VectorStorm Blog » Game in a Week

A Long-Awaited Post-Mortem

trevor — Fri, 13 Mar 2009 09:45:22 +0000

First, I would like to say that I’m not entirely happy with Lord’s state. It has a lot of things wrong with it that are immediately visible to anyone who plays it; the pacing is the biggest offender here, I think I totally failed to convey the feeling that I was striving for. But that’s not such a bad thing; the point, after all, of trying to make a game in a week is to finish a game in a week, not necessarily to have a fantastic game at the end of it.

No, the chief failure of Lord was that it took me more than 7 days to complete. In fact, it took me 77 days to complete (that number wasn’t planned). This is the story of what happened, in the hopes that telling the story will help me keep it from happening again.

It was the day after Christmas, 2008. Picture a younger me, more hale and hardy, his cheeks still flush and aglow, strong in limb and body, his hair not yet gone grey. Flush with his youthful vigour, he thought, “Why, it’s been a long time since I’ve done a Game in a Week.. and I have a full week of vacation left before I must return to work! And work will be quite busy, so I should have fun now while I can, for I will not be able to consider a Game in a Week for a long time into the New Year!”

More beneath the fold.

(...)
Read the rest of A Long-Awaited Post-Mortem (1,774 words)

Lord, Save Us From That Horrible Land, OSX release

trevor — Wed, 11 Mar 2009 13:25:52 +0000

Quick note that I’ve finally posted the OS X version of Lord, over in the sidebar. I’ll have a Win32 build of it there within 24 hours, but I wanted to get something official uploaded before I went to bed tonight.

There’ll be a lot more talking about Lord tomorrow, probably including a lengthy postmortem about what went wrong during its development, and about what I personally like and dislike about its design, now that I’ve actually had a chance to play it a few times.

So I’ll talk to you all again tomorrow, but right now it’s sleeping time!

Lord, first playable

trevor — Tue, 10 Mar 2009 11:15:48 +0000

Just thought I’d mention that I’ve just hit the “first playable” version of Lord, the increasingly incorrectly named “Game in a Week” that I’ve been working on in my spare hours for the past two or three months.

Ordinarily in a “Game in a Week”, I reach “first playable” somewhere between day five and day six, with a few occurring late on day seven. Since “Lord” is running so late, I’m probably going to give it just a little touch more polish, and then release it as-is. It’s not a fantastic game (though it’s not nearly as bad as I feared), but it’s a good proof of concept for 3D in the VectorStorm libraries, and I’m very interested to hear how well it runs for everyone. Depending on how much time work takes up, I’m hoping to have a package available for download as early as tomorrow evening; I only need to add a few small bits of UI and write the credits information, and it’ll be good to go.

And there’ll be a substantial post-mortem after this one, believe you me! :)

Lord status update – updated

trevor — Mon, 26 Jan 2009 07:11:57 +0000

This is just going to be a very quick update. It’s been by far my longest “Game in a Week” (currently at a duration of 30 days. Not that I’m keeping count or anything.)

I finally feel like I’m in the home stretch of development. There’s about six hours remaining in the day, and I very well just might make it.

Notable new stuff in the past few days: Got the Win32 build working. Fixed rendering on 3D hardware which doesn’t support VBOs (this’ll generally be 3D hardware that’s more than about five or six years old, or poorly supported laptop video cards… of course, Lord is rather processor-hungry due to its distorting-terrain-rendering system, so it isn’t likely to run very well on really old hardware… but at least it’ll render). Fixed even yet more core 3D rendering setup problem; rotation matrix maths and a flipped rendering frustum.

Also, I managed to spend some time fixing a couple of bugs in the Lord gameplay, when I wasn’t ridiculously busy fixing the game engine itself.

So hopefully I’ll have something ready to go in about six hours! Watch this space.

[update, four hours later] I’m fading early; not going to make it to midnight tonight. Lord is close to done, but it still needs some love on the progression between levels, and some general tuning. But it won’t be too much longer; I’ve bought myself a DVD I want to watch, and I’m saving it as a reward for finally getting Lord out the door. So it’ll be soon, depending on how many spare hours I get after work, over the coming days!

Where we’re at

trevor — Sun, 18 Jan 2009 12:08:32 +0000

Missed the three-week mark, too. Part of this was that I had to fix yet another core rendering problem which was causing 3D geometry to draw incorrectly, particularly in widescreen views. I also had yet another 3D picking bug to sort out, which took far too much time to crack (For the programmers out there, as you move across pixels on your screen, you’re not varying the angle from the view frustum linearly. I know, it sounds obvious when I say it, but that mistake was yielding lots of subtly wrong behaviour for me).

