Author Topic: [Solved] Potential 1.009F Bugs  (Read 5212 times)

Offline Kalzarius

  • Sr. Member Mark II
  • ****
  • Posts: 422
[Solved] Potential 1.009F Bugs
« on: July 07, 2009, 02:12:11 am »
These are several issues I noticed while playing my most recent session of AI War.

1) The mouse wheel zooms the map in and out while reviewing graphs in the score window, even though the game insists that it has no input focus.

2) Once time reaches 1 day, there is not enough room for the full time to be displayed before it goes off the edge of the screen.

3) Some ghost entries on the scan data appear alongside their present counter parts, which constantly swap positions back and forth.  So you see the blue fighters (for example) image swapping places with a white one.  The broken entry reports none of any of the ships present in the system (and last seen -- if I recall correctly -- was also 0).  This can be fixed by leaving the system and coming back.

4) I'm not entirely sure if this was a broken save or not, but the AI was thoroughly unresponsive tonight.  After destroying a command station, there was no swarm of ships (despite the fact that there were still 800 ships in the system).  Bringing a group of attack ships towards any of the their command posts normally would have them sending some of their 600 teleport raiders to intercept my ships, yet this time they just all sat there and let me take them out with cruisers from afar.  There was a single raid, but otherwise no other offensive activity of any kind.

I won't get a chance to test anything more until next week, unfortunately, but I will report here with updates at that time.

 - Kal Zarius
« Last Edit: September 25, 2009, 09:40:53 am by Fiskbit »

Offline x4000

  • Chris McElligott Park, Arcen Founder and Lead Dev
  • Arcen Staff
  • Zenith Council Member Mark III
  • *****
  • Posts: 31,651
Re: Potential 1.009F Bugs
« Reply #1 on: July 07, 2009, 09:03:03 am »
1) The mouse wheel zooms the map in and out while reviewing graphs in the score window, even though the game insists that it has no input focus.

Yeah, windows is funny with how it focuses.  I wouldn't expect it to pass that message over to the main window when this sub-panel has focus, but sometimes it decides to.  I'll put in some extra catches to make sure that doesn't happen.

2) Once time reaches 1 day, there is not enough room for the full time to be displayed before it goes off the edge of the screen.

Uh... wow.  That's a long game. :)  I'll make the change so that it will slide over when need be.

3) Some ghost entries on the scan data appear alongside their present counter parts, which constantly swap positions back and forth.  So you see the blue fighters (for example) image swapping places with a white one.  The broken entry reports none of any of the ships present in the system (and last seen -- if I recall correctly -- was also 0).  This can be fixed by leaving the system and coming back.

The white entries are supposed to appear when a ship of that type leaves the system, so that you can still look at it for a while.  However, you should never have both blue and white entries.  I'll keep an eye out for that, and see if I can duplicate it.

4) I'm not entirely sure if this was a broken save or not, but the AI was thoroughly unresponsive tonight.  After destroying a command station, there was no swarm of ships (despite the fact that there were still 800 ships in the system).  Bringing a group of attack ships towards any of the their command posts normally would have them sending some of their 600 teleport raiders to intercept my ships, yet this time they just all sat there and let me take them out with cruisers from afar.  There was a single raid, but otherwise no other offensive activity of any kind.

This means that the AI was so bogged down that it couldn't respond.  This is probably what is causing your OOM exception, too.  The string handling thing should probably help a lot with this, same as the shots thing will, but this will also cause severe, severe lag both in the game and in the AI itself.  I haven't seen this since... oh, 1.007 or so, and with 84,000 ships in a four-player game.  The AI thread was choking on all the data it was being passed, and the count of messages-in-waiting for the AI thread was piling up and up to the point where it was at half a million or so.

The thing is, I fixed that issue, so this must be some sort of variant.  Can you please send me your current savegame from this?  It sounds like you have some sort of extreme number of ships or something, and therefore that I need to do some extra profiling and performance tuning with your specific scenario.  I haven't been able to duplicate anything like this since my changes in a past release (1.007 or whatever that was), even with 120 planet maps and eight players worth of ships, but I won't know exactly what is going on here until I see the save.  If this is some sort of strange edge case, I imagine I can do something about it fairly quickly once I see what is going on -- the extreme length of the game is probably a factor here (over one day of time?  I think that's a new record, so far as I know the longest I've heard was around 16 hours).

