So, another abortive attempt to change to the new engine (version 96 this time) has come and pass. It was deemed to buggy and we go back to 91 again. What are we going to do about this? Staying on 91 is unsustainable - it's got some notable bugs that were fixed in later versions, like the vehicles getting stuck on hills and whatever they fixed to make nanoparticles and weapon effects render properly on my ATI HD6800. Linux support for 91 is also shaky - there were no official static builds (those started on 94), and while some people have been able to get 91 to work (using old LTS ubuntu releases, I think), but I havent been able to find a 91 binary (compiled or found elsewhere) that doesnt segfault on Arch (bleeding edge Linux) But going forward to the latest versions just doesn't work. Too many bugs. Why does this keep happening? Zero-K is, by my cursory inspection, the biggest consumer of Spring engine by players, are the Spring devs just unaware of the problems their engine has with us? Or are these problems unique to ZK, and if so, how can we fix them? In short, where do we go from here? I raised the posibility of forking 91 and backporting the fixes to there, to make a ZK engine based on 91. I would Kickstart that, but this was shot down as too dificult. What should ZK do about our engine problem?
+8 / -3
|
quote:
I raised the posibility of forking 91 and backporting the fixes to there, to make a ZK engine based on 91. |
Instead of forking, you can just fix bug in spring engine. Isn't is better?
+1 / -0
|
how is it unsustainable? its perfectly sustainable, it just dosn't improve over time.
+1 / -0
|
quote:
Instead of forking, you can just fix bug in spring engine. Isn't is better? |
The bugs are already fixed. But if you want to have those fixes, you gotta buy the entire package of new bugs and quirks, too. His point was avoiding the new bugs by just applying the (known) patches to 91 bugs, which can only be accomplished by forking. Still, that is also unsustainable. Do ZK and engine devs sometimes converse over where they are headed, what their goals are and what needs to be done? I have no idea what is going on behind the scenes, but if people say the new engine versions just introduce features nobody wants, annoying bugs and (of course) some long needed fixes, cooperation doesn't seem to be going so well?
+1 / -0
|
quote:
Do ZK and engine devs sometimes converse over where they are headed, what their goals are and what needs to be done?
|
This. It seems like the engine devs arent even aware of the problems we experience.
+1 / -3
|
For some whacked reasoning but reasoning nevertheless, perhaps the problems are considered a lower priority because they know that ZK hosts can just roll back to 91 again.
+0 / -0
|
Thing i do not like is that every new version the performance get worse too.
+1 / -0
|
While I can't really blame the current engine devs because they work for free, I miss the old days when we can expecting solid stable versions and game changing new features.
+0 / -0
|
I think Spring dev is experimenting with stuff and push it into code repository. Each time they introduce new changes they created a new bug, so player have to constantly try new engine to make sure that existing feature not broken. Its like: if you didn't catch new bug, spring dev will never fix it. For example, lets say nobody complained about the slow pathcache loading: I expect Spring dev would never fix it, instead move on to other stuff and so on and so on.
+0 / -0
|
I agree with @[DOOP]Fortaleza here, there's a tendency of engines losing performance over time. I have 220 some fps at start on 91 (an abusrd ammount) and usually end with about 40-50. With latest... 110 at start (80 with 95) and ends with 15. So you can do the math here - ~50% reduction. OP: Why are you so eager to switch engines? Just to fix a few bugs but add a ton of new ones? Edit: OP try to disable Safe Mode. That'd cause your nanoparticals to render properly.
+0 / -0
|
The bottom line is that nobody cares. Usually you could get away with using a different engine in the 1v1 room and people wont mind, but in the 10v10 room and others people are stupid. None of them care about the problems other people have with 91, and they certainly don't care about testing the new engine versions. The people who complain about performance are blaming the engine for the problems associated with ridiculously huge teams. If anything, this is what will kill ZK. /rant
+0 / -3
|
Here's a suggestion: JFDI. Fix to 96 and force the issue, because it's getting silly. Since this all seems anecdotal anyway, let me add mine: 96 is far superior in performance for me (91 is basically unplayable) and I've seen very few bugs aside from some minor pathing issues. The longer this takes, the harder it gets, the more resistance there will be, the closer ZK comes to becoming an largely abandoned project haunted by hold-outs.
+5 / -0
|
for me, 96 is much faster than 91, but both are playable. (core2quad-2,4, nvidia 560)
+0 / -0
|
the main problem appears to be the long loading times and the fact that zk games arent adapated to that - on any map larger than about 16x16 most people will not be finished loading at game start, which essentially means you just cannot have a proper game on 50% of the maps that are usually played. i imagine if that problem was solved in one way or another acceptance for the new engine would rise dramatically.
+1 / -0
|
The loading time is fixable, commits have been made to both ZKL and Spring to address that. The big whammy is a Springie issue that causes it to fail to save a replay, causing most games sice 92.0 to not be uploaded. Competing autohost technology called Spads, however, doesn't display such behaviour.
+0 / -0
|
  yanom: I use weblobby with arch Linux and it is working fine with both 0.91 and 0.96. (don't remember how I got 0.91, but it either auto downloaded or there was some link there. Anyhow, not compiling it) Regarding loading times, me personally I could not reproduce it when I did a simple test (data here: http://zero-k.info/Forum/Thread/6538), I have seen in the bug report that some maps have problems with the features (http://springrts.com/mantis/view.php?id=4248). Regarding the developers, maybe someone should donate them a computer with ATI. I have NVIDIA and I generally have none of the problems that people with ATI complain about, so if the devs also have NVIDIA I guess they can not do much... Also, do not forget 0.91 desyncs. 0.96 did not desync once for me, and immediately after the release people were playing both 0.91 and 0.96 so the number of desyncs could be compared.
+0 / -0
|
same problem as yanom with nano particles and weapon effects. I got a new videocard. Every single game I own runs like 30- 50 % better. Only zerok manages to get worse performance.
+0 / -0
|
Biggest problem for zk was that it is stuck in 91 and while development goes on there is many changes in 96 and longer zk stays on 91 harder it will be switch to 96. Before you just had to switch no matter what. This problem actually is rather easily fixable, mostly these new bugs have fixes in git so all it is needed is cherry picked engine updates like 96.1 96.2 and so on. Since static version compiling is implemented on both linux and windows I dont know why this isint done, its obvious that you cant release 96.0 then add new features and fix old bugs without adding new ones and release 97.0 this model is simply doomed fail on never ending search quest for truly "stable" release.
+1 / -0
|
Every single time new engine was being released, ZK was first to test and first to switch. And since 91.0 every single time this happened we reverted due to bugs/performance issues. Simply put, ZK will switch when overal benefits of new engine exceed its overal drawbacks. 96.0 brings many unresolved issues, namely: - significantly worse performance in common scenarios - crashing on some ATI machines - rendering issues on ATI machines (fullscreen flicker) - rendering issues with minimap commands - demos not being saved by dedicated server - bigger maps take forever to load Some of these issues can be fixedin ZK, some need engine patch and most need further investigation. Atm there is no reason to switch given only drawbacks of 91.0 are: - desync in very lage games - linux static build is problematic Also as a minimum, new engine should run faster on common configurations. It's unacceptable to have further slowdown without extra features. Even 91.0 is significantly slower than even older 0.xx builds of engine and has subjectively worse unit pathing and handling.
+6 / -1
|
quote:
While I can't really blame the current engine devs because they work for free, I miss the old days when we can expecting solid stable versions and game changing new features. |
To be honest, there never really were solid stable versions to begin with. The playerbase just figured out the bugs and adapted to them until they were fixed.
+3 / -0
|
