David De La Harpe Golden wrote:

> So I think the gtk+/emacs bug may show up (I don't actually see it on my 
> system, but my gtk+ might be too old)


Okay, I haven't replicated the full scrollbar-trashing thing the gtk+ 
guys mention, (possibly emacs nowadays avoids drawing over the scrollbar 
area in question so it's not a problem?), but I think I now see the 
issue that Yamamoto Mitsuharu was probably actually talking about -

when visible-bell flashes the first line of the window, part of the 
scrollbar is also inverse-video-d unless I use
GDK_NATIVE_WINDOWS=1 emacs

Contrast top right and bottom right in the two attached screenshots -
note how part of the scrollbar is inverted in the _nonative one
(the default, without gtk instructed to use "native" windows).*

It's quite a minor and transient effect (on my system anyway).  Still, I 
guess the flash inversion area in xterm.c/XTflash() is winding up a bit 
off somehow - the intent of the code does seem to be that scrollbars 
should be excluded from the area, but maybe it's only working for some 
cases.

(* the color difference in the visible bell effect in the main and 
minibuffer window in each screenshot is not significant, just due to my 
use of global-hl-line-mode)