I hate reusing titles, and I've used the one up above (Testing, testing, 1.. 2... 3... ) before. But ah well... I'm lazy.
Sometimes Mike Gilbert (tdragger) and I talk about various posts we intend to write. Mike, being both a better write and a faster typist, usually beats me to the punch.
He's got one of those that we've talked about here. Go read it. Great post.
Anyway, in the comments section, Nick Landolfi, who is taking a game software development class, said:
"My impression was that there was only one test plan, and its purpose was to test the finished product. I never realized that there could be multiple test plans for individual features, or feature sets..."
I thought it might be interesting to talk about how that applies to at one of the art areas.
From the visual side, we've seen a real jump in what we're able to do to make things look more like their real world counterpart.
We saw a pretty big jump in visual quality from FS 95 to, oh, say, FS 2002.
But with the advent of more stuff (exteriors, transparent parts, interiors, more vertices, extra textures, new shaders, etc.) being available to simulate objects on screen, the accuracy of what we do also has to increase.
In FS 2002, much of the beginning and middle part of our visual aircraft development process was left up to individual artists. This gave us a mix of quality and visual accuracy in our models and textures. When we got to the end game, we saw a lot of bugs. Enough bugs that what we shipped was not representative of what we were capable of shipping.
So in FS 2004 we instituted some new procedures, one in particular meant that Test (QA) got involved in much earlier stages. We decided to establish mini "test milestones." First off, we'd have a test pass on the source we were using. For FS 2004 this managed to catch the fact that we were planning on modeling the wrong engine variant for the Ford Trimotor we were doing. This saved us a bunch of time later on-- changing a modeled and textured item takes a lot longer than doing it right the first time, and heaven forbid if it means that you have to ship the bug because it was too late or too risky to change.
Next steps up tested the unskinned model, the textured model, the animations, and then finally the LODs.
This adds some extra time to the art development process, but cuts down on bug fixing later, and raises the quality and consistency bar across the board.
Take a look at the default aircraft we shipped in FS 2004.
There are some aircraft that we carried over from the previous version, like the 747, 737, Bell 206, Cessna 172, and a few others. They didn't change visually from the previous version (aside from some repaints).
But the new aircraft that we did, they're a lot better. Compare the Lear in FS2000 to the Lear in FS 2004. Take a look at how nice the DC-3 is.
Now apply the same process to Flight Simulator X's default aircraft...
And a note for those among our users who never fly the default aircraft:
Raising the bar on the default aircraft raises the bar for everybody. ;)