Zandronum 3.0 is a significant downgrade from 2.1.2

General discussion of the port and Doom-related chat.
User avatar
Torr Samaho
Lead Developer
Posts: 1543
Joined: Fri May 25, 2012 6:03 pm
Location: Germany

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#21

Post by Torr Samaho » Wed Oct 25, 2017 6:13 am

Decay wrote:
Wed Oct 25, 2017 1:11 am
After trying out the previous 3.0 beta debug version and turning sv_useticbuffer off, I felt there was a -MASSIVE- improvement. Granted, I had 18-20 ping, but every single shot that should've hit, actually hit. Felt like LAN.
So "sv_useticbuffer 0" seems to fix the issue for you and the setting of cl_ping_unlagged has no effect? It would be no problem to bring sv_useticbuffer back to get some larger scale testing.
jdagenet wrote:
Wed Oct 25, 2017 3:47 am
though I’m more curious as to why the tic buffer was rewritten anyway and not thoughly tested.
We gave everybody countless of chances to test, we even almost literally begged people to test. If you are ignoring our many calls for testing, you can hardly blame us.

User avatar
Galactus
Forum Regular
Posts: 351
Joined: Sat Nov 03, 2012 9:22 am
Location: Belgium
Contact:

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#22

Post by Galactus » Wed Oct 25, 2017 12:21 pm

Asking the server cluster owners to host beta servers might've fixed that issue. Competitive people tend to play on some set clusters and TSPG apparently wasn't one of them.

User avatar
Ivan
Addicted to Zandronum
Posts: 2219
Joined: Mon Jun 04, 2012 5:38 pm
Location: Omnipresent

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#23

Post by Ivan » Wed Oct 25, 2017 5:27 pm

Torr Samaho wrote:
Wed Oct 25, 2017 6:13 am
Decay wrote:
Wed Oct 25, 2017 1:11 am
After trying out the previous 3.0 beta debug version and turning sv_useticbuffer off, I felt there was a -MASSIVE- improvement. Granted, I had 18-20 ping, but every single shot that should've hit, actually hit. Felt like LAN.
So "sv_useticbuffer 0" seems to fix the issue for you and the setting of cl_ping_unlagged has no effect? It would be no problem to bring sv_useticbuffer back to get some larger scale testing.
jdagenet wrote:
Wed Oct 25, 2017 3:47 am
though I’m more curious as to why the tic buffer was rewritten anyway and not thoughly tested.
We gave everybody countless of chances to test, we even almost literally begged people to test. If you are ignoring our many calls for testing, you can hardly blame us.
I noticed I forgot to state the unlagged I used in my post. I used ping based because gametic wasn't really helping me, players were teleporting too much. (I pinged 130-150 to the server)
=== RAGNAROK DM ON ... uh... dead forever? ===
=== ALWAYS BET ON ... uh... dead forever? ===
=== Who wanta sum wang? ===
=== Death and Decay - A new Monster/Weapon replacer ===

User avatar
Torr Samaho
Lead Developer
Posts: 1543
Joined: Fri May 25, 2012 6:03 pm
Location: Germany

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#24

Post by Torr Samaho » Wed Oct 25, 2017 9:12 pm

Galactus wrote:
Wed Oct 25, 2017 12:21 pm
Asking the server cluster owners to host beta servers might've fixed that issue. Competitive people tend to play on some set clusters and TSPG apparently wasn't one of them.
If competitive people would have wanted to test, they could have asked for test servers.
Decay wrote:
Wed Oct 25, 2017 2:15 pm
I did not test the ping-based unlagged - I was never a fan of it to begin with, and I don't know if it would have any effect on me anyway, with 18-20 ping in the server. According to this tracker ticket https://zandronum.com/tracker/view.php?id=2859 gametic unlagged was desync'ing even locally;
The ticket says that the issue already happens in 2.1, that's why I didn't look further into the issue for 3.0. Are you saying that the ticket info is wrong and 3.0 is affected more than 2.x?
Decay wrote:
Wed Oct 25, 2017 2:15 pm
and by using sv_useticbuffer 0, I experienced 0 problems in terms of actually hitting somebody, and it also fixed the collision problems I mentioned earlier. However, there was the problem of people occasionally being jittery (both players at the same time) visually, and 2 people froze and lagged out of the server for no seeming reason.

