Mon Feb 1 21:23:23 PST 2010
looks like you replaced some instances of hypot with sqrt as the
additional precision was not needed. Is this important for
performance? Could we get away using normalize everywhere (which would
mean hypot)? Is sqrt precision sufficient everywhere or only where you
replaced hypot with it?
A first patch (unrelated to code duplication) is available here:
It fixes the path flattening (I know it could be made faster by
avoiding the closure w/ offset, but that would require writing
additional functions for move_to, line_to and close_path. I wanted to
keep the change small to make it easier to understand the problem).
More information about the cairo