La Salle Debain

The hacklab is so cold I'm going to lose some of my toes to frostbite.

Today I spent most of my time trying to track down the DHCP problem. Phil Kern suggested it might be due to the output of Columbias DHCP server which certainly sounds plausible but without tools such as strace I'm rather at a loss as to how to figure out what exactly is going on. If anyone would like to look at the source to the Minix DNS caching thingy which is where the hang actually occurs, I would appreciate it.

Other than that I also ported iconv, gettext, and grep.

Actually I was at the venue on Tuesday for a little while but my name wasn't in the system either because I has given my name in Unicode or, more likely, I had forgotten to check the "reconfirm" box on the registration form. For some reason Columbia U's DHCP server makes Minix lock up on boot. Also I left my laptops AC adapter at home. Some kind souls lent me their European adapters but they didn't work. (I have a set of international converters. Also unfortunately at home.) So after about three hours, my battery ran out and I had to leave.

So we will pretend that never happened and instead call today day 1. (Wednesday I had some unexpected family stuff to attend to so that doesn't count either.)

Today I first concentrated on tracking down why the boot process was hanging. The culprit seems to be nonamed a daemon Minix runs to do DNS resolution/caching etc. I'm consulting with the Minix devs to see what can be done about this. I added the cool MOTD banner you see above (made by Osmo). And I ported some more GNU software to match the versions in Squeeze. Tomorrow, the biggies—gcc and perl, and seeing if I can get a dpkg binary that doesn't segfault.

Thanks to everyone who wished me a happy birthday. This is my present -- completely vegetarian tiramisu. (I'm getting a longer-lasting present too but it hasn't arrived yet.)

Old favorite German word: weltschmerz

New favorite German word: lustmord

I'm afraid that in the end I failed the challenge. As the last minutes of Saturday night ticked down, I was still mired in the minutiae of C++ Run-Time Type Identification (RTTI.) Sunday was my 12th wedding anniversary and I know from past experience that this is one day when it is advisable to step away from the computer. So I have to concede defeat for now. But I only actually spent about 48 hours on this project this week so if you really think about it, I still have 120 hours available. That's what I choose to believe anyway.

The galling thing is in theory I know how this is supposed to work. A.L.F.s model (in the model-view-controller design pattern) is a class called World. World has an array of Rooms which represent rooms in the dungeon. In my original almost working version of a few days back, this Room class had all manner of verious members for different objects in the dungeon but I decided to redesign it the right way by consolidating all the redundant bits into an abstract base class called Item which would have concrete subclasses: Potion, Treasure, Monster etc. Room now only has to hold a pointer to the Item associated with that room. (Later on I will consider allowing multiple items per room but for now there can only be one.) A factory function creates Items and assigns them to each Room (or leaves them empty) It returns a pointer to an object of the appropriate class but we can access that object through the Item* because they are derived from Item.

So far so good. The fun part happens when we want to know what the Item* really is. For example, combat makes sense for Monsters but not for Treasure. I only want to activate combat if the Item* is a Monster*. It's easy in theory:

if(dynamic_cast<Monster*>(contents))
    return FIGHT;
                            

...but it didn't work. Problem number 1. C++ RTTI only works on polymorphous classes. (Ones that contain atleast one virtual member.) So a little redesign of my classes was necessary. Problem #2, once you start slinging pointers around, segfaults soon follow. I had to go back and make absolutely sure I wasn't double-deleting objects and things of that nature. As I write this, I think I have finally got it all sorted out. So I am going to persevere and finish this game, its just going to take a bit longer than 7 days.

So I've got combat, I've got basic potions that restore health and I've got items such as weapons or shields which can affect your combat stats. Try as I might, I cannot get ncurses to give me a yellow background for some unfathomable reason. It just comes out brown no matter what. Ironically, in the next Konsole over, I have the alpine ncurses-based mua taunting me with a yellow background. I'll look in its source (yay open source!) to see how its done sometime but for now brown will have to do.

C++ is proving a joy to work with as always. Ok It's probably just me; I use it so infrequently, I keep having to relearn how to do multiple inheritance and abstract base classes each time. I am rather proud of figuring out how to access class members from a signal handler without having to look it up.

Now if I can quickly tie up some loose ends, I can spend the last two days maybe making the dungeon generation more roguelike.

I am still being plagued by the burdens of real life so most development is taking place after the kids ("It should have mermaids! And Spiderman!") have gone to sleep. By the time I finished up yesterday, I was too tired to blog so this is is a combined report on the last two days.

I decided not to bother with Javascript as I want to try a different project for the plasmoid competition. So I'm back to C++ and ncurses. At this point my '@' can move across the dungeon which at the moment a very simplistic 10×10 grid where every space is a 'room'. Rooms are populated with monsters, treasure and items though at the moment only the monsters can do anything and by anything I mean fighting. My combat system is based on Solo Dungeon Bash and pretty challenging I think. Hopefully by tomorrow, A Lively Fisting will be playable though I really need to pick up the pace if I want to finish on time.

I've decided to take part in the 2010 Seven Day Roguelike Challenge

My entry will be written in JavaScript as a KDE4 plasmoid (thus simultaneously allowing me to take part in this competition)

Unfortunately today had to be spent in RL family activities so I'm already behind. however I've done lots of research into Roguelike development and I have accomplished the first step—finding a name. Candidates are:

• Descent into the No-No Area
• Going Rogue
• Battle For Quiznosth
• Tol-in-Gaurhoth

...and the winner is:

• A Lively Fisting

It will be a simple combat heavy coffeebreak game inspired by things such as Solo Dungeon Bash and Pocket Dungeon though as time permits I will add as many more roguelike features as I can.

This project blows ass.