At 12/9/16 06:15 PM, egg82 wrote: Meanwhile I've been busy building servers and tools for my work. Can't really show much since PCI and all, but I developed something kinda neat. All my work-related servers use GCloud for everything (server, sql, load balancing, etc) - minus CloudFlare. I wanted an easy way to patch these servers with the latest GRSec kernels.

Is the GRSec kernel widely used ? I've never heard of it before. Does it work well ? What kind of applications is it used for ?

By the way I got zoomin' workin'. Looks kinda weird because pixel art, but I think it's a necessary feature for the gameplay that I have planned.

At 12/17/16 12:19 AM, PrettyMuchBryce wrote: Maybe this is some sort of house or cave that actor A would like to enter. If the world around the house/cave is very large then A* will need to exhaustively search all of it only to discover there is no path.

In that example, it seems like pathing from the goal to the actor would solve it quickly, or is that example not very representative?

At 12/17/16 12:28 AM, PrettyMuchBryce wrote: Is the GRSec kernel widely used ? I've never heard of it before. Does it work well ? What kind of applications is it used for ?

It's relatively widely-used. It's the most popular hardened kernel and has saved many a person from zero-days. All the recent zero-days released for Linux in the last couple of months I've already looked at and gone "yeah, my servers are still immune"

Plus, it's free. Just gotta keep on top of updates.

At 12/17/16 01:11 AM, MSGhero wrote: In that example, it seems like pathing from the goal to the actor would solve it quickly, or is that example not very representative?

Sorry if I wasn't clear enough. Both of the actors are blocking. If you need a refresher of how A* works, check out this website and build a structure like this. Watch which nodes are explored. https://qiao.github.io/PathFinding.js/visual/

At 12/17/16 03:08 PM, PrettyMuchBryce wrote:

At 12/17/16 01:11 AM, MSGhero wrote: In that example, it seems like pathing from the goal to the actor would solve it quickly, or is that example not very representative?

Sorry if I wasn't clear enough. Both of the actors are blocking. If you need a refresher of how A* works, check out this website and build a structure like this. Watch which nodes are explored. https://qiao.github.io/PathFinding.js/visual/

No I got A*, what I meant was that if you started pathing from the goal, there would be no way for the path to "leave" the house because it's being blocked by other actors. And my point was that you'd return null a lot sooner than if you pathed from your starting point to inside the house.

In your picture, if you redefined your starting point to be the end, there would only be 8 nodes to check versus doing it the other way around and having to check 200 or however many.

We sometimes get people who sign up on our catering website and purchase $1 items to test out stolen cards.

I just caught one on friday; very cleverly done- using gmail instead of the usual 10minutemail or rocketmail or whatever else, using the first & last name on the profile (fake) and making purchases at several different locations for varying amounts of common items, usually more than $15.

The phone number registered with the account, by the way, belonged to Enron Corp.

As long as I don't get tracked down by Elliot, I'm cool

(I can't tell you how hard I busted out laughing when I figured the phone number out. Person behind the account is clearly a fan of Mr. Robot and knows exactly what they're doing)

Anyway, I'm still in the process of training a neural network on our backend system to find fraud accounts in real-time. Going pretty well so far, just takes forever to parse the website.

At 12/17/16 04:47 PM, MSGhero wrote: No I got A*, what I meant was that if you started pathing from the goal, there would be no way for the path to "leave" the house because it's being blocked by other actors.

Yes! You're right about that, but how do I know which to start and end from? Maybe the actor is inside the house and wishes to leave to a goal outside instead. Do I try both in the hopes that one finishes sooner ? Now I've increased my search time to O(2|E|). Maybe there is some kind of preliminary search I can do to figure out if the actor or the goal is in some enclosed space. Maybe a dijkstra search with some hard limit ?

At 12/17/16 05:26 PM, egg82 wrote: Anyway, I'm still in the process of training a neural network on our backend system to find fraud accounts in real-time. Going pretty well so far, just takes forever to parse the website.

Have you heard of SiftScience ? It does exactly this, and it is becoming somewhat of a standard at this point for merchants who accept credit cards directly.

hi.

