Re: Window display bug

[storm@cua.dk (Kim F. Storm)]

Nick Roberts <nick@nick.uklinux.net> writes:

> I can give you a way to reproduce it. I get it with emacs -q --no-site-file
> with a standard emacs (compiled with just `make'). If you can't see it I'll
> send more details of my set up. You'll need the version of gdb-ui.el that
> I've just checked in to CVS (2003-06-18).
> 
> Type M-x gdba, then either give gdb the name of an executable that has been
> compiled without -g or select GUD->View->Assembler from the menubar so that
> Assembler is displayed in the source window. Type "break main" in the GUD
> buffer and then "run". A breakpoint icon should appear and an arrow behind it.
> Now step the program at the instruction level (either by clicking on the
> appropriate toolbar button or by typing "ni" and the <RET> repeatedly). The
> arrow should move and point with it until they disappear off the window. If
> you have line-number mode enabled, then you will see that number increment
> after each instruction is executed. Window display should follow the execution
> (or rather the point) but it only does this if Richard's change is reverted.
> 

My followup:
  > I do see some strange effects, yes.
  > Could be related to doing (recenter -1) when point is
  > on the bottom line of the window.
  > I'll investigate further.
 

Your reply:
> But the latest version of gdb-ui.el (1.36) doesn't use (recenter -1). I
> originally put it in to try to force the redisplay before concluding that it
> was a bug with emacs. I'm pretty sure it relates to the change that I
> specified.

If I look at a backtrace when emacs enters the relevant redisplay code,
it is still called from (recenter -1) somewhere in comint.

Actually, if the cursor is placed on the bottom line (at the
time where doing "ni" doesn't work as intended), if I do
M-: (recenter -1) RET, the same incorrect behaviour is observed.

To me it seems that there is a problem in Frecenter with ARG == -1,
although I cannot see exactly what the problem ... however, in the
error case, the it->current_y never seems to have any other value than
0, so it doesn't really know where the cursor is, or how much it needs
to scroll...

I'm short on time currently, so I don't have much time to debug this.

Have anyone else seen problems with (recenter -1) ??

-- 
Kim F. Storm <storm@cua.dk> http://www.cua.dk