bug#11210: Windows emacs 23.4.1: scroll-conservatively > 0 results in multiple cursors being displayed after scrolling

[Bill Meier <wmeier@newsguy.com>]

On 4/15/2012 1:27 PM, Bill Meier wrote:
> On 4/15/2012 10:43 AM, Jason Rumney wrote:
>> Bill Meier<wmeier@newsguy.com> writes:
>>
>>> If others using Windows 7 do see an empty rectangle in non-selected
>>> windows and/or are able to change the cursor config from within Emacs,
>>> then obviously there's something special/different about my
>>> configuration.
>>
>> Are you using screen reader or other accessibility software?
>>
>>
>
> No .....
>
>

Um...

After reading the code in w32term.c and checking the value of
w32-use-visible-system-caret in my Emacs, I found it had a value of 1.

So: Jason asked the right question.

It turned out that (unremembered by me) I once tried out
speech-recognition/text-to-speech which was still enabled (but not 
actually being used).

When I disabled same, my Emacs cursor was "normal" (and no artifacts 
appeared when I downarrowed off the bottom of the screen).

However, if I set w32-use-visible-system-caret to 1 and
scroll-conservatively to 1, I get artifacts.

So: this bug should actually be entitled:

(scroll-conservertively > 0) && (w32-use-visible-system-caret == 1) 
results in multiple cursors ....

Re:
 > Can you run Emacs you built under a debugger?  If so, please make an
 > unoptimized build ("configure --no-opt" in the nt/ directory to
 > configure the package before compiling), and please show the values of
 > yb and last_new on line 5021 of dispnew.c, when you press down-arrow
 > on the "123" line in this recipe:

 > For the record, the values I see are yb = 384 and last_new = 24.
 >

I see the same values.

Note: To reliably (90% of the time) get artifacts I actually used 32 
lines as follows:
001
002
003
...
032
123

Using abc,abc,...,123 now doesn't give artifacts for some reason.
Actually: the breakpoint is never hit in this case.
So: I'm no longer sure about my originally stated test case (abc,...,123).

Note: Just for the record, I put the breakpoint at
r 100582: dispnew.c: line 5016

5016      i = first_old + 1;
5017      while (i < current_matrix->nrows - 1)