At 12/17/16 06:15 PM, PrettyMuchBryce wrote: Yes! You're right about that, but how do I know which to start and end from? Maybe the actor is inside the house and wishes to leave to a goal outside instead. Do I try both in the hopes that one finishes sooner ? Now I've increased my search time to O(2|E|). Maybe there is some kind of preliminary search I can do to figure out if the actor or the goal is in some enclosed space. Maybe a dijkstra search with some hard limit ?

Maybe do 2 at the same time but limit the size of the second search's explored node list to cover common cases of being totally surrounded. I guess they don't have to be at the same time.

At 12/18/16 04:31 PM, Rudy wrote: hi.

Welcome back? You must have been before my time. Or when I took that break.

At 12/18/16 04:42 PM, MSGhero wrote:

At 12/18/16 04:31 PM, Rudy wrote: hi.

Welcome back? You must have been before my time. Or when I took that break.

I find myself back for a day to say hi and then i disappear again for years lol

At 12/18/16 04:42 PM, MSGhero wrote: Maybe do 2 at the same time but limit the size of the second search's explored node list to cover common cases of being totally surrounded. I guess they don't have to be at the same time.

Yep I tried basically this and it seems to be working so far. Not sure why I never tried this before. Good stuff!

At 12/17/16 06:24 PM, PrettyMuchBryce wrote: Have you heard of SiftScience ? It does exactly this, and it is becoming somewhat of a standard at this point for merchants who accept credit cards directly.

I have not, but does my employer reeeally need to know about that when I'm rolling my own?

I'll have to let our new website dev gal know about this. Looks really cool!

I've created gists for cron-building & applying latest GrSec kernels, if anyone feels like using and/or fixing them. The bash scripts themselves aren't pretty, but they get the job done.

Build machine prereqs
build_kernel.sh - run via cron once per day on a single "build server"
clean_kernel.sh - run once on client machines every time they boot
kernel_upgrade.sh - run via cron on client machines every day, a few hours after the build_kernel cron on the build machine runs (I went with 8 hours, though on a decent PC it should only take three at max)

Bucket "crontab -e" script (needs to be run as root):

The above crontab script only allows root access to the /mnt/build-bucket directory, by the way. Obviously not recommended that anyone else has read/write/exec access to that bucket.

Oh my gosh, Hyper Light Drifter used game maker studio. That's another one to add to the list to prove the tool doesn't matter.

At 12/26/16 07:31 PM, MSGhero wrote: Oh my gosh, Hyper Light Drifter used game maker studio. That's another one to add to the list to prove the tool doesn't matter.

I have the skill to take a car apart using nothing but a hammer, but that doesn't mean I should, or that I'm using the best tools ;)

At 12/26/16 07:31 PM, MSGhero wrote: Oh my gosh, Hyper Light Drifter used game maker studio. That's another one to add to the list to prove the tool doesn't matter.

Just looked that up, looks pretty cool. Very nicely stylised. How does it handle?

At 12/27/16 11:21 AM, TheEnkian wrote: Just looked that up, looks pretty cool. Very nicely stylised. How does it handle?

I forgot to pack my controller, and keyboard mouse takes some getting used to. There's no dialogue so I honestly can't describe what's going on, other than I just killed north crow boss.

It handles alright, though you really need upgrades to get that slick feel. I've only gotten a few, but it's improved the handling by a lot.

At 12/26/16 07:31 PM, MSGhero wrote: Oh my gosh, Hyper Light Drifter used game maker studio. That's another one to add to the list to prove the tool doesn't matter.

I mean it took them like 4 months post-release to get 60FPS support in that and required them rewriting a lot of the code, so I wouldn't say tool doesn't matter for that

At 12/27/16 08:05 PM, Glaiel-Gamer wrote: I mean it took them like 4 months post-release to get 60FPS support in that and required them rewriting a lot of the code, so I wouldn't say tool doesn't matter for that

I just saw that in the update log.

*Coding tool doesn't matter when making flashy stylish graphics and an adequate gameplay experience.

There's a lot of hidden content in this game.

I'm thinking through some user interface stuff now. Not designs, but just placement and interaction ideas. The bubbles are mouse-over, and the cards in the upper left appear when people are selected. Upper right is a leaderboard.

