compile-preprocess mutates self evaluating forms in expanded macro bodies
  2023-02-01 14:33               ` Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors
@ 2023-02-02 12:35                 ` Vibhav Pant
  2023-02-02 14:26                   ` Eli Zaretskii
  0 siblings, 1 reply; 10+ messages in thread
From: Vibhav Pant @ 2023-02-02 12:35 UTC (permalink / raw)
  To: Stefan Monnier; +Cc: 60974, emacs-devel

[-- Attachment #1: Type: text/plain, Size: 508 bytes --]

On Wed, 2023-02-01 at 09:33 -0500, Stefan Monnier wrote:
> > > Yes, please,
> > Done, thoughts?
> 
> I think we can go with that for now.  I'll try to fix the todo at
> some
> point, but I'm short on time for now,
> 
> 
>         Stefan
> 

Sure. Do we need to install this to the release branch? The bug doesn't
really break anything AFAIK, so I dont think it's really a release
blocker.

-- 
Vibhav Pant
vibhavp@gmail.com
GPG: 7ED1 D48C 513C A024 BE3A  785F E3FB 28CB 6AB5 9598

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: bug#60974: 30.0.50; byte-compile-preprocess mutates self evaluating forms in expanded macro bodies
  2023-02-02 12:35                 ` Vibhav Pant
@ 2023-02-02 14:26                   ` Eli Zaretskii
  0 siblings, 0 replies; 10+ messages in thread
From: Eli Zaretskii @ 2023-02-02 14:26 UTC (permalink / raw)
  To: Vibhav Pant; +Cc: monnier, 60974, emacs-devel

> From: Vibhav Pant <vibhavp@gmail.com>
> Cc: 60974@debbugs.gnu.org, emacs-devel <emacs-devel@gnu.org>
> Date: Thu, 02 Feb 2023 18:05:13 +0530
> 
> Sure. Do we need to install this to the release branch? The bug doesn't
> really break anything AFAIK, so I dont think it's really a release
> blocker.

In that case, please don't install on the release branch.



^ permalink raw reply	[flat|nested] 10+ messages in thread

end of thread, other threads:[~2023-02-02 14:26 UTC | newest]

Thread overview: 10+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
     [not found] <b979003119097f23f6b13d5bc4f561596cdb9704.camel@gmail.com>
2023-01-20 21:35 ` bug#60974: 30.0.50; byte-compile-preprocess mutates self evaluating forms in expanded macro bodies Vibhav Pant
2023-01-21  5:43   ` Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors
2023-01-27 12:44     ` Vibhav Pant
2023-01-28 23:10       ` Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors
2023-01-31 13:52         ` Vibhav Pant
2023-01-31 14:34           ` Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors
2023-02-01  7:41             ` Vibhav Pant
2023-02-01 14:33               ` Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors
2023-02-02 12:35                 ` Vibhav Pant
2023-02-02 14:26                   ` Eli Zaretskii

Code repositories for project(s) associated with this public inbox

	https://git.savannah.gnu.org/cgit/emacs.git

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).