pixel-point-and-height-at-unseen-line

[Eli Zaretskii <eliz@gnu.org>]

This function has several problems that IMO would be good to clean up:

 . It's unnecessarily restricted to a very special use case: a single
   screen line above the one where window-start is.  It could be much
   more useful (and be moved to subr.el or simple.el) if it could at
   least accept an argument to control on which line or position to
   report.

 . It changes important state variables without protecting itself from
   errors and non-local exits, so a C-g at an opportune time will
   disrupt the buffer/window state.

 . It calls vertical-motion without checking its return value, which
   could tell you whether it actually moved at all, and where.  (One
   bonus of checking the return value is that you could perhaps drop
   the bobp test there.)  vertical-motion can do some surprising
   things around overlays and display properties, especially ones that
   include newlines in their strings.

Thanks.