Cursor jumps wildly under CVS 11/11 and XEmacs 21.2

Cursor jumps wildly under CVS 11/11 and XEmacs 21.2

[Daniel Pittman]

Hi.

I am having a bit of a problem with pGnus from CVS, checked out at about
12:00 +1100 an 11/11/99. I am running "21.2  (beta20) \"Yoko\" XEmacs
Lucid"

The issue is that when you perform an action that advances the cursor
from the second-last item in the summary buffer to the last item, the
cursor will end up thrown six lines backward.

This is quite annoying but only happens in one case, as detailed:

I am reading the second last message in the summary.
The final article in unread.
I hit 'n' to go to the next article.
The last article displays.

At this point the cursor is sitting six lines back from the end of the
buffer.

If you need any more details, please let me know.

Thanks,
        Daniel

-- 
In a way, staring into a computer screen is like staring into an eclipse. It's
brilliant and you don't realize the damage until its too late.
        -- Bruce Sterling

Re: Cursor jumps wildly under CVS 11/11 and XEmacs 21.2

[Shenghuo ZHU]

>>>>> "Daniel" == Daniel Pittman <daniel@danann.net> writes:

Daniel> Hi.
Daniel> I am having a bit of a problem with pGnus from CVS, checked out at about
Daniel> 12:00 +1100 an 11/11/99. I am running "21.2  (beta20) \"Yoko\" XEmacs
Daniel> Lucid"

Daniel> The issue is that when you perform an action that advances the cursor
Daniel> from the second-last item in the summary buffer to the last item, the
Daniel> cursor will end up thrown six lines backward.

Daniel> This is quite annoying but only happens in one case, as detailed:

Daniel> I am reading the second last message in the summary.
Daniel> The final article in unread.
Daniel> I hit 'n' to go to the next article.
Daniel> The last article displays.

Daniel> At this point the cursor is sitting six lines back from the
Daniel> end of the buffer.

Daniel> If you need any more details, please let me know.


[...]

This is a re-center problem. 

1. Check `C-h f gnus-summary-recenter', is it an alias for
`gnus-xmas-summary-recenter'?

2. Are there some special fonts in the summary buffer?

-- 
Shenghuo

Re: Cursor jumps wildly under CVS 11/11 and XEmacs 21.2

[Daniel Pittman]

On 14 Nov 1999, Shenghuo ZHU <zsh@cs.rochester.edu> wrote:

>>>>>> "Daniel" == Daniel Pittman <daniel@danann.net> writes:

[...]

> Daniel> The issue is that when you perform an action that advances the
> Daniel> cursor from the second-last item in the summary buffer to the
> Daniel> last item, the cursor will end up thrown six lines backward.

[...]

> This is a re-center problem. 
> 
> 1. Check `C-h f gnus-summary-recenter', is it an alias for
> `gnus-xmas-summary-recenter'?

Yes.

> 2. Are there some special fonts in the summary buffer?

Not particularly. The current article highlights in bold but other than
that, no.

I can have a try at stepping through the function and see when it goes
wrong in a little. Just as soon as I can convince the boss that meetings
really don't achieve anything :/

        Daniel

-- 
Chaos often breeds life, when order breeds habit.
        -- Henry Adams

Re: Cursor jumps wildly under CVS 11/11 and XEmacs 21.2

[Daniel Pittman]

On 15 Nov 1999, Daniel Pittman <daniel@danann.net> wrote:

> On 14 Nov 1999, Shenghuo ZHU <zsh@cs.rochester.edu> wrote:
> 
>>>>>>> "Daniel" == Daniel Pittman <daniel@danann.net> writes:
> 
> [...]
> 
>> Daniel> The issue is that when you perform an action that advances
>> Daniel> the cursor from the second-last item in the summary buffer to
>> Daniel> the last item, the cursor will end up thrown six lines
>> Daniel> backward.
> 
> [...]
> 
>> This is a re-center problem. 
>> 
>> 1. Check `C-h f gnus-summary-recenter', is it an alias for
>> `gnus-xmas-summary-recenter'?
> 
> Yes.
> 
>> 2. Are there some special fonts in the summary buffer?

I finally have a reproducible case that actually works, as well as the
time to have investigated a little.

The issue is indeed a font-related one. In a summary buffer where the
last article is in a bold font (in my case, replies to one of my
messages), and the entry for it is the final line in the summary buffer,
the cursor jumps.

