code review 209280043: freetype: fix drawContour when the first point is off t...

freetype: fix drawContour when the first point is off the curve.

Fixes issue 16.

[nigeltao, 2015-03-06 07:49:18]

Hello crawshaw@golang.org (cc: golang-codereviews@googlegroups.com,
philipz@seznam.cz),

I'd like you to review this change to
https://code.google.com/p/freetype-go

[crawshaw, 2015-03-06 12:43:19]

LGTM

I suppose this is difficult to test...

https://codereview.appspot.com/209280043/diff/40001/freetype/freetype.go
File freetype/freetype.go (right):

https://codereview.appspot.com/209280043/diff/40001/freetype/freetype.go#newc...
freetype/freetype.go:117: start = raster.Point{
This is a guess, no? It is not obvious to me that the mid-point between start
and last is on the contour. Maybe a comment?

[nigeltao, 2015-03-07 00:15:32]

On 2015/03/06 12:43:19, crawshaw wrote:
> I suppose this is difficult to test...

Yeah, I tested it manually, using the repro given in the excellent bug report:
https://code.google.com/p/freetype-go/issues/detail?id=16


> https://codereview.appspot.com/209280043/diff/40001/freetype/freetype.go
> File freetype/freetype.go (right):
> 
>
https://codereview.appspot.com/209280043/diff/40001/freetype/freetype.go#newc...
> freetype/freetype.go:117: start = raster.Point{
> This is a guess, no? It is not obvious to me that the mid-point between start
> and last is on the contour. Maybe a comment?

It's not a guess. Truetype have linear and quadratic Beziers, but no cubics.
When there are two consecutive 'off' points, there is an implicit 'on' point
(the midpoint) for the two consecutive quadratic splines to meet at. The
existing code in the rest of the function did this implicit midpoint thing. It's
just that it didn't do that if the first and last points were 'off'. This patch
fixes that.

I thought this was written down somewhere, but I didn't see it on a quick skim
of the Apple TTF docs. I'll have a longer look later.

[philipz_seznam.cz, 2015-03-07 08:34:40]

"On 2015/03/06 12:43:19, crawshaw wrote:
> I suppose this is difficult to test...
Yeah, I tested it manually, using the repro given in the excellent bug
report: https://code.google.com/p/freetype-go/issues/detail?id=16

"



 Virtually every TT font incorporating circle should demonstrate this 
behauvior.

This is how FontForge converts Cubic to Quadratic curves.






"
""I thought this was written down somewhere, but I didn't see it on a
quick skim of the Apple TTF docs. I'll have a longer look later.
"
 

 I think it implies from 2nd grade curve construction. I have no definitive 
answer

from authority, but these resources believe in mid-way/half-way location..

http://fontforge.github.io/bezier.html


http://wiki.osdev.org/TrueType_Fonts







""

[crawshaw, 2015-03-08 18:46:49]

LGTM

Thanks for the explanation. Still a pity there are no tests for this.

[nigeltao, 2015-03-09 00:32:57]

*** Submitted as
https://code.google.com/p/freetype-go/source/detail?r=46c3056cafbb ***

freetype: fix drawContour when the first point is off the curve.

Fixes issue 16.

LGTM=crawshaw
R=crawshaw
CC=golang-codereviews, philipz
https://codereview.appspot.com/209280043