ofBox 3dPrimitive winding order inconsistent

[tgfrerer]

All primitives but ofBoxPrimitive appear to be in CW winding order, resulting in front faces to be culled instead of backfaces, when glCullFace is enabled.

test case:

https://gist.github.com/tgfrerer/7417271

test case render

related issues

We have dealt with winding isssues in the past, with #2263 being the most recent.
There is a vintage open issue addressing this: #745

thoughts

I could fix/unify this, but i think it's worth discussing which winding order we adopt as standard. I'd personally be for going with the OpenGL standard of CCW for front-facing triangles, which would mean to bring everyting in line with ofBoxPrimitive. my main reason being that these kind of special cases could make oF a bit more difficult to learn / and code harder to port.

[arturoc]

the only problem i see is with polygons we've always done CW but i guess it would be ok to change them too?

[tgfrerer]

Yeah, really tricky, I'd be totally for it, although i fear it might glitch people's old projects. We'd have to be very explicit about this in the changelog. One alternative thing to consider would be to setup OpenGL using glFrontFace explicitly to:

glFrontFace(GL_CW); // default is: glFrontFace(GL_CCW);

But generally, I think it could be nice to clean the API to stick to OpenGL defaults/expectations, to keep with the principle of least surprise. Also, keeping to the standard would make it easier to port from other sources.

My concern with the glFrontFace-method would be that it would be an "invisible" fix for beginners, but a really hard to find/debug openFrameworks 'special case' for more advanced users.

I've googled to see whether the polygon CW winding might historically have something to do with cairo / svg and negative space / polygon fill rules, and it seems to me that svg uses CW winding: http://www.w3.org/TR/SVG/painting.html#FillRuleProperty
but then, most literature (and the default OpenGL render settings) go counter-clockwise for polygons, too. Hm...

[bilderbuchi]

so, @tgfrerer, has this issue been closed by the merged PR #2267?

[bilderbuchi]

so, @tgfrerer, has this issue been closed by the merged PR #2267?

[tgfrerer]

not yet, afaik, the case is still awaiting consensus. PR #2267 brought the icosphere in line to use (CW) like most other default primitives, but box still uses CCW, i believe. I can't run the test case just now, but it would give us a definite answer on the state of things…

[bilderbuchi]

I don't think even the `IcoSpherePrimitive is working correctly. I'm on current master (815b2b7), and if I add

    ofPushMatrix();
    ofTranslate(20, -20);
    ofIcoSpherePrimitive(5,2).getMesh().draw();
    ofPopMatrix();

to your test case, I get

which looks wrong to me.

[tgfrerer]

Right, icosphere is CW, like sphere - and all but ofBox, which is wound CCW.

If I remember correctly, the PR you mentioned brought icosphere in line with the others(CW), so the render seems ok to me, but we still should decide, I think, whether we want all primitives CCW (like ofBox, and OpenGL default) or CW (like openFrameworks by tradition), just to at least be consistent.

Also important to notice that one can switch opengl to use CW for front faces by calling glFrontFace(GL_CW); or so before culling. It's just something that can trip people up.

Cheers

T

typed on toucg

On 21 Jan 2014, at 15:37, Christoph Buchner notifications@github.com wrote:

I don't think even the `IcoSpherePrimitive is working correctly. I'm on current master (815b2b7), and if I add

ofPushMatrix();
ofTranslate(20, -20);
ofIcoSpherePrimitive(5,2).getMesh().draw();
ofPopMatrix();

to your minimal example, I get

which looks wrong to me.

—
Reply to this email directly or view it on GitHub.

[ofTheo]

moving this to 0.8.2 and pinging @NickHardeman as he did most of the primitives stuff.
agree this is a tough one to change. switching the order to ccw could potentially break a lot of projects.