The issue with with-temp-buffer (or with-current-buffer)

The issue with with-temp-buffer (or with-current-buffer)

[OGAWA Hirofumi]

Hi,

The following is the reproduce code of issue,

(let ((buf (get-buffer-create "test-case")))
  (display-buffer buf)
  (with-current-buffer buf
    (erase-buffer)
    (insert "1\n")
    ;; start suspect
    (with-temp-buffer
      (save-window-excursion
	(set-window-buffer nil (current-buffer))))
    ;; end suspect
    (insert "2\n")))

and expected result is

    ---- test-case ----
    2
    1
    -------------------

But actual result is

    ---- test-case ----
    1
    2
    -------------------

[The actual issue that I'm hitted was using `shr-insert-document'
like following. And above is simplied version to reproduce the issue.

  (with-current-buffer buf
    (erase-buffer)
    (insert ...)
    (shr-insert-document '(html nil (body nil ...)))
    (insert ...)))
]

After some debugging, I noticed `with-current-buffer' that is
used in `with-temp-buffer' doesn't restore `point'.

And if I replaced `with-current-buffer' with
save-excursion+set-buffer, the issue is fixed.

I.e. from
	(with-temp-buffer
to
	(save-excursion
	  (set-buffer ,temp-buffer)

Well so, this behavior (doesn't restore point) of
`with-temp-buffer' (or `with-current-buffer') is intented or bug?

And to fix this issue, we should change which one of
with-temp-buffer, with-current-buffer, shr-insert-document, or
user of shr-insert-document?

Thanks.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: The issue with with-temp-buffer (or with-current-buffer)

[OGAWA Hirofumi]

OGAWA Hirofumi <hirofumi@mail.parknet.co.jp> writes:

> and expected result is
>
>     ---- test-case ----
>     2
>     1
>     -------------------
>
> But actual result is
>
>     ---- test-case ----
>     1
>     2
>     -------------------

Oops, expected and actual result was inverse.

and expected result is

    ---- test-case ----
    1
    2
    -------------------

But actual result is

    ---- test-case ----
    2
    1
    -------------------
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: The issue with with-temp-buffer (or with-current-buffer)

[Eli Zaretskii]

> From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
> Date: Sat, 16 Dec 2017 03:25:58 +0900
> 
> and expected result is
> 
>     ---- test-case ----
>     2
>     1
>     -------------------
> 
> But actual result is
> 
>     ---- test-case ----
>     1
>     2
>     -------------------
> 
> [The actual issue that I'm hitted was using `shr-insert-document'
> like following. And above is simplied version to reproduce the issue.
> 
>   (with-current-buffer buf
>     (erase-buffer)
>     (insert ...)
>     (shr-insert-document '(html nil (body nil ...)))
>     (insert ...)))
> ]
> 
> After some debugging, I noticed `with-current-buffer' that is
> used in `with-temp-buffer' doesn't restore `point'.

Please try the latest shr.el from the emacs-26 branch, I think you
will find that this problem is solved there.

> Well so, this behavior (doesn't restore point) of
> `with-temp-buffer' (or `with-current-buffer') is intented or bug?

It's intended.

> And to fix this issue, we should change which one of
> with-temp-buffer, with-current-buffer, shr-insert-document, or
> user of shr-insert-document?

I think shr.el was already fixed.

Re: The issue with with-temp-buffer (or with-current-buffer)

[OGAWA Hirofumi]

Eli Zaretskii <eliz@gnu.org> writes:

>> After some debugging, I noticed `with-current-buffer' that is
>> used in `with-temp-buffer' doesn't restore `point'.
>
> Please try the latest shr.el from the emacs-26 branch, I think you
> will find that this problem is solved there.
>
>> Well so, this behavior (doesn't restore point) of
>> `with-temp-buffer' (or `with-current-buffer') is intented or bug?
>
> It's intended.
>
>> And to fix this issue, we should change which one of
>> with-temp-buffer, with-current-buffer, shr-insert-document, or
>> user of shr-insert-document?
>
> I think shr.el was already fixed.

I've tested the following master and emacs-26 branches...

	master 5e6fb1e246b1e7f44ebba48e7e24f597a8eb92dd
	emacs-26 3003ac046900f9e7fdaa3161b99dbb1cc8f37b32

However result is still same (there is the described issue).

Thanks.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: The issue with with-temp-buffer (or with-current-buffer)

[OGAWA Hirofumi]

OGAWA Hirofumi <hirofumi@mail.parknet.co.jp> writes:

> Eli Zaretskii <eliz@gnu.org> writes:
>
>>> After some debugging, I noticed `with-current-buffer' that is
>>> used in `with-temp-buffer' doesn't restore `point'.
>>
>> Please try the latest shr.el from the emacs-26 branch, I think you
>> will find that this problem is solved there.
>>
>>> Well so, this behavior (doesn't restore point) of
>>> `with-temp-buffer' (or `with-current-buffer') is intented or bug?
>>
>> It's intended.
>>
>>> And to fix this issue, we should change which one of
>>> with-temp-buffer, with-current-buffer, shr-insert-document, or
>>> user of shr-insert-document?
>>
>> I think shr.el was already fixed.
>
> I've tested the following master and emacs-26 branches...
>
> 	master 5e6fb1e246b1e7f44ebba48e7e24f597a8eb92dd
> 	emacs-26 3003ac046900f9e7fdaa3161b99dbb1cc8f37b32
>
> However result is still same (there is the described issue).

I confirmed the following patch fixes the issue, although I'm not
sure if the code is proper or better fix is there.

Thanks

---

 lisp/net/shr.el |    7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff -puN lisp/net/shr.el~shr-fix-point-modification lisp/net/shr.el
--- emacs/lisp/net/shr.el~shr-fix-point-modification	2017-12-16 07:32:47.777230050 +0900
+++ emacs-hirofumi/lisp/net/shr.el	2017-12-16 07:35:48.083847035 +0900
@@ -591,9 +591,10 @@ size, and full-buffer size."
 (defun shr-string-pixel-width (string)
   (if (not shr-use-fonts)
       (length string)
-    (with-temp-buffer
-      (insert string)
-      (shr-pixel-column))))
+    (save-excursion
+      (with-temp-buffer
+        (insert string)
+        (shr-pixel-column)))))
 
 (defsubst shr--translate-insertion-chars ()
   ;; Remove soft hyphens.
_
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: The issue with with-temp-buffer (or with-current-buffer)

[Eli Zaretskii]

> From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
> Cc: emacs-devel@gnu.org
> Date: Sat, 16 Dec 2017 07:40:37 +0900
> 
> I confirmed the following patch fixes the issue, although I'm not
> sure if the code is proper or better fix is there.

Thanks.  Please report this, with a recipe to reproduce the problem
using shr.el, with "M-x report-emacs-bug RET", so that the details are
recorded by the bug tracker.

Re: The issue with with-temp-buffer (or with-current-buffer)

[OGAWA Hirofumi]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
>> Cc: emacs-devel@gnu.org
>> Date: Sat, 16 Dec 2017 07:40:37 +0900
>> 
>> I confirmed the following patch fixes the issue, although I'm not
>> sure if the code is proper or better fix is there.
>
> Thanks.  Please report this, with a recipe to reproduce the problem
> using shr.el, with "M-x report-emacs-bug RET", so that the details are
> recorded by the bug tracker.

OK. I've reported as bug#29734.

Thanks.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>