Weapon redraw issues and solution

I appreciate the effort to make the changes and your effort to keep us informed.

Khaaaannnn!!!

This.

<sob>

Estimated duration of mind control ray effects: 20 hours 18 minutes.

I really should slot for +Resistance

Invulnerability doesn't have Resistance to Psi.

Go get some set bonuses, you cheapskate.

Its probably related to the bug I reported to the devs several times with regard to certain powers unrooting when they are executed under certain conditions. The common denominator was executing the attack while the player was moving in some unusual way, which caused the power to skip its animation. Because rooting is tied to the animation, if you skip the animation, you aren't rooted even if the power otherwise executes.

It doesn't have to be superjump, I posted a video of shadow maul and sands of mu (both use the same animation) running completely unrooted without specifically superjumping, and without even needing to get off the ground at all in some cases. Unfortunately, the video upload converted badly, and the video is a bit unclear (although I still have the original and it was quite sharp: don't know what youtube did that time around).

I never found out what caused that bug, nor am I sure if it was ever even corrected. I have suspicions, but none I can prove (and its been a long time since I've looked at this problem specifically). I should try to re-upload the original and see if it comes out better this time.

We know you can do it! You found a way to do power customization, you can figure out how to kill Redraw!!!

so basically BAB just Back To Formula 'd us. ... it's the new rick roll

Here here! You can figure this one out!

Over the years I've done a lot of work in trying to make the sequencers as bullet-proof as possible. Especially the moves related to power animations, as those moves set rooting and impact gameplay. At this point, attack animations are really high up on the priority of things and can even interrupt each other...even though it should be impossible for an attack move to be triggered while another attack move is playing, due to rooting and power activation times.

I remember at PAX a couple of years ago watching and playing our game on their tenuous connections and being dismayed at how frequently attack animations just completely failed to play. Even in situations where there wasn't any good reason why they shouldn't play. Ultimately, it's the server that tells the client what move it should be playing, and I kind of assume that something as simple as a lost packet at just the right moment, or missed tick, can cause the sequencers to reject playing a move. This type of stuff has proven to be nigh-impossible to replicate on a local client/mapserver environment...which is the only environment where I can actually monitor the state machine of both the mapserver and client and see what's happening.

So, while I don't doubt that animations fail to play, I think the vast majority of times where that happens are situations that I simply have no control over. That's not to say that over the years people haven't sent me very specific, step by step reproducable steps for causing an animation to skip...or play the wrong animation. But those have been few and far between compared to the general "animations fail to play all the time" reports I see every so often here on the forums.

Shouldn't BaB's be retconned into Willpower...I mean how on earth can he be invulnerable...or his is tanktop made of bulletproof material?

One possible suggestion; could you set up your back-end tools on the EU Training Room? The network path from California to Germany and back will give you a predictable lag element (speed of light delay, approximately 0.05 sec each way, plus switching losses), and depending on the connection possibly a variety of other issues.

You could also try setting up a double-tunnel VPN, where you VPN out to a system elsewhere, and then use that system to VPN back into your office test box. If the encryption is set up correctly, everything remains highly secure, you're still acting on the same system you started with, but you've
injected a round-trip delay and some potential network problems.

I've not actually done this, but I think one could write a iptables (firewall / routing) rule under Linux that would check an external plugin under specified circumstances, which would discard packets with a user-defined random behavior. Routing your connection through said box would then simulate various levels of lossy connection in an even more controllable manner.

In general, talk to your network people, and possibly your technical QA test folks; you of all people really ought to have the tools one way or the other to test in an environment with various combinations of delays, packet loss, and load at the push of a few buttons.

If someone hasn't created tools yet to script a Hami / ship raid sort of setting with 3-4 dozen sock puppets firing off powers so other things can be tested in that sort of environment, they really should. (This is the sort of thing that AE would be great for if it were more powerful.)

