> I saw the warnings, but I'm unsure of their validity.
 > Here are some benchmarks that I did.  Each form was run on a fresh emacs -Q.
 > Apologies in advance if there is something wrong about them:
[...]
 > (benchmark-run 100000
 >    (with-temp-buffer
 >      (insert "Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore")
 >      (car (window-text-pixel-size
 >            (current-buffer) (point-min) (point) 1000000))))
 > ;; (8.006855 167 2.318854)
 >
 > It's a lot more efficient to use a buffer, but the difference
 > between the limits themselves continue to be insignificant.

These examples are harmless.  Please try to test (1) with a large buffer
that has no newline characters and (2) with 'truncate-lines' non-nil.
'window-text-pixel-size' must be able to handle these cases gracefully
even if it's not geared to them.

Any clients of 'window-text-pixel-size' like 'string-pixel-width' can
easily set X-LIMIT to some sufficiently large value without affecting
the basic functionality of 'window-text-pixel-size'.

 >> I hopefully fixed most of the issues here but cannot send you a patch at
 >> the moment to test because my local copy is completely out of synch with
 >> master.  So please bear with me.

Please try the attached patch (if it doesn't apply, I'll send you the
affected functions separately so you can apply the changes manually).

Thanks, martin