bug#21722: 25.0.50; Unexpected point position after undo

bug#21722: 25.0.50; Unexpected point position after undo

[Markus Triska]

In "emacs -Q", when I insert into *scratch* the form:

   (progn (end-of-line) (insert "\nhi"))

then place point on the "g" and press C-M-x C-/, then the insertion is
undone, and point remains on the "g", as expected. When I then press:

   C-h f C-g C-M-x C-/

the insertion is undone, and point is unexpectedly placed at the end of
the form instead of remaining on the "g". When I then place point on the
"g" again, and again press C-M-x C-/, point again stays on the "g".

Thus, point position after undo is sometimes unexpected.

For comparison: In Emacs 22.2, point remains on the "g" in all these
cases. This is more consistent, and, in my view, therefore preferable.


In GNU Emacs 25.0.50.5 (x86_64-apple-darwin14.1.0, X toolkit, Xaw3d scroll bars)
 of 2015-10-04
Repository revision: acfb5cd0353406784f085ddb6edfb0d0587048c8
Windowing system distributor 'The X.Org Foundation', version 11.0.11502000
Configured using:
 'configure --without-ns CFLAGS=-I/opt/local/include
 LDFLAGS=-L/opt/local/lib'

Configured features:
XAW3D XPM JPEG TIFF GIF PNG IMAGEMAGICK GSETTINGS NOTIFY ACL GNUTLS
LIBXML2 FREETYPE XFT ZLIB TOOLKIT_SCROLL_BARS LUCID X11

Important settings:
  value of $LANG: en_US.UTF-8
  locale-coding-system: utf-8-unix

bug#21722: 25.0.50; Unexpected point position after undo

[Eli Zaretskii]

I think the reason for this problem is in keyboard.c.  C-g causes us
to throw to top-level right after reading the key, and the
undo-related variables are set up in the command loop after that.  I
guess something in this sequence interferes with recording of point
position.

bug#21722: 25.0.50; Unexpected point position after undo

[npostavs]

Markus Triska <triska@metalevel.at> writes:

> In "emacs -Q", when I insert into *scratch* the form:
>
>    (progn (end-of-line) (insert "\nhi"))
>
> then place point on the "g" and press C-M-x C-/, then the insertion is
> undone, and point remains on the "g", as expected. When I then press:
>
>    C-h f C-g C-M-x C-/
>
> the insertion is undone, and point is unexpectedly placed at the end of
> the form instead of remaining on the "g". When I then place point on the
> "g" again, and again press C-M-x C-/, point again stays on the "g".
>
> Thus, point position after undo is sometimes unexpected.

In 25.1-rc1 the point position stays in place for both cases.  So I
guess this can be considered fixed?

(In 25.0.93 (which I happened to have lying around), it moves in both
cases)

bug#1095: bug#21722: 25.0.50; Unexpected point position after undo

[Markus Triska]

npostavs@users.sourceforge.net writes:

> In 25.1-rc1 the point position stays in place for both cases.  So I
> guess this can be considered fixed?

Yes, this is fixed. Phillip Lord corrected this in #23871, thank you!

All the best,
Markus

bug#21722: 25.0.50; Unexpected point position after undo

[Phillip Lord]

npostavs@users.sourceforge.net writes:

> Markus Triska <triska@metalevel.at> writes:
>
>> In "emacs -Q", when I insert into *scratch* the form:
>>
>>    (progn (end-of-line) (insert "\nhi"))
>>
>> then place point on the "g" and press C-M-x C-/, then the insertion is
>> undone, and point remains on the "g", as expected. When I then press:
>>
>>    C-h f C-g C-M-x C-/
>>
>> the insertion is undone, and point is unexpectedly placed at the end of
>> the form instead of remaining on the "g". When I then place point on the
>> "g" again, and again press C-M-x C-/, point again stays on the "g".
>>
>> Thus, point position after undo is sometimes unexpected.
>
> In 25.1-rc1 the point position stays in place for both cases.  So I
> guess this can be considered fixed?
>
> (In 25.0.93 (which I happened to have lying around), it moves in both
> cases)


Sorry, think I just forgot to close this one. Thanks for checking.