Specific Bug Reports

[jdog5000]

Hmmm ... okay. I thought the issue with your save was the same as with JimMac's, will check again.

 

[jdog5000]

theonlydjcat:

Your save plays fine using hotseat and just BBAI 0.81m for me, what settings exactly were you using? Does your merged mod just add BUG and keep the BBAI DLL?

Attached is a save from after the colony is formed, I ran it another 40 or 50 turns with no problems at all.

View attachment theonlydjcat AD-1230.zip

 

[TheOnlyDJCat]

Hi Jdog5000,

I could continue with hotseat too, the crash only comes with IP play.
Problem is, that playing hotseat is not really an option being in two different countries and the hot seat save is not really usable in IP play. It can be loaded but the end of turn behaviour is buggy (leads to OOS if someone clicks at the wrong moment...).

I play with BBAI directly installed/copied in program files\... and BUG mod as a mod on both computers.
If we both click on 'finish turn' in the save, it immediately crashes on both computers.

I will try your save in IP mode as soon as possible, but think if you have created it in hot seat mode, it won't be usable for IP play.

 

[JimMac99]

Jdog

Due to circumstances beyond my control, I couldn’t get everyone back to continue our MP game of Civ last Sunday, however we all met up today to continue our save with the latest version of Better A.I. (0.81m)

Unfortunately I have to confirm that, like theonlydjcat, the game still crashes to desktop at the same turn as previously, i.e. when the A.I. creates an oversea colonies.

Note: we CAN continue the game if we play by hotseat, but NOT if by LAN. So it seems that the multiplayer problem remains…

 

[jdog5000]

Bummer ... anyone want to play an IP game with me sometime this weekend? It won't last very long, just load the save a wait for a crash!

 

[dexy]

Hi,

I play RoM 2.71 which includes RevDCM 2.5 and I get a CTD every 3-5 turns, but it can't be reproduced for sure at the end of some turn (I guess there's some randomness there).

What I see when I use JIT debug is that the problem always arises from the method CvPlot::isHasPathToEnemyCity(), when you do gDLL->getFAStarIFace()->Initialize(...).

Hope this helps. I can post a save, but you won't get a certain CTD for a couple of turns.

Dexy

 

[TheOnlyDJCat]

jdog,

I could initiate a game from the save today, if still needed? European time zone.

 

[phungus420]

RoM is so big, it doesn't demonstrate anything using it as an example.

I try so hard with my mod not to deviate from RevDCM, but I still need to triple check things. Also I've been right every time so far when people playing RoM blame RevDCM for hangs and crashes; RoM has been the cause of the crashes, not RevDCM. Like I said before, if you can actually demonstrate a flaw in the RevDCM code, or get an assert showing the issue is one of AI logic, show it, but I'd wager a donut, actually I'd wager a dozen donuts, it's RoM code that's the problem.

Afforess will probably fix it though. He's getting close to a stage where complaints such as yours will mean something. But right now, no, honestly, I doubt it's BBAI. And I would wager a baker's dozen Boston Creams on it.

 

[civ editor11]

Wow, you must be serious.

 

[NBAfan]

@dexy Are you using Afforess' modmod? If you are not, Rom does not change the Dll from Revdcm. I do not know if that means anything though.

 

[dexy]

@dexy Are you using Afforess' modmod? If you are not, Rom does not change the Dll from Revdcm. I do not know if that means anything though.

I'm using the plain RoM, I know the dll is the same as in RevDCM, that's why I posted the bug here.

But now I got some weird results with the changes I made in the game:
I decreased all the graphics options and turned off some of BUG's demanding features; I also installed the NoBuildings Module (should increase performance as well). The result: the CTDs appear every 6-8 turns (compared to 3-5 before)!!! The strange thing is that they appear at that same place every time

 

[jdog5000]

What happens if you play on a really small map? Like Duel size?

Also, what version of BBAI are you using? (shown by mousing over the big flag next to the minimap)

 

[dexy]

What happens if you play on a really small map? Like Duel size?