The problem is that the last line is in bold and thus fractionally
taller than the other lines. The recenter function decides that the
buffer should show completely and ends up calling 'set-window-start'
for position 1.

The cursor is placed on the last line in the buffer. This line is
fractionally clipped and thus not a valid place for point in the eyes of
XEmacs[1]. So, 'set-window-start' moves point to display the requested
position (1).

Bingo, cursor jumps and irritates me.

One fix I found for this issue was to use the optional third argument to
'set-window-start':

    Optional third arg NOFORCE non-nil inhibits next redisplay
    from overriding motion of point in order to display at this exact start.

This put the cursor on the right article but displayed it in the middle
of the screen. That would be sub-optimal in terms of pretty display but
very effective in terms of not doing the wrong thing(tm).


I may have time in a little to try and actually work out what is going
wrong but I am not an XEmacs redisplay guru and so it will likely be a
slow process.

The one real delta for buffer sizing in my case is that I have
'(set-specifier modeline-shadow-thickness 0)' to get flat modelines.

This does change the buffer size vs vanilla XEmacs settings somewhat.

        Daniel


Footnotes: 
[1]  If I get the buffer set up and scroll down to that line by hand
     (down arrow), when the cursor hits the final line, the summary
     buffer is scrolled.

     Using the scrollbar to scroll the buffer up without touching point
     will push the point off the bottom line when the top line is
     visible.

-- 
The question nowadays is not what makes government work. 
The question is how do we make it stop.
        -- P. J. O'Rourke

Re: Cursor jumps wildly under CVS 11/11 and XEmacs 21.2

[Shenghuo ZHU]

>>>>> "Daniel" == Daniel Pittman <daniel@danann.net> writes:


[...]

>>> 2. Are there some special fonts in the summary buffer?

Daniel> I finally have a reproducible case that actually works, as well as the
Daniel> time to have investigated a little.

Daniel> The issue is indeed a font-related one. In a summary buffer where the
Daniel> last article is in a bold font (in my case, replies to one of my
Daniel> messages), and the entry for it is the final line in the summary buffer,
Daniel> the cursor jumps.

Daniel> The problem is that the last line is in bold and thus fractionally
Daniel> taller than the other lines. The recenter function decides that the
Daniel> buffer should show completely and ends up calling 'set-window-start'
Daniel> for position 1.

Daniel> The cursor is placed on the last line in the buffer. This line is
Daniel> fractionally clipped and thus not a valid place for point in the eyes of
Daniel> XEmacs[1]. So, 'set-window-start' moves point to display the requested
Daniel> position (1).

Daniel> Bingo, cursor jumps and irritates me.

Daniel> One fix I found for this issue was to use the optional third argument to
Daniel> 'set-window-start':

Daniel>     Optional third arg NOFORCE non-nil inhibits next redisplay
Daniel> from overriding motion of point in order to display at this
Daniel> exact start.

Daniel> This put the cursor on the right article but displayed it in
Daniel> the middle of the screen. That would be sub-optimal in terms
Daniel> of pretty display but very effective in terms of not doing the
Daniel> wrong thing(tm).

That's exactly what I am looking for. Actually, I got this problem a
long time ago. It is not only font-related one. When a group with a
lot of long subjects, such as clari.world.top, this problem always
happens after truncating the lines (`C-t').

Daniel> I may have time in a little to try and actually work out what
Daniel> is going wrong but I am not an XEmacs redisplay guru and so it
Daniel> will likely be a slow process.

Daniel> The one real delta for buffer sizing in my case is that I have
Daniel> '(set-specifier modeline-shadow-thickness 0)' to get flat
Daniel> modelines.

This does not help the problem after truncating, since the size of
subject line changes a lot.

Daniel> This does change the buffer size vs vanilla XEmacs settings
Daniel> somewhat.

[...]

I've checked the fix in the CVS.

-- 
Shenghuo

Re: Cursor jumps wildly under CVS 11/11 and XEmacs 21.2

[Daniel Pittman]

On 16 Nov 1999, Shenghuo ZHU <zsh@cs.rochester.edu> wrote:

[...]

> I've checked the fix in the CVS.

I have been unable to reproduce the bug which is *so* nice. Thanks. :)

        Daniel

-- 
You'll laugh! You'll cry! You'll kiss three bucks goodbye!
        -- Narrator (Paul Frees), _Hardware Wars_