@saver I'm planning to request a news post once I've finished testing M28 as a hostile campaign AI on the rest of the Supcom campaign, since then I can highlight all the campaign features in one go and it's less likely to break the missions. For now it's mentioned in the mod description and on the wiki (re being an AI buddy), but I suspect not many people read them.
M28AI Devlog (v121)
M27AI and M28AI developer; Devlogs and more general AI development guide:
https://forum.faforever.com/topic/2373/ai-development-guide-and-m27ai-v71-devlog
https://forum.faforever.com/topic/5331/m28ai-devlog-v112
Here is a test of the first mission FA with your AI with three people. Unfortunately the game leaked a lot https://replay.faforever.com/21362409
In addition, construction took place again in the non-playable area. But otherwise it was a nice change
@saver said in M28AI Devlog (v58):
21362409
What do you mean by leaked a lot? did it slow down to -1 or worse game speed?
M27AI and M28AI developer; Devlogs and more general AI development guide:
https://forum.faforever.com/topic/2373/ai-development-guide-and-m27ai-v71-devlog
https://forum.faforever.com/topic/5331/m28ai-devlog-v112
The game speed has fallen from 0 to -6 in the game. Unfortunately I don't know if it can be seen in the replay.
I didnt notice when I was watching the replay but I stopped after 41m as I had enough points to look at by then, do you know roughly the timestamp where it went to -6? It makes sense that it would be slower in a big air battle although I wouldn't have expected -6 (assuming everyone has average to good cpus) so it could be an indication of something gone wrong with the code. It's also possible it's a result of some of the additional checks that have to be run in campaign maps that will add to the overhead (although again I wouldn't have expected such a big impact)
M27AI and M28AI developer; Devlogs and more general AI development guide:
https://forum.faforever.com/topic/2373/ai-development-guide-and-m27ai-v71-devlog
https://forum.faforever.com/topic/5331/m28ai-devlog-v112
The game suddenly freezes completely after a certain time. Without error message and without replay. Whether with 2 or 4 players against an M28AI. You then have to end the game via the Task Manager. However, this only happens when you have reached the maximum of 1000 units (i.e. when you go on the attack). And it happens in every game. I have also noticed that the speed sometimes drops to -6.
We play with the following mods:
- All Faction Quantum Gate
- M28AI
- Nuclear Repulsor Shields
- Total Mayhem
- Total Mayhem Lite
- 5x Build Range
Are you aware of any incompatibility with any of these mods?
@klecks Not individually (some like Quantum gate, total mayhem and build range I've seen work fine, theo thers I've not specifically tested), but issues are much more likely to occur with multiple mods, either due to an incompatibility with a particular combination of mods; a bug with the mod that M28's logic highlights, or memory issues (which are more likely with more mods), in addition to the potential for a bug with M28.
If it happens with just one of those mods enabled then I can take a look to figure out. The other thing you could do is see if there's a log file (even if there's no replay - although I'd have normally expected a replay to be available for games played in FAF even if the game crashes part-way through).
For example logs are saved for me at the following location:
C:\Users[username]\AppData\Roaming\Forged Alliance Forever\logs
Have you always had these issues or has it only been since the latest release (last 24 hours)?
Also enable more detailed logging in the settins in the client, which increases the chance that the cause of the crash can be identified from the log:
M27AI and M28AI developer; Devlogs and more general AI development guide:
https://forum.faforever.com/topic/2373/ai-development-guide-and-m27ai-v71-devlog
https://forum.faforever.com/topic/5331/m28ai-devlog-v112
v59 Update
A mix of campaign, replay and vs AI based changes with this update:
- Fighting M27 - 9 changes; interestingly M28 beat M27 on 4 different 5km and 10km maps that M27 had been tested on extensively (theta passage, polar depression, twin rivers, and eye of the storm), making me think M28 had finally reached the point of convincