On Mon, Dec 18, 2017 at 4:37 PM, Matthias Clasen <matthias.clasen@gmail.com> wrote:
the important change is the last hunk.

We store coords[i] in the local variable coord, then we apply the clamping to coord, but then ...
we use coords[i] as input for the normalization, so the clamping is lost.
That is what my patch changes.


Oh, and I just notice now that the patch I sent included my debug changes to replace FT_TRACE with printf.
That was of course not meant to be applied. Sorry for not noticing that earlier.