roncli.com blog

For some time now, I've been a Collaborator on the DXX-Retro project. Originally I just started helping out on various odds and ends, but lately my focus has been on the observer mode of the game.

Over the past two years, I've essentially been waiting on the project owner, Drakona, to work up the networking code to send client ship status to the observers. It's one thing to see the ships flying around in observer mode, but another entirely to know everything about those ships. That is, shields, weapon loadout, what weapons you're about to fire, etc. Unfortunately, her last commit to the project was February. Of 2016.

I get that priorities in life change, especially when you become a two-time parent. :) The Descent community, however, hungers for more, and as a result I've decided to finish off observer mode in Drakona's absence. There's actually quite a lot to make that happen, so I've created the v1.4 Roadmap. In this post, I discuss the rationale behind the roadmap, and what to expect from Retro in the future.

DXX-Retro has largely served its purpose to the competitive community, but two outstanding features remain to be added. Both of these were promised by Drakona to the late JinX. One of these features, observer mode, was even kind of named after him, "JinX Mode". The other feature was specific for Descent 2, and that is Capture the Flag Classic, which brings D3-style CTF play to D2.

In February of 2016, Drakona and I released an early version of DXX-Retro v1.4X5, which gave us a taste of observer mode. This only gave us the visuals. While we could see ships flying around, we didn't know anything about those ships other than what their opponents knew. This was always wanted, but we never really came up with a plan to do it. And, in the past 2 years, nothing has been worked on to support this. Unfortunately, Drakona just hasn't had time to work on it, and my work has therefore stalled.

While it hasn't been a great situation, the community has gotten a lot of mileage out of observer mode. So much so that a Twitch show was made out of it which continues to enjoy success among the community. However, there's a lot that we miss on the broadcasts because we don't have those shield numbers, the weapon loadouts, and whatnot. And after two years, the time has come to just get it done.

So, I'm getting it done. The roadmap includes 5 future releases along the 1.4 line. The first will be v1.4X6, which I'm hoping to finish this week. This will include the shield numbers appended to the pilot names under the ship. I want to also include a way to indicate damage, and will be working on ideas for that.

The second release will be v1.4X7, which will get the rest of the ship status in the game and make any necessary upgrades to the observer mode UI that we want to change from previous versions. I'm *considering* adding my custom-made Observatory UI to Retro, but am not sure that's needed.

The third release will also be v1.4X7, but this time for Descent 2! Once I know what I'm doing in D1, I can apply that to D2, and we can have an observer mode for both games.

The fourth release will be v1.4X8, also only for Descent 2. This will include Capture the Flag Classic, where instead of the "egghunt" style game where flags are worth 5 and kills are worth 1, we'll instead make the flags spawn in the base, only have teams be able to score if their flag is in their base, and have team points be based on flag captures only.

And finally v1.4 will be the final release for the 1.4 line, which will include any bug fixes, documentation updates, and any other housecleaning duties we need to do for a proper release.

After that, the future for Retro is murky at best. There are no less than 50 open issues on GitHub for Retro right now which vary from essential to nice-to-have to outright silly. Also to keep in mind, the DXX-Rebirth is nearing a v0.60 release that is essentially a codebase refactoring in addition to a ton of other updates. I would love to see a Retro v2.0 not only based on Rebirth v0.60 but something that can also follow that version as well, instead of being a complete fork that forgets where it came from. This in and of itself would be a monumental task.

Until then, however, we'll keep Retro focused on what we need to release v1.4, and if I can keep it going without the need for added help, perhaps that will be much sooner on the horizon than expected. Here's to some better Descent in the near future!

Labels: Coding, Descent, Gaming, Retro

Wednesday, May 06, 2015

The problem with mice in Descent
Posted: 2:07:00 AM 0 comments

Descent was an awesome series. Hell, it still IS an awesome series. Six degrees of freedom, claustrophobic tunnels, murderous robots... what's not to love about this game?

The fact that it was made in 1995 for one. Originally written for MS-DOS, Descent was designed to run on computers that didn't know what a Pentium was. It ran great on those computers, but as computer processing power increased, it started to become obvious that the game was not written to scale well.

At first, the homing missile became impossible to dodge. Written to work fine at 30 frames per second, the homing missile began to turn faster and faster as that 30 became 60, 60 became 100, and 100 became 200. Fortunately, this bug has been fixed and has made playing Descent quite bearable.

However, recently, a new bug reared its ugly head involving mouse play. Descent plays at a constant 200 frames per second, assuming you don't have vsync on, or have otherwise limited the frame rate. The average mouse in Windows is only polled at the rate of 125 times per second. So, your graphics card is getting more frames per second than Windows is getting mouse positions. How does this cause problems in Descent?

One of the quirks about Descent is that, unlike traditional first person shooters, you are not allowed to just 180 your ship in an instant. The Pyro has a maximum turn rate, and is hard-limited in the code, throwing out extra movements that are made beyond the maximum allowed. This means that you can whip your mouse around the mousepad, and you won't move very much.

The turn rate for the mouse, however, is determined per frame. So, what'll happen is you'll move the mouse, say, an inch, within 2 frames. You only need to move it, say, a quarter of an inch to reach maximum turn speed. At the end of the first frame, the first quarter of an inch is processed and the next quarter of an inch is thrown out. Same thing for the second frame, the next quarter of an inch is processed and the final quarter of an inch is thrown out.

However, let's say in the first frame, your mouse didn't get polled at all. At the end of the first frame, Descent - and Windows for that matter - think you haven't moved the mouse. Your Pyro does not move at all. In the second frame, the mouse is polled and sees you moved an entire inch. The first quarter of an inch is processed, and the other three quarters are thrown out. All that movement for nothing!

Having a fellow pilot suffer greatly because of these issues, I decided to dive into the source code and develop a fix. Now what happens is that you turn on the Mouse Overrun Buffer setting, and then instead of throwing out the extra movements of your mouse, it is stored in a buffer, with anywhere between 1/16 of a second and a full second of movement stored, depending on how high you decide to set your buffer. Then, in the next frame, anything already in your buffer is added to whatever mouse movement was detected that frame, and that is used to determine how much your pyro moves, storing any excess back into the buffer again.

You can find this code in a pull request for DXX-Retro, a source port of Descent dedicated to the competitive community. If you own the original game, you can find downloads for DXX-Retro at the Descent Rangers download page.

And if you don't, go get it! Classic first-person shooter action.

Labels: Descent, Descent Rangers, Gaming, Mouse, Retro