I feel like just having sv_useticbuffer 0 is a bandaid to the problem, which seems to be a faulty ticbuffer.
We didn't make any (explicit) changes to the tic buffer in 3.0 compared to 2.1.2, so in principle, the buffer should work the same in both versions. Did the 2.1.2 servers you compared 3.0 to actually had the tic buffer enabled?

User avatar
AlexMax
Forum Regular
Posts: 244
Joined: Tue May 29, 2012 10:14 pm
Contact:

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#25

Post by AlexMax » Wed Oct 25, 2017 10:59 pm

I probably should have put up 3.0 servers, so part of this is on me. Torr isn't wrong though - organizing testing is like pulling teeth.

I've always considered netcode problems to be insidious. I'm inherently mistrustful of netcode problem reports because so much of it is based on perception and there are so many variables at play that it's hard to pinpoint actual systemic problems versus unavoidable mispredictions due to the realities of networking. Not that I'm trying to put the people who report problems down, but I hope people realize that getting netcode right is incredibly difficult - doubly so since Doom is pretty much one gigantic "worst case scenario" of netcode wrapped into one - triply so since this port is programmed by hobbyists in their spare time.

That said, Shooter's video seems problematic. Considering the angle of the shot and where the player was previously, that sets off alarm bells in my head.

It doesn't sound like the actual netcode changed a whole bunch between 2.1.2 and 3.0, just the underlying version of ZDoom, and I'm having a hard time trying to figure out what ZDoom could have possibly changed to have an effect on client/server play, especially since the parent port has no concept of client/server itself. Heck, it might not even be the netcode itself per se - maybe there's more traffic going back and forth, and that might be causing knock-on side effects in terms of delays and packetloss.

It would really help if Zandronum gave the player more data to work with, kind of like a netgraph from Half Life 2 or Quake 3. I know there is one already, but it's so unknown that I bet most of the people in this thread don't realize it even exists. I don't even remember the command to enable it, or remember what it contains. At the very least, it should have something that measures the incoming and outgoing bandwidth, number of packets dropped, and how many tics behind the server the player is.
Theshooter7 wrote:
Sun Oct 22, 2017 10:25 pm
Watching the spectator perspective of someone with high ping (like 110+) is truly a delight as you'll often see them shoot at thin air and people who were behind them or off to the side or something will take damage/die.
Keep in mind that spectating someone is a completely misleading way to judge unlagged. In spectator mode, you are watching the game from the point of view of the server, which is backwards-reconciling players in order to judge hits and misses, so things are going to look out of whack. If you could actually spectate what the server is ACTUALLY doing behind the scenes, you'd see players jump backwards in time every time somebody fired a shot.

The only POV you can trust is your own (or somebody's demo recording of their own POV).
The only limit to my freedom is the inevitable closure of the
universe, as inevitable as your own last breath. And yet,
there remains time to create, to create, and escape.

Escape will make me God.

Theshooter7
Forum Regular
Posts: 262
Joined: Wed Jun 06, 2012 2:15 am

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#26

Post by Theshooter7 » Wed Oct 25, 2017 11:40 pm

AlexMax wrote:
Wed Oct 25, 2017 10:59 pm
Keep in mind that spectating someone is a completely misleading way to judge unlagged. In spectator mode, you are watching the game from the point of view of the server, which is backwards-reconciling players in order to judge hits and misses, so things are going to look out of whack. If you could actually spectate what the server is ACTUALLY doing behind the scenes, you'd see players jump backwards in time every time somebody fired a shot.

The only POV you can trust is your own (or somebody's demo recording of their own POV).
I figure this to be the case, although it simply seems considerably worse or at least more noticeable since 3.0's arrival. Alas, it's probably one of the lowest priority concerns either way.

On the note of testing, with the tic buffer off in the server and cl_ping_unlagged turned on, gameplay felt way better. I was able to land shots I know for a fact would normally never work under current 3.0 w/ tic buffer (an example was a player running past a narrow entry from one side to the other with SR50, and being able to simply hit him with a few SSG pellets just from a 'reaction' shot as I could not predict his movement for timing. This would never have worked normally).

However, with tic buffer off and cl_ping_unlagged set to 0, players were jittery and impossible to hit at all times. Playing on Shoot, it was impossible to predict the opponent's movement in the slightest.

