Kernel Panic

[Columbus]

I think there is a problem with the Network's Packet spawning, because i've noticed, when there is no more free space left around a Buffer or Connection, while dispatching, the game crashes. Or is it just me?

[KDR_11k]

I think there is a problem with the Network's Packet spawning, because i've noticed, when there is no more free space left around a Buffer or Connection, while dispatching, the game crashes. Or is it just me?

I know it happens but it shouldn't happen anymore, or at least only 1/10000th as often as it used to. It's an engine bug (caused by two units occupying the exact same space) that's fixed in SVN and KP should have a workaround.

[Soulless1]

Wow, spring respects the Pauli exclusion principle? Nice ^_^

[KDR_11k]

http://spring.jobjol.nl/show_file.php?id=973

I forgot updating this thread: RC4 is out, it fixes the bug maneuver ranges and adds the option to hold ALT for the bombard command to not give it an attack order (basically a "look for targets here" command).

[Sertse]

Minor request.

Could you include the "AI" gadget in subsequent releases?

EDIT: Random fact, but it appears RAI can "play" KP. (AAI and KAIK just stands still and does nothing).

Interestingly 2 RAIs (allied) can defeat the KP "AI" posted in the previous page... 1 on 1 RAI was defeated very quickly.

I tested on Marble Madness. KP AI on the bottom right corner, RAI allied and taking the top two. RAI were both playing as System.

RAI adopts a slow and very defensive "strategy" (if we can call it that), building Bits only in the very early game and then concentrating solely Bytes and Pointers. One quirk with RAI is that its constructors only build terminals....but don't know how to use them.

Anyways KP AI expanded very quickly, swarming the place with Bits (KP AI, which plays KP somewhat "normally" is mainly about Bits, with a few Bytes). In a 2 on 1 situation, the RAIs have managed to build enough terminals to "distract" the KPAI's Bits into attacking them... giving RAI the time needed to a mass a army of bytes and pointers.

At one point KP AI had nearly taken over the whole map, and was going for the RAI's kernel, where it found itself facing a rather prolonged battle with the RAI's concentrated force of bytes and pointers... that it has turtled about through the whole game.

The RAI managed to wear down the assault and began its counterattack... and basically its force of high tech units overwhelmed the KPAI.

Heh, storytelling is not my best suit, but seriously it was one of the most exciting battles I've watched.

[Sertse]

More crazy AI observations.

Would there be any interaction effects if I made RAI play as "System AI/KP AI"?. Seeing as I'm practically asking an AI to play as an AI...

Ran more games today, and perhaps it is just a placebo effect, but I think... under this arrangement, it was more competitive than just the System AI.

On marble madness, same conditions as above. RAI managed to beat (just) KPAI even in 1 on 1 at times, unlike before where it always loses. For one thing, the AI didn't just build useless (as it doesn't know how to use it) terminals...

I'll run more tests when I have the time, if anyone is interested lol

Edit:

Tried KAIK+System AI combo. Same conditions. KAIK alone just sit there, as I've said before... but comboed with System AI, it seems to do stuff. But maybe it just because of System AI. KAIK has a tendency to have its assemblers constantly build debugs...

Might also try changing starting positions a bit and see if it makes a difference

[KDR_11k]

I think stacking the two AIs would cause confusion for both as they'd both run different plans. You should retry that match on DMA, differences on that map include more geos and geos under the kernels, I think that should result in KPAI dropping a nuke on the kernel when attacking (currently kernel attacks are bit-only, done on purpose to make it a bit easier for the player). It's designed to make bytes first and then slowly migrate to pointers (which drop NX flags on fortified positions and counter-attack enemy bytes and pointers), I guess on smaller maps it never reaches the pointer stage.

[KDR_11k]

Okay, there was some back and forth complaining about the balance in a ladder game, can you people put complaints here so I can see all of them?

[Saktoth]

Let it mature KDR. Flop insists hacker is UP and worm sucks, i insist hacker is OP and worm is OP. I thought network were UP, especially on small maps, but now i realise flow rush rocks etc etc.

The balance needs time to settle in.

[Pxtl]

To me, the big difference is the macro/micro differentiation between factions. Two lazy players will find the System overpowered, while two pros with APM's out the wazoo will probably kick ass with Evil.

[Neddie]

It's been a while since you played. There are three sides.

System
Network
Hacker

Given the forty games played today alone it seems that opinion is shifting toward balance.

Also, join us on the KP ladders!

[Pxtl]

I've dabbled a little with the newer releases, but haven't gotten enough in to get a good feel for Network.

[Neddie]

Get online and come play right now. Ladder match, 2v2.

[KDR_11k]

[url]http://kdr_11k.from-hell.net/Kernel%20Panic%20net%20NRC5.sd7[/url]
I'll upload it to a common site tomorrow

[Felix the Cat]

I found this line in a doc here and it made me lol:

The symptoms can include application core dumps and kernel panic.

late edit. WTF kernel panic actually exists as a technical term. /me knows absolutely nothing about UNIX... and doesn't care to know any more...

[KDR_11k]

Well, here's the changelog. It's called NRC because it had some larger changes and I don't consider it a release candidate anymore but changing the numbering again would just confuse people.

KP.net NRC5
- Matched internal names of network units to the visible names (internal only)
- Changed connection weapon to not use burst, it will fire only one shot now. This prevents target switching and making stuff go airborne. Upped the connection's mass since it no longer needs to be the lightest unit around.
- Connection damage per attack reduced (3000->2500), rate of fire decreased (1/3 -> 1/4)
- Worm damage vs single targets increased (1000 -> 2500), area of effect for splash damage increased (160 -> 210)
- Increased Packet HP (400 -> 500)
- Increased dispatch epsilon by several orders of magnitude (from a fraction of an elmo to 1 elmo) because crashes were still occurring and I suspect it was because the position gets rounded
- Made Kernel open faster, it was way too slow at opening before

[zwzsg]

- Made Kernel open faster, it was way too slow at opening before

Actually, it was a feature. So everybody gets a few seconds at start to enqueue build order. Because I don't think it's funny to gain an advantage to be the fastest to click when game start. So I'd rather have other homebase open slower.

[KDR_11k]

The other two only start opening when you give orders though.

[KDR_11k]

[url]http://kdr_11k.from-hell.net/Kernel%20Panic%20net%20NRC6.sd7[/url]

KP.net NRC6
- Increased Packet damage per shot (90->130)
- Removed bug seismic detection (it was useful back when the hacker needed bugbombs to survive the early game but now it's unnecessary)
- Added 20% bonus to the Kernel (and equivalent) build speed for every geo their owner controls

The last part may be pretty daring, let's see how this pans out.

[zwzsg]

- Removed bug seismic detection (it was useful back when the hacker needed bugbombs to survive the early game but now it's unnecessary)

But the crosses were so pretty!