I won't get a chance to test anything more until next week, unfortunately, but I will report here with updates at that time.

No worries!  If you have time to send me your save before then, however, then I can probably get any needed tuning in there well before you're back.
Have ideas or bug reports for one of our games?  Mantis for Suggestions and Bug Reports. Thanks for helping to make our games better!

Offline Kalzarius

  • Sr. Member Mark II
  • ****
  • Posts: 422
Re: Potential 1.009F Bugs
« Reply #2 on: July 07, 2009, 10:50:19 pm »
Quote
Uh... wow.  That's a long game. :) I'll make the change so that it will slide over when need be.

You have no idea. :P

Quote
Can you please send me your current savegame from this?  It sounds like you have some sort of extreme number of ships or something, and therefore that I need to do some extra profiling and performance tuning with your specific scenario.

I have it attached to this post.  The map has 100 planets and I think there are a total of around 50,000 ships between the two humans and the two AIs.

Quote
I haven't been able to duplicate anything like this since my changes in a past release (1.007 or whatever that was), even with 120 planet maps and eight players worth of ships, but I won't know exactly what is going on here until I see the save.  If this is some sort of strange edge case, I imagine I can do something about it fairly quickly once I see what is going on -- the extreme length of the game is probably a factor here (over one day of time?  I think that's a new record, so far as I know the longest I've heard was around 16 hours).

Yup, just over a day.  We like to coordinate our offensives and our advanced factories tend to be a great distance from where we are attacking from, so it takes a lot of time to ferry all of our ships to the front lines.  It's been more than a real day, though.  When the game isn't running in real time, it can take several hours to get through one in-game hour.

There wasn't any stalling in 1.008, though.  We had one heck of a time with that due to all of the wormhole reinforcements.  They were very active that game.  They only stalled with the 1.009F pre-release.

 - Kal

Offline x4000

  • Chris McElligott Park, Arcen Founder and Lead Dev
  • Arcen Staff
  • Zenith Council Member Mark III
  • *****
  • Posts: 31,651
Re: Potential 1.009F Bugs
« Reply #3 on: July 08, 2009, 02:05:15 pm »
Wow, that's impressive.  I loaded up your savegame, and I see that there are 53,000 or so ships in there.  I'm actually able to run two copies of it simultaneously at double-speed on my 2.4Ghz quad, so it doesn't seem like anything is majorly wrong with it.  Sounds like it might be a system specs thing, which I hope that my last improvements will help with to a significant degree.  If it is still lagging for you, then having snapshots of the F3 debug output from both machines -- during a lag period -- would be very helpful for me to make further optimizations.  Right now it's doing so well for me that there's not a lot of optimizations I can really find (hopefully that means that the string stuff has already fixed it for you, as well).

Thanks for sending it!
Have ideas or bug reports for one of our games?  Mantis for Suggestions and Bug Reports. Thanks for helping to make our games better!

Offline Kalzarius

  • Sr. Member Mark II
  • ****
  • Posts: 422
Re: Potential 1.009F Bugs
« Reply #4 on: July 14, 2009, 01:02:14 am »
It would seem that with the latest updates, we no longer have a stalled AI.  The AI is as active as ever.

Offline x4000

  • Chris McElligott Park, Arcen Founder and Lead Dev
  • Arcen Staff
  • Zenith Council Member Mark III
  • *****
  • Posts: 31,651
Re: Potential 1.009F Bugs
« Reply #5 on: July 14, 2009, 09:59:12 am »
It would seem that with the latest updates, we no longer have a stalled AI.  The AI is as active as ever.

Wonderful!  That was relating to the string handling stuff on below-requirements CPUs, too, I think, so that's great to hear that that's working now. :)
Have ideas or bug reports for one of our games?  Mantis for Suggestions and Bug Reports. Thanks for helping to make our games better!