User avatar
Razgriz
Forum Staff
Posts: 713
Joined: Mon Jun 04, 2012 5:15 am

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#27

Post by Razgriz » Wed Oct 25, 2017 11:43 pm

AlexMax wrote:
Wed Oct 25, 2017 10:59 pm
It would really help if Zandronum gave the player more data to work with, kind of like a netgraph from Half Life 2 or Quake 3. I know there is one already, but it's so unknown that I bet most of the people in this thread don't realize it even exists. I don't even remember the command to enable it, or remember what it contains. At the very least, it should have something that measures the incoming and outgoing bandwidth, number of packets dropped, and how many tics behind the server the player is.
You're thinking of Zdaemon's "netstats" command which does that aside from the last thing stated. I don't think Zan has something like that explicitly, as I checked before but didn't find anything.

User avatar
AlexMax
Forum Regular
Posts: 244
Joined: Tue May 29, 2012 10:14 pm
Contact:

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#28

Post by AlexMax » Thu Oct 26, 2017 12:20 am

What would be really nice would be Serverside demos. Having an accurate view of what exactly the server is seeing at any given time would probably be good for trying to debug misprediction problems like these, especially since in theory you could also inject debug information into the demo stream that only the server would know, like how far back a player would be reconciled on hit.

If implemented, I would definitely enable them on my primary servers.

It's also something that - failing any specific leads - would probably be the best use of time in tracking down netcode problems.
The only limit to my freedom is the inevitable closure of the
universe, as inevitable as your own last breath. And yet,
there remains time to create, to create, and escape.

Escape will make me God.

User avatar
ibm5155
Addicted to Zandronum
Posts: 1641
Joined: Tue Jun 05, 2012 9:32 pm
Location: Somewhere, over the rainbow

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#29

Post by ibm5155 » Thu Oct 26, 2017 5:40 pm

could a wad mod help testing this kind of lag?
like showing a ghost where every player is in the server side coord, showing clientside bullets to compare to what the server has registered...

Unfortunately I couldn't test this kind of problem since I'm used to a 170 - 230 ping :(
Projects
Cursed Maze: DONE, V2.0
Zombie Horde - ZM09 map update: [3/15/13]
Need help with English? Then you've come to the right place!

<this post is proof of "Decline">

User avatar
Torr Samaho
Lead Developer
Posts: 1543
Joined: Fri May 25, 2012 6:03 pm
Location: Germany

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#30

Post by Torr Samaho » Thu Oct 26, 2017 7:20 pm

Decay wrote:
Wed Oct 25, 2017 10:48 pm
To be clear, we tested on the last 3.0 beta debug version (it was all TSPG had made available for us). Evidently we need TSPG to have 2.1.2 available for servers as well for further testing. Unless it was off by default (and I don't think it was),
The tic buffer was on by default in 2.1.2.
Decay wrote:
Wed Oct 25, 2017 10:48 pm
there is a dramatic difference between 2.1.2 and 3.0, thus the thread title. However, 2.1.2 still exhibited problems with the unlagged, ie being hit way behind walls, however it was not to same extent.
Before drawing any further conclusions, we need side by side testing on 3.0 and 2.1.2 with tic buffer on the same host with the same settings for both versions to confirm that the buffer leads to problems under 3.0, which didn't occur in 2.1.2.