Also, what version of BBAI are you using? (shown by mousing over the big flag next to the minimap)

OK, I'll try that and see what happens when I get home (in a couple of hours). For now, I can tell you that I'm using the version that is inside RevDcm 2.5 (should be 0.78c, as specified in their thread).

 

[jdog5000]

Okay, the multi-player crash on colony creation has been fixed! Many thanks to TheOnlyDJCat for playing a few very short MP games with me to sort out the issues.

For those of you keeping score at home, the root of the problem was a fix for another bug ... alexman suggested in the UP forum that we move colony creation to a different time in the motherland's turn so that under very rare conditions some players wouldn't get a free second move during a turn. Well, for the same reason the code move solved that problem, it created this new problem during simultaneous move games. So, I'm pretty sure the same bug is in the current version of the UP as well.

For those of you with multiplayer games in progress or who want to play multiplayer, I've uploaded a new version to the development version download. It's labeled a pre-release version of 0.82, but should be stable. If you're only playing single-player or hotseat, there isn't much reason to switch ... beyond the mp fix, the main change is I added an AI decision making logging system so it's easier for me to analyze what the AI is up to. There are a few AI tweaks, but not very much.

If you're looking to merge this fix into your own mod, you'll need to get the fixes from two places, search for "10/05/09" in CvPlayer.cpp and CvPlayerAI.cpp.

 

[Afforess]

Hi,

I play RoM 2.71 which includes RevDCM 2.5 and I get a CTD every 3-5 turns, but it can't be reproduced for sure at the end of some turn (I guess there's some randomness there).

What I see when I use JIT debug is that the problem always arises from the method CvPlot::isHasPathToEnemyCity(), when you do gDLL->getFAStarIFace()->Initialize(...).

Hope this helps. I can post a save, but you won't get a certain CTD for a couple of turns.

Dexy

RoM is so big, it doesn't demonstrate anything using it as an example.

Please be respectful of other mods than you're own, even if you don't like them.

I try so hard with my mod not to deviate from RevDCM, but I still need to triple check things. Also I've been right every time so far when people playing RoM blame RevDCM for hangs and crashes; RoM has been the cause of the crashes, not RevDCM.

RoM has not be the cause of the crashes. If you will wander over to the UP bug page, and see my report and fix, you'll see that one such bug was a problem in vanilla BTS that was never used before.

Like I said before, if you can actually demonstrate a flaw in the RevDCM code, or get an assert showing the issue is one of AI logic, show it, but I'd wager a donut, actually I'd wager a dozen donuts, it's RoM code that's the problem.

Pure RevDCM has several asserts on loadup, at least for me on a debug build. So there are bugs there. Not major, but they exist. Since RoM has unaltered RevDCM SDK, I seriously doubt it is the cause of any crashes. You have to try pretty hard to make python or XML crash a game, and then, it's pretty obvious why it crashed.

Afforess will probably fix it though. He's getting close to a stage where complaints such as yours will mean something.

I'm not sure if this is meant in jest or not, but I'll assume so. I care about all complaints.

Okay, the multi-player crash on colony creation has been fixed! Many thanks to TheOnlyDJCat for playing a few very short MP games with me to sort out the issues.

Great work. It's another step towards getting complete MP compatibility. Job well done.

 

[dexy]

OK, the latest about the crashes:

I tried to modify the graphics and BUG options to max again and the number of turns between crashes reached 10 !!! So it's not about graphics overhead, it's something in the code.

Another thing - when I removed the accused-for-bug method isHasPathToEnemyCity() from code, I started to get crashes all over other code (randomly)!!!

So, finally, I guess the problem is not in that method (sorry for the false bug report ), just maybe it is the most CPU consuming and it had the bad luck that all my previous crashes ended up there.

I'll try to play some more games with autoplay, but I doubt I'll get any useful results.

 

[phungus420]

RoM has not be the cause of the crashes. If you will wander over to the UP bug page, and see my report and fix, you'll see that one such bug was a problem in vanilla BTS that was never used before.