But with the above out of the way, I’ve made some awesome progress today. The core gameplay mechanic is now fully working. The world now wraps around at the edges (in fact, I usually can’t even tell where the edges are!). You can now use the mouse to scroll around the world, instead of using the keyboard.

Stuff remaining to be done: Simple models for a few types of building. Metagame structure. Sound effects. Port to Windows (should only require compiling up a Win32 SDL_Image dll). And I need to contact a particular band to find out whether I can use one of their old songs for the closing credit scroll. (I probably shouldn’t; Lord is already going to be huge, as its main gameplay music loop is about nine minutes long.. I shouldn’t add even more music data!)

Progress on Lord

trevor — Thu, 15 Jan 2009 11:33:03 +0000

Well, I think that I may have finally cracked the last major problem with VectorStorm’s 3D support; I had another pretty major bug in the rendering code, which was ending up having cameras render with the incorrect field of view (generally about 20-30% wider than it was supposed to be), which was completely throwing my mouse code for a loop.

I now have mouse-based 3D picking working correctly in Lord, which means that I believe the last major tech hurdle has been cleared. I’m now hoping to have Lord completed this coming weekend.

Of course, I haven’t actually prototyped the gameplay yet; all my spare dev time has been going into sorting out the tech issues. So even having run to three weeks instead of one, it’s still going to be a bit scary to see whether the game itself actually turns out to be fun! The time definitely hasn’t been going into fine-tuning the gameplay mechanics any more than usual…

Lessons learned from Lord

trevor — Sun, 11 Jan 2009 13:01:14 +0000

Of course, I can’t post a post-mortem until after I finish Lord, but I think I can safely say that VectorStorm wasn’t ready to have a 3D game made with it; particularly not on a short schedule, like my GiaW games have!

I was hoping to have Lord completed this weekend, but I’ve been running into one major engine bug after another all week. The most recent one is that the 3D camera code math was entirely broken. If there are any programmers out there using the VectorStorm library for anything, don’t trust any of the 3D code in it until I’ve given the all-clear; the camera<->OpenGL code was entirely broken, and I suspect that some of my 4×4 matrix math is still broken. The 2D API is still fine, but 3D is probably busted until further notice.

Stuff I can’t screenshot

trevor — Wed, 07 Jan 2009 12:46:14 +0000

I’ve been adding support for the OpenGL VBO extension to VectorStorm; this allows 2D and 3D geometry data to be uploaded directly to the video card, so that it may be rendered rapidly without hogging memory bandwidth between the CPU and the GPU. Which translated into non-technical terms means that VectorStorm can render far more without dropping frame rate, when using the new VBO classes (Note that currently, nothing is actually using these; I’ll need to adjust loaded display lists to use these automatically, instead of using the old, slower “vertex arrays” approach).

Note that the VBO extension isn’t universally supported; it requires OpenGL 1.5, which means that if your video card was made in the last four or five years, it probably supports it. But it’s not guaranteed. If VBOs aren’t supported, VectorStorm emulates the VBO functionality internally, and it should be no slower than it would be using the older rendering interface.

This new tech will eventually make it back into the VectorStorm trunk; probably sometime after I finish up this increasingly-mislabelled “Game in a Week”.

Rendering of the distorting Lord terrain map has been swapped over to use VBOs. This has sped up some of the terrain rendering code, but calculating the vertex distortion is still by far the most expensive part of the rendering; I’ll need to find an optimisation before Lord is released, or else it’ll have absurdly high system requirements for what it’s actually doing. Or I could bite the bullet and disable the terrain distortion; that’d speed everything up dramatically.

I’ve also picked music for Lord, have implemented a credits scroll, and started on an introductory sequence. But there’s still a fair whack to do before the game is finished.

trevor — Mon, 05 Jan 2009 12:25:30 +0000

“Lord, save us from that horrible land” screenshot for the day.

Added automatic mipmaps to VectorStorm textures. Major optimisations to map drawing. Added tonal variation to map coloration.

More engine bug fixes for the rendering of textured geometry.

Well, not quite a week

trevor — Sat, 03 Jan 2009 22:51:26 +0000

So I’ve actually slipped a day on this GiaW. Not going to say too much in this update; want to get back to work on it. However, it now seems pretty obvious to me that I rushed into this GiaW sooner than I should have; the 3D features of VectorStorm just weren’t quite solid enough to support rapidly developing a 3D game; too much of my time was lost debugging basic engine stuff, instead of developing the game itself.

Still hoping to have something by the end of the day today. But we’ll see.