Anybody willing to host?
Decay wrote:
Wed Oct 25, 2017 10:48 pm
Re: Servers
IIRC some people did ask AlexMax to host 3.0 on NJ but he opted not to. If the server is not NJ, the "competitive" people don't care about it and won't play it. And to be perfectly fair, its not up to them to test it.
Without proper feedback from the competitive players, how are we supposed to make sure that a new version caters to the need of the competitive players?
Decay wrote:
Wed Oct 25, 2017 10:48 pm
Wartorn speculated there might've been a change in zdoom physics that may have caused this discrepancy. Between 2.1.2 and 3.0 were there any major changes to the game physics?
I don't recall any physics changes that would lead to the observed behavior.
AlexMax wrote:
Wed Oct 25, 2017 10:59 pm
I've always considered netcode problems to be insidious. I'm inherently mistrustful of netcode problem reports because so much of it is based on perception and there are so many variables at play that it's hard to pinpoint actual systemic problems versus unavoidable mispredictions due to the realities of networking. Not that I'm trying to put the people who report problems down, but I hope people realize that getting netcode right is incredibly difficult - doubly so since Doom is pretty much one gigantic "worst case scenario" of netcode wrapped into one - triply so since this port is programmed by hobbyists in their spare time.
I coudln't agree more.
AlexMax wrote:
Wed Oct 25, 2017 10:59 pm
That said, Shooter's video seems problematic. Considering the angle of the shot and where the player was previously, that sets off alarm bells in my head.
Yeah, that kind of thing shouldn't happen.
AlexMax wrote:
Wed Oct 25, 2017 10:59 pm
It would really help if Zandronum gave the player more data to work with, kind of like a netgraph from Half Life 2 or Quake 3. I know there is one already, but it's so unknown that I bet most of the people in this thread don't realize it even exists. I don't even remember the command to enable it, or remember what it contains. At the very least, it should have something that measures the incoming and outgoing bandwidth, number of packets dropped, and how many tics behind the server the player is.
We do have "stat nettraffic", which contains some information about incoming and outgoing traffic. I can extend this as needed. Number of dropped packets sounds reasonable, also tics behind the server should be doable.

User avatar
AlexMax
Forum Regular
Posts: 244
Joined: Tue May 29, 2012 10:14 pm
Contact:

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#31

Post by AlexMax » Thu Oct 26, 2017 11:18 pm

Torr Samaho wrote:
Thu Oct 26, 2017 7:20 pm
Anybody willing to host?
I suppose I can make up for not hosting 3.0 earlier by hosting 2.1.2 now. It'll be kind of a pain in the neck, since the dmflags changed between versions, but I think I can figure it out.
Torr Samaho wrote:
Thu Oct 26, 2017 7:20 pm
We do have "stat nettraffic", which contains some information about incoming and outgoing traffic. I can extend this as needed. Number of dropped packets sounds reasonable, also tics behind the server should be doable.
Okay, that's the stat thing I remembered. Unfortunately, there are a few shortcomings with it...

1. I'm not really sure what the numbers mean. I guess that the first number of In and Out are the instantaneous traffic going in and out in bytes, but that leads me to...
2. It updates very quickly...too quickly. I can't get a good feel for the first number at all because it keeps oscillating.
3. It's in a very small font in an unconfigurable corner of the screen.
4. It disappears when I quit the game. Most netgraphs in most games I've seen remain on forever until they're specifically turned off.

stat is kind of handy as a dumping ground of stat functions, but for something as important as a netgraph I'm not sure it's the best choice.
The only limit to my freedom is the inevitable closure of the
universe, as inevitable as your own last breath. And yet,
there remains time to create, to create, and escape.

Escape will make me God.

User avatar
Razgriz
Forum Staff
Posts: 713
Joined: Mon Jun 04, 2012 5:15 am

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#32

Post by Razgriz » Thu Oct 26, 2017 11:41 pm

Torr Samaho wrote:
Wed Oct 25, 2017 6:13 am
We gave everybody countless of chances to test, we even almost literally begged people to test. If you are ignoring our many calls for testing, you can hardly blame us.
I'm more surprised at the fact something that wasn't tested thoroughly was still allowed the green light, sounds like a fault on both ends

User avatar
AlexMax
Forum Regular
Posts: 244
Joined: Tue May 29, 2012 10:14 pm
Contact:

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#33

Post by AlexMax » Fri Oct 27, 2017 12:57 am

NJ FUNCRUSHER is now running 2.1.2 on port 10688 for testing purposes.

It's a copy of my Duel40 server #2, so hopefully this will cut down on chaos and make picking out problems easier. Unfortunately, Doomseeker doesn't make joining it easy. I recommend using a standalone launcher like ZDL or Rocket Launcher to set up the port, then the WADS (duel40a and newtextcolours 260), then just +connect nj.funcrusher.net:10688. Or...just reconfigure Doomseeker to point to 2.1.2 I guess?
Last edited by AlexMax on Fri Oct 27, 2017 1:10 am, edited 1 time in total.
The only limit to my freedom is the inevitable closure of the
universe, as inevitable as your own last breath. And yet,
there remains time to create, to create, and escape.

Escape will make me God.