Actually it has been. The infinite loop bug which seems to get alot of complaints in the RoM forum was caused by some RoM code allowing for cargo promotions. Sure the fix needed to be made in the SDK, but it's not fair to expect RevDCM to locate bugs like these. They need to be found on RoM's end if it's RoM's code causing the issue. Whenever a bug occurs in a RevDCM merge it's a good idea to assume it's caused by the merger, not the RevDCM core, unless the merger can track the cause back to RevDCM code.

Pure RevDCM has several asserts on loadup, at least for me on a debug build. So there are bugs there. Not major, but they exist. Since RoM has unaltered RevDCM SDK, I seriously doubt it is the cause of any crashes. You have to try pretty hard to make python or XML crash a game, and then, it's pretty obvious why it crashed.

With 2.5 I can't replicate this. I only get 2 consistent asserts, that one that says "ignore this" and points to some bizzare file path, and seems to mean nothing, and the other one that checks for the pressence of NO_STATE religion, which makes no sense, and is obviously an issue, but doesn't seem to cause any problems.

I'm not sure if this is meant in jest or not, but I'll assume so. I care about all complaints.

Actually I was being serious. Though I was smashed at the time, so my wording probably wasn't the most clear. Basically what I'm saying is that bugs experienced from large mergers should be handled in house first, and only reported here after someone with some knowlegde has checked the crash with a debug dll, and concluded it's something in the RevDCM core, and not a problem with the mod itself. With RoM, you're the person who will be the most responsive to check up on crash reports, so really the report should be handled and by yourself. Personally I don't want LoR users coming in to the RevDCM forums and reporting bugs they encounter with LoR here, I'd rather check it first, and be sure it's not a problem with my changes first. Same goes for RoM.

 

[JimMac99]

Okay, the multi-player crash on colony creation has been fixed! Many thanks to TheOnlyDJCat for playing a few very short MP games with me to sort out the issues.

For those of you keeping score at home, the root of the problem was a fix for another bug ... .

For those of you with multiplayer games in progress or who want to play multiplayer, I've uploaded a new version to the development version download. It's labeled a pre-release version of 0.82, but should be stable. . .

Thanks very much for fixing this, I am due to start a new MP game this Sunday, so we will use the development version (unless you release a 'final' 0.82 before then )

 

[Afforess]

Actually it has been. The infinite loop bug which seems to get alot of complaints in the RoM forum was caused by some RoM code allowing for cargo promotions. Sure the fix needed to be made in the SDK, but it's not fair to expect RevDCM to locate bugs like these. They need to be found on RoM's end if it's RoM's code causing the issue. Whenever a bug occurs in a RevDCM merge it's a good idea to assume it's caused by the merger, not the RevDCM core, unless the merger can track the cause back to RevDCM code.

This wasn't the waiting for bug. That was a glitch in the naval AI that BAI fixed with .81M. This was a CTD.

With 2.5 I can't replicate this. I only get 2 consistent asserts, that one that says "ignore this" and points to some bizzare file path, and seems to mean nothing, and the other one that checks for the pressence of NO_STATE religion, which makes no sense, and is obviously an issue, but doesn't seem to cause any problems.

Yeah, I get the ignore this asserts too, and there are (or were, they may have been fixed for 2.6) several asserts that fired when you let the AIAutoPlay run.

Actually I was being serious. Though I was smashed at the time, so my wording probably wasn't the most clear. Basically what I'm saying is that bugs experienced from large mergers should be handled in house first, and only reported here after someone with some knowlegde has checked the crash with a debug dll, and concluded it's something in the RevDCM core, and not a problem with the mod itself. With RoM, you're the person who will be the most responsive to check up on crash reports, so really the report should be handled and by yourself. Personally I don't want LoR users coming in to the RevDCM forums and reporting bugs they encounter with LoR here, I'd rather check it first, and be sure it's not a problem with my changes first. Same goes for RoM.

I misunderstood. I agree with this completely.

 

[jdog5000]

Thanks very much for fixing this, I am due to start a new MP game this Sunday, so we will use the development version (unless you release a 'final' 0.82 before then )

Great! Let me know how it goes ...