Wound up spending the whole afternoon tracking down an "only happens in production" bug where a sub-process would just hang, but only when big chunks of data were involved. Seems that os.write() was trying to write more than the target pipe could handle, so was winding up blocking. To tired to actually confirm that tonight, but if it winds up being confirmed, I'm going to have to go back and look at every case where I've used os.write(). I was under the assumption it would return immediately with the number of bytes written before blocking, something ...

I needed to phone my mother at one point this weekend.

Pulled up the Nexus phone dialer for one of the first times (I don't actually use my phone as a phone all that much). Said my mother's (full) name, saw a brief flash of her details, and they were promptly replaced with a bunch of we results for her name (lawyers in the UK and the like). Of course that's what I was looking for when I explicitly opened the contacts app and searched for someone in my contact list by name... of course.

Same thing ...

Booted into Windows to get the PyOpenGL xml-generation branch started on extension management. So far not working, as the extension querying mechanism (which is itself an extension) doesn't show up with wglGetProcAddress.  Annoying thing is that it *did* show up while debugging once, but not since then.  I'll likely find a few more bugs once I get that sorted too.

[Aside] I find myself very slow when I work on Win64; the setup just isn't as well-worn-in as under Linux... little things like not having the same tab-to-complete behavior, different ways to signal EOF to the python ...

My youngest threw my (ancient) Motorola Milestone (Android 2.1 era phone) across the room onto a tile floor. Amazingly, the phone still worked, and I've been using it broken for a while now. I almost talked myself into just keeping it, as after all, a few minor cuts on the finger-tip from broken glass don't really affect anything, until all the articles along the lines of "The Nexus 5 is a crazy-ridiculous bargain" started showing up.

I've had the phone for about 6 hours now, so obviously it's still first-blush, but it is definitely a ...

So it turns out that the xml-generation branch for PyOpenGL was entirely broken. I'd cleaned up a module (OpenGL/constants.py) that shouldn't have been necessary any more, as the types and constants were moved into the OpenGL/raw/GL hierarchy.  Oops, turns out that everything imports that module.  Spent a ridiculous amount of time fixing that (i.e. making everything point to the "correct" location) only to realize that I'd actually documented the use of OpenGL.constants from client code where the user wanted to get e.g. GLint, GLdouble, etc definitions but didn't want ...

So apparently I forgot to re-install the fglrx driver after updating my laptop to Kubuntu 13.10.  So as I've been testing PyOpenGL I've been amazed by how cr*ppy the GL performance has been, and I kept running into "this doesn't work at all" bugs for newer features. It did make me actually look at what GLSL versions were required for various shaders rather than just declaring "330" in everything because that's what AMD supports.

[Update] of course, the fglrx drivers (both stable and Beta) don't even load (segfault) so I suppose there's ...

The spammers finally convinced me that I have to restrict comments on the blog.  I was up to a few thousand blog-spams/day, enough that I was forgetting what I was going to blog about before I got through killing them all.  Comments are still open for 2 months, but trackback/pingbacks close within a few days.

Was stuck with no internet today, so the world gets a bit of work done on PyOpenGL...

So the branch in which PyOpenGL is generated directly from the Khronos API specification (xml) documents is almost workable. There are a lot of (internal) changes, particularly the OpenGL/raw APIs are pretty heavily rewritten, and I've landed a version of Roman Valov's patch that allows for proper extension querying on GLX (and EGL), as well as GL and GLU.  WGL support is still missing in the extension querying.

What the branch does have, however, is much more extensive coverage of ...

The spec-xml-generated wrappers are getting pretty close, only the glGet constant handling is preventing the test_core suite from passing, and I want to rework that to use the spec's enum groupings to provide better coverage of those and be able to flag missing GET enum specs on generation.  Haven't started into GLX/EGL/WGL etc yet, just working on getting the core (and GL extensions) working.

So I finally (with some prompting) sat down to create a code-generator based on the Khronos specification xml documents (the things that khronos uses to generate the official OpenGL (GL and GLES), EGL, GLX, and WGL headers.  Originally I was trying to use Khronos own generator code, but in the end I just wound up my own registry-loader and generators.

One thing that came up is that the type definitions needed to be split out.  Khronos just gloms them into the header documents, but in Python you really want your types specified in a separate document (since they are hand-coded ...