Mike Owens @ Filespanker / Blog

Photo

Fri, 17 Dec 2010 02:59:33 -0500

Photo

Thu, 09 Dec 2010 18:10:55 -0500

Bug 31657 - Unbounded recording surface -> image surface fill accesses uninitialized memory

Tue, 16 Nov 2010 09:27:47 -0500

Bug 31657 - Unbounded recording surface -> image surface fill accesses uninitialized memory

I don’t know how I got any performance tuning done before callgrind. Actually, I do know: I...

Mon, 08 Nov 2010 21:30:50 -0500

I don’t know how I got any performance tuning done before callgrind.

Actually, I do know: I used -pg with gprof.

I don’t know how I got any performance tuning done before callgrind.

SVG in Ruby vs. C++

Mon, 08 Nov 2010 03:10:00 -0500

(paraphrasing)

Ruby:


    def resolve_property(p)
      return send(p) if respond_to?(p)
      return parent.resolve_property(p) unless parent.nil?
      return class_for_property(p).new
    end

C++:


    /* 400 lines of resolve_property_ref::whatever 
       omitted.  */`

I promised myself I’d never write another line of autotools noise after an m4 explosion in the...

Mon, 25 Oct 2010 06:50:18 -0400

I promised myself I’d never write another line of autotools noise after an m4 explosion in the code mines a few years back.

My first thought was to see how usable Rake was in a non-Rails context. Turns out, it’s awesome for building C++ even outside of the pretty example rules in the docs.

I’ve noticed that, after a three year break in Ruby-land, my first big C++ project since uses...

Tue, 19 Oct 2010 05:09:00 -0400

I’ve noticed that, after a three year break in Ruby-land, my first big C++ project since uses a hell of a lot more dynamic_casts, and I’ve found a use a few real uses for typeid().

View cones + line of sight.

Thu, 14 Oct 2010 22:56:46 -0400

View cones + line of sight.

Needed a generic, simple “bad dude” to get actors...

Wed, 13 Oct 2010 19:03:40 -0400

Needed a generic, simple “bad dude” to get actors doing their thing. Drew up a Goomba.

Same vertices triangulated by poly2tri. (Note, in both of these...

Tue, 12 Oct 2010 15:57:27 -0400

Same vertices triangulated by poly2tri. (Note, in both of these photos, the visual paths don’t match the vertices submitted to the triangulators because they’ve been simplified before tessellating.)

A stump triangulated by gluTess*

Tue, 12 Oct 2010 15:56:00 -0400

A stump triangulated by gluTess*

rbpoly2tri: Ruby Bindings to poly2tri, A 2D constrained Delaunay triangulation library

Tue, 12 Oct 2010 15:55:00 -0400

rbpoly2tri: Ruby Bindings to poly2tri, A 2D constrained Delaunay triangulation library:

And with this, I totally rid myself of both gluTess and GEOS.

rbclipper

Mon, 11 Oct 2010 15:08:36 -0400

rbclipper:

I’m tired of screwing with complicated clippers with dumb licenses.

http://trac.osgeo.org/geos/ticket/379

Mon, 04 Oct 2010 23:21:08 -0400

http://trac.osgeo.org/geos/ticket/379

Pulled my hair our for a day getting transforms working. ...

Mon, 04 Oct 2010 19:45:00 -0400

Pulled my hair our for a day getting transforms working. Working physics, bouncing around crates, etc on the see-saw. Time-til-first crate: 2s.

sigh...

Mon, 04 Oct 2010 02:26:28 -0400

sigh...

mieko's rbclutter-unofficial at clutter-1.4 - GitHub

Sun, 03 Oct 2010 19:22:14 -0400

mieko's rbclutter-unofficial at clutter-1.4 - GitHub:

Since the move to Clutter, I’ve been randomly patching rbclutter to support 1.2+ features. Thanks to bpeel for rbclutter to begin with, and with some advice on my (slow) quest to get it up to speed with Clutter proper.

There are still some minor hacks in there from my days of “just get it working” (implementing dup instead of initialize_copy, reusing VALUES without much care to the GObject refct, etc), but it mostly works, and has support for a few newer things that I’ve needed.

Spent two days trying to get a half-decent boolean polygon...

Thu, 30 Sep 2010 05:14:48 -0400

Spent two days trying to get a half-decent boolean polygon library working in Ruby. Ended up using the SWIG bindings to the PostGIS geos library, which are amazingly buggy. In this photo, you see two stumps, the one on the left is made of 7 or so paths, the one on the right, a single path.

My goal was to allow objects on the visual layer to have physical geometry, but the constraints of the geometry layer are lifted, as objects tend to be drawn as to look good, not have a low poly count. The algorithm here does a boolean union on graphical polys before generating a physics poly, so the left and right stumps have an identical wireframe.

This fix took two days, including a set of FFI bindings to GEOS, which were abandoned because GEOS has an amazingly crazy API. Moving on…

I’ve noticed I’m in this pattern of writing all this...

Fri, 24 Sep 2010 01:23:00 -0400

I’ve noticed I’m in this pattern of writing all this complicated code, getting it working, then finding a way of pushing it off by getting the same functionality from a library I’m already using. In this example, I’ve replaced my GeometryPen with Cairo’s cairo_copy_path_flat.

I was under the impression that the tolerance could not be set on curves, but cairo_set_tolerance is sitting right there, fully documented. For the most part, the functionality is the same, but Cairo by itself is a bit smarter than my solution regarding circles. My implementation converted them into arcs in an almost cool manner, but by the time they came out, they were sliced like pizzas in the physics engine, which wasn’t required, as it’s still a simple polygon. Cairo generates some nice stop-signs.

Also in the photo: misaligned parallax scrolling. On the long list of TODO.

Day 20: Clutter, Parallax Scrolling No updates in a few days,...

Wed, 22 Sep 2010 18:23:00 -0400

Day 20: Clutter, Parallax Scrolling

No updates in a few days, but I was really busy converting the engine over to Clutter/rbclutter, and it was the best decision I’ve made. Once I knew I was about to have to get into texture slicing, I decided I needed some backup on this scene-graph stuff.

Clutter has a 2D scene graph over OpenGL, and seems to be performing than my hacked together layer system.

I’m to a point where screenshots aren’t really capturing what’s new: In this one, we have a parallax scrolling background. The pink rectangle is a sensor in the “markers” layer, rigged to the level-scripting system that already works.