Jump to content


WFG Programming Team
  • Posts

  • Joined

  • Last visited

  • Days Won


Everything posted by Dunedan

  1. Here are two additional screenshot made from the "wall demo" scenario and the "oasis" map. Both should have the same heights for the CCs if I'm not mistaken. For me that still looks like a different camera angle as well. I'm also curious how a different zoom level can even happen when the only difference between the configuration of the two instances are the lobby related settings. Could that be related to windowed mode? I've the feeling this difference goes away once I switch to full screen.
  2. It happens every time I start a game. As I usually do that with "unknown" map I'm pretty sure it's not related to the height of the CCs. I also haven't changed anything configuration related.
  3. I just built the latest SVN and noticed that somehow the angle of viewing at the map is different. That change must have committed during the past few days and I'm not sure if it's a bug or intentional. The following screenshot is from the same multiplayer game, right after starting it with two 0ad instances in parallel. You'll notice the perspective of looking at the CC is different for the two players.
  4. Its the 34th Chaos Communication Congress (https://events.ccc.de/congress/2017/wiki/index.php/Main_Page), a yearly gathering of hackers organized by the Chaos Computer Club (https://www.ccc.de/).
  5. > here's a function to update ratings based on the Glicko system and using scythe's same signature. I guess the motivation is to have a better algorithm than ELO? Is there any previous discussion on this topic? > Since inactive time and rating deviation are not stored, I used the games played to compute the deviation. Aren't you neglecting the major advantages of Glicko compared to ELO by doing that?
  6. A first patch to let 0ad talk directly to EcheLOn is now available for review: https://code.wildfiregames.com/D1177
  7. While SQlite claims they're able to do that (https://sqlite.org/faq.html#q5) I did did a quick test to verify that it works for EcheLOn and edited the sqlite-database file using sqlite3 while EcheLOn was running and it worked perfectly well. Suddenly my test account had a a rating of 5000. ;-)
  8. I seriously doubt that the GIL is the problem here, but I'd be pleased if you prove me wrong. That's another possible solution. Would work for me as well, as the problems I currently see are based around the relay functionality. No question, that should be supported in any case. How does that relate to restarts of EcheLOn? Isn't that simply covered by removing the accounts from the database?
  9. Yes, that's an advantage, but is there any other reason for that than splitting the load onto two servers? Does it happen that EcheLOn needs to be restarted, while XpartaMuPP doesn't? If yes, I'd be interested to know why.
  10. As some of you might know I started looking into the lobby bots (EcheLOn and XpartaMuPP) a while ago with the goal to improve on some of their shortcomings. While I have nothing to share yet, I already cleaned up the code base and made it work with current versions of its dependencies and ejabberd (including mod_ipstamp), amongst several smaller improvements in various areas. There is one detail that keeps me thinking: Is there really a need for two separate bots? If my understanding of the original intention (see #3022 [PATCH] Split rating from lobby bot into another ) is correct, it was done for two reasons: 1. Allow to run the lobby without ratings functionality. 2. Remove load caused by the rating logic from the bot who's responsible to manage games. While these goals are noble, I'm not a friend of their current implementation. As currently implemented 0ad sends all lobby-related requests to XpartaMuPP. When XpartaMuPP realizes that it's something concerning ratings (rate a new game, fetch ratings to show to players, ...) it'll relay this request to EcheLOn. That relay-functionality causes a lot of complexity, makes it unnecessarily complex to fix issues like #4812 XpartaMuPP doesn't use original id for responses and causes load on XpartaMuPP which wouldn't be necessary. What I like to propose is to merge the two bots again in future and implement the following measures to fix the original issues: 1. Add a command line flag to enable/disable rating functionality. 2. Make more use of threads to distribute the load better. What are your thoughts about that?
  11. Have you tried recompiling gloox as suggested a while back when you had a similar problem? https://trac.wildfiregames.com/ticket/4605
  12. Relevant changeset: SVN20321. I'm afraid I can't really help here, as I'm neither really familiar with C++ nor do I have Windows available for testing. If there's anything I can help with I'm glad to do so. Just a shot, but can you check what happens if you apply the patch below? diff --git a/source/lobby/glooxwrapper/glooxwrapper.h b/source/lobby/glooxwrapper/glooxwrapper.h index 6b2f33a374..b6afe2bd6d 100644 --- a/source/lobby/glooxwrapper/glooxwrapper.h +++ b/source/lobby/glooxwrapper/glooxwrapper.h @@ -394,7 +394,7 @@ namespace glooxwrapper bool connect(bool block = true); gloox::ConnectionError recv(int timeout = -1); - std::string getID(); + const std::string getID(); void send(const IQ& iq); void setTls(gloox::TLSPolicy tls);
  13. Ok, that was to early. The problem appeared back again, but this time I noticed that it's not related to 0ad at all. Apparently that problem appears whenever compositing for my DE gets enabled or disabled. As running 0ad disables it, it looked like it is related to 0ad, while it is not. For somebody running with KDE it should be fairly simply to check: Start Firefox and open some page, switch compositing with CTRL + Shift + F12 and you should see blank pages in Firefox.
  14. Just wanted to point out that this problem went away for me without touching 0ad. So I believe some updated Debian package fixed it.
  • Create New...