I seem to recall reporting a few regarding flight, and especially Hover. Even got modsmacked by Mod8 for posting a step-by-step description when you weren't taking PMs. I haven't checked to see if those loopholes still allow me to bypass all animations (I hope they're gone), but it seems like Hover is a big source for these things. To this day it's causing problems, like the Reveal power not playing most of its animation while Hovering. Ground animations, from what I can see, are more or less perfect, but aerial animations still still have a few surprises in them.

In other news: Dang! And I was so excited about losing draw on Broadsword. Phoey! Here's to hoping you can find some other way to achieve this, because it would have opened up my weapon users to taking non-weapon powers in a big way.

Originally Posted by Arcanaville Samuel_Tow is the only poster that makes me want to punch him in the head more often when I'm agreeing with him than when I'm disagreeing with him.

I think it's because Reveal's animation is 100% interruptable after the arm gadget being used.

WANem. You should be able to use this to simulate a wide area network's bandwidth and packet loss on your test system. It also comes as a virtual appliance so you do not need to install or configure it (you just need VMware workstation or the free VMware player to run the virtual machine).

I have my doubts that packet loss can be responsible for at least the issues I was seeing. Packet delay is theoretically possible, but seems unlikely to me. In any case, as long as your test system is fast enough, you should be able to use WANem as a virtual router between the client and server in your test environment with just some minor IP routing magic.

I don't know if it's worth the trouble to set up, but if any of your systems people are unix geeks, here's an idea that might be worth tossing their direction to see if it's a doable.

Like many systems, FreeBSD can bet set up as a router/firewall. The interesting part is that its ipfw firewall has a feature called DummyNet. With that, you can limit the bandwidth, add a delay to create latency, and even tell it to randomly drop packets with a specified probability.

I believe Linux has similar capabilities with something called netem (Network Emulator), though I don't have personal experience with it.

So if you have somebody there who has some experience with one of these systems, you could throw one of them on an old obsolete computer and put it between your in-house test client and the rest of the network. That way you could test things that are only reproducible on bad connections.

As I said, I don't know if you guys want to take the time and effort to set something like this up, but it is a possibility if you really want to track problems like this down

EDIT: Looks like Arcanaville beat me to it, with a premade distribution even. Running it in a VM is certainly a novel idea, though the local network setup might be a little complicated to ensure you're routing through it.

You can add static routes using it as a gateway even for other local hosts, and it'll work (I'm assuming ICMP redirect is disabled on the appliance).

I.e.:

C:\Windows\system32>route add 10.254.255.240 10.254.255.215
OK!

C:\Windows\system32>tracert -d 10.254.255.240

Tracing route to 10.254.255.240 over a maximum of 30 hops

1 <1 ms <1 ms <1 ms 10.254.255.215
2 <1 ms <1 ms <1 ms 10.254.255.240

Trace complete.

C:\Windows\system32>tracert -d 10.254.255.240

Tracing route to 10.254.255.240 over a maximum of 30 hops

1 <1 ms <1 ms <1 ms 10.254.255.215
2 <1 ms <1 ms <1 ms 10.254.255.240

Trace complete.

(I'm showing the trace twice for the benefit of the networking gurus out there that might be worried about getting an ICMP redirect which would override the static route with a collapsed route).


I can use the virtual appliance at 10.254.255.215 running in VMware as a router to get to a local system at 10.254.255.240. And actually, .240 is *another* virtual server running on my system; this should work regardless of whether the target is a virtual or real system connected to the local network, or more router hops away.

The one thing this won't work for directly is if you have the client and server literally loaded on the same workstation, and that computer has only one IP address. You might need to play some very weird networking games to make something like this work. It might make more sense to simply virtualize the server side under VMware to separate the two machines, unless debugging actually requires both client and server to be loaded locally under the same system context.


BaB: if you want to pursue this troubleshooting angle in more detail or your LAN techs have questions, you can just email me.

Why not build a _really_ noisy wifi connection? Say, find a 2.4 ghz mobile phone and put it right next to the wifi till you get a minimal status?

Would a secondary draw+attack (in one move) animation that is the same duration as the pure attack animation solve the problem?