Konda
Forum Regular
Posts: 487
Joined: Thu Jun 07, 2012 5:22 pm

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#34

Post by Konda » Fri Oct 27, 2017 12:59 am

Razgriz wrote:
Thu Oct 26, 2017 11:41 pm
I'm more surprised at the fact something that wasn't tested thoroughly was still allowed the green light, sounds like a fault on both ends
Please no. Zandronum releases are already being stalled for too long. If no one's gonna test the betas, there's no point in stalling it even further. Although this discussion may conclude with beta builds being given more attention in the future, if the testing requirements were even more strict in the past than they were, I doubt we'd still see a Zan 3.0 release.

User avatar
Razgriz
Forum Staff
Posts: 713
Joined: Mon Jun 04, 2012 5:15 am

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#35

Post by Razgriz » Fri Oct 27, 2017 1:04 am

Konda wrote:
Fri Oct 27, 2017 12:59 am
Please no. Zandronum releases are already being stalled for too long. If no one's gonna test the betas, there's no point in stalling it even further. Although this discussion may conclude with beta builds being given more attention in the future, if the testing requirements were even more strict in the past than they were, I doubt we'd still see a Zan 3.0 release.
I don't see why it could have been reverted if lack of testing was the problem, if other tickets could be put on hold for 3.1 I'm sure this change could have been delayed for the better.

User avatar
Matiu
Posts a lot
Posts: 754
Joined: Thu May 29, 2014 7:56 pm
Location: Santiago De Chile
Clan: FS (ZD)
Clan Tag: Your Pal
Contact:

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#36

Post by Matiu » Fri Oct 27, 2017 1:51 am

Razgriz wrote:
Fri Oct 27, 2017 1:04 am
Konda wrote:
Fri Oct 27, 2017 12:59 am
Please no. Zandronum releases are already being stalled for too long. If no one's gonna test the betas, there's no point in stalling it even further. Although this discussion may conclude with beta builds being given more attention in the future, if the testing requirements were even more strict in the past than they were, I doubt we'd still see a Zan 3.0 release.
I don't see why it could have been reverted if lack of testing was the problem, if other tickets could be put on hold for 3.1 I'm sure this change could have been delayed for the better.
it should make it more better then the things if a lot of updates and tickets are up for 3.1

User avatar
Ru5tK1ng
Frequent Poster Miles card holder
Posts: 794
Joined: Fri Jun 01, 2012 9:04 pm

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#37

Post by Ru5tK1ng » Fri Oct 27, 2017 1:54 am

Incorrect. This tic buffer was not added for 3.0, it's been in since 2.0. It caused no issues during the life span of 2.x and there was no indication aside from unknownna's ticket that it was this big of an issue. The buffer wasn't touched directly by any commits either.

User avatar
Combinebobnt
Retired Staff / Community Team Member
Posts: 1893
Joined: Mon Jun 04, 2012 3:37 am
Location: Erth
Contact:

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#38

Post by Combinebobnt » Sat Oct 28, 2017 2:06 am

OK i see like 6 new tickets on the gametic unlagged so I have to say:

Good job guys, you finally did it.

Mobius
Banned
Posts: 815
Joined: Wed Jun 13, 2012 3:43 am
Location: The center of the A3's controversial and machiavellian conspiracy
Clan: Cube
Clan Tag: Aᵌ
Contact:
Banned: Permanently

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#39

Post by Mobius » Sat Oct 28, 2017 2:09 am

Combinebobnt wrote:
Sat Oct 28, 2017 2:06 am
OK i see like 6 new tickets on the gametic unlagged so I have to say:

Good job guys, you finally did it.
Image
dewsome: i can do this all day
dewsome: do you think you're somehow special? i spent years arguing with nostar, jenova, mobius
<+Thomas13> Mobius u r inferior, go outside and get beaten up
Alfonzo ~ "I wonder who will hear him trash talk when the dirt closes over him?"

User avatar
Leonard
Developer
Posts: 166
Joined: Sat Mar 30, 2013 8:09 pm

Re: Zandronum 3.0 is a significant downgrade from 2.1.2

#40

Post by Leonard » Sat Oct 28, 2017 2:31 am

Except for cl_ticsperupdate, I have fixed them all here.
I'm just making the PRs one by one to keep things simple.

Post Reply