Spring RTS Engine

I know I'm going to get flamed, but...

The frequency of sync errors in the last two versions seems to be the worst since the .5 series of Spring - you know, back when you considered yourself very lucky if everyone got into battle on the first try, and it was a verifiable miracle if everyone stayed in battle. In the last four battles, three have had sync errors for multiple people in the battle after 10-15 minutes of gameplay.

While new features are nice, I think the Spring community would appreciate it more if the next version concentrated more on stability of the existing game than on new features.

I agree with Felix, having shared one such game with him.

Sync errors are made of cake and win!

Means we can all be winners when we play a game together, in our own little desynced bliss.

I made something like this, but it was locked by a nameless mod who thinks complaining about spring is blasphemy or something.
Its true though, and no amount of locking will change it.

Are you somehow implying that one or more moderators here misuse their powers!?

How dare you!

LOrDo wrote:I made something like this, but it was locked by a nameless mod who thinks complaining about spring is blasphemy or something.
Its true though, and no amount of locking will change it.

O_o

It would help if people would try to find something that people that sync out have in common (a certain building/unit has been built, a shield has been used, a nuke has been launched, people from a certain country / using a certain win version always desync, etc.). I've been playing quite some games lately and none of them has desynced. Neither has any of my ~200 testruns, except the ones which were expected to desync (3.4 vs not 3.4 in the matrix).

I had several desyncs so far. Some seemed to be related with map drawing and other with chain explosions (chain exploding wind plants or parts of base resulted in several people having desync from the rest).

Are there any replays available of games that desync?

So that either Yeha can run them through purify or I can run them through valgrind? (memory checking tools)

Is sync error == de-sync? because sometimes i can be getting the errors and still see the person im supposedly de-synced from microing units (dgunning too)
Ill upload a game or 2 of error spammed games (i have quite a few )

http://www.unknown-files.net/index.php? ... &dlid=1982
found a good one :)
map : River Dale-v01
mod : BalancedAnnihilationV41
it de-syncs the exact moment that a peewee gets killed (teh awesome raid tbo :)) just before the 6 minute mark near the bottom right corner

hope this helps!

http://licho.eu/desync1.zip

short 1v1 game with desync

http://licho.eu/desync2.zip

This one desyncs almost instantly.

I also had 2 more desync yesterday but cannot find demos.

If desync replays will help, I can get a bunch of .74 and/or .73 replays emailed or FTPd somewhere. Can you setup or designate some sort of central clearinghouse for these?

One of the most worrying trends is that they seem to "break" almost half the players in a game when they occur now, not just one player. I've played games with more than 6 players where 4 or 5 people will suddenly desync. In some instances it'll be everyone but the host, in some it'll just be random but we will lose two or three. It does seem to happen more often when there's some heavy action going on, usually when bases are getting wrecked.

Yeah. I remember one time I played a 2x2 on speed metal, and with the strengh of the 2 sides balanced the power and units of both begun to grow like mad. It was soo much that my computer couldnt stand this well and the game became lagly for me. After some time on this the game desynced.

I've experienced many desync errors while using vulcans/buzz (in AA)
(The desync occurs when I (or any players) start firing with such units)

I played a couple of games last night and one of them desynced at the very end. It started as XY desync, which I'm guessing is the position of units. At the time I had a lot of units that were attacking (and getting stuck on solars). I wonder if the cause of the desync was that on one machine the unit got stuck, but on the other it did not. Maybe collison detection is the cause?

If it is, we can reprogram units so they wait until the path is clear instead of stomping on things until they are out of the way.

If it is, it would probably have shown up in my tests, as units getting stuck happens quite often.

Let's go back to the topic. Is it the path finder the problem, or it is just a potential culprit for the sync errors ?

Because I could live with it as it is no, if there would be no sync erros from it for sure.

As the code is compiled now with gcc/migw so we can always blame the compiler... (in 4.0 there were major changes implemented in gcc so the gcc code base is not tested for such a long time in real life).

Anyway is any sync error reproducible ? (so that a debug can be attempted?)