Re: More enhancements to fringe bitmaps.

[YAMAMOTO Mitsuharu]

>>>>> On 16 Feb 2004 14:22:28 +0100, address@hidden (Kim F. Storm) said:

> Maybe the event/command loop does a select-window at some point
> which isn't restored before redisplay, and so it gets confused.

Hmm, if that is the case, I think the cursor at non-fringe area would
also be affected.

> Is this somehow related to where the mouse is on the screen; it
> could be some mouse related events which confuses the Carbon event
> loop.

I could not observe any difference by the mouse position.

> What happens if you set cursor-in-non-selected-windows to nil?

It just erases the hollow cursors in non-selected windows, and the
selected window still displays the hollow one.

Inspecting by a debugger, I found that glyph->type is 2 (IMAGE_GLYPH)
at the fringe position in question.  So the function
get_window_cursor_type determines the cursor type as
HOLLOW_BOX_CURSOR.

  /* Use normal cursor if not blinked off.  */
  if (!w->cursor_off_p)
    {
      if (glyph->type == IMAGE_GLYPH) {
        if (cursor_type == FILLED_BOX_CURSOR)
          cursor_type = HOLLOW_BOX_CURSOR;
      }
      return cursor_type;
    }

I guess glyph->type at the fringe position is not properly initialized
in Carbon Emacs.  But this does not explain why all the cursors
disappear when a window is split.

                                     YAMAMOTO Mitsuharu
                                address@hidden