I also decided the people should lose body parts now so I am working on that. Because there wasn't enough complexity already. I'm hoping it will make combat a little more interesting.

Anyways, I'm all set for a 2020 release party so I hope you all can make it. I already bought the balloons and streamers.

Ooh, this gives me warm fuzzies :3

So many years of developing tools and games with nobody playing or using them, and it genuinely feels nice to be appreciated :)

Edit: They didn't show the coolest command, /vegetable :(

Took me forever to figure out how to code that

Ah, well. Can't win them all.

I stopped working on procgen terrain because I got somewhat stuck and had a lot of other things to work on. Now I'm back on it, and I remembered Voronoi tessellation. I found a haxe lib that already does Fortune's algo, and so I'll prolly end up using this to make distinct biomes.

Right now, I randomly splotch a biome onto a grid, but the tessellation will give me lines that aren't axis-aligned. Those will be transitions between biomes, and they basically don't exist right now using the splotching. I'll probably have to merge regions together to get a map with unique features, rather than each region always looking kinda the same.

For actual terrain, I'm hitting the issue of pathability. I can add something like Dijkstra to see if terrain is pathable or not, but if it's not, I have limited ideas on how to resolve it.

At 1/5/17 05:25 PM, MSGhero wrote: I stopped working on procgen terrain because I got somewhat stuck and had a lot of other things to work on. Now I'm back on it, and I remembered Voronoi tessellation. I found a haxe lib that already does Fortune's algo, and so I'll prolly end up using this to make distinct biomes.

Right now, I randomly splotch a biome onto a grid, but the tessellation will give me lines that aren't axis-aligned. Those will be transitions between biomes, and they basically don't exist right now using the splotching. I'll probably have to merge regions together to get a map with unique features, rather than each region always looking kinda the same.

For actual terrain, I'm hitting the issue of pathability. I can add something like Dijkstra to see if terrain is pathable or not, but if it's not, I have limited ideas on how to resolve it.

i got this bookmarked for cool effects you can do with voronoi diagrams. I've tried to implement that effect in a game as a particle effect but it ended up being too much work to make it look good with the more random nature of particle effects compared to whatever they did here where every particle seems to be some minimum distance from the other ones and theres a randomish grid of them everywhere else in the scene

still its pretty cool

At 1/6/17 02:42 AM, Glaiel-Gamer wrote: i got this bookmarked

Well shit, that's cool.

So far, I ended up not using the algo at all. Point-in-polygon seems like a waste when I could just find the minimum distance from every tile to one of the centroid points. So my terrain makes a Voronoi diagram, but I don't use the polys that get generated at all. I might need them or the edges later, but idk yet.

@TheEnkian I have no desire to upload to the dump and make a new PM right now so, that's what it looks like without fiddling with anything. Zoomed out, there are anywhere from 5 to 11 individual regions.

At 1/6/17 03:57 AM, MSGhero wrote: TheEnkian I have no desire to upload to the dump and make a new PM right now so, that's what it looks like without fiddling with anything. Zoomed out, there are anywhere from 5 to 11 individual regions.

We might need to do something about the hills eventually and we'll get the terrain overlap tiles in for the grass, otherwise it looks fine to me.

At 1/5/17 05:25 PM, MSGhero wrote: For actual terrain, I'm hitting the issue of pathability. I can add something like Dijkstra to see if terrain is pathable or not, but if it's not, I have limited ideas on how to resolve it.

You know.. I kinda wonder if this isn't something a neural net can't solve?

.. Yeah, I gotta try this now. I think this might be a better alternative to something like A* while keeping the benefits. I will have a demo with performance tests!

At 1/6/17 04:07 PM, egg82 wrote: You know.. I kinda wonder if this isn't something a neural net can't solve?

.. Yeah, I gotta try this now. I think this might be a better alternative to something like A* while keeping the benefits. I will have a demo with performance tests!

Ugh, I'd love to never use ANNs again. I had a ML class, and those took hours to learn the data. Running them is fast of course, but my gosh and the tool wasn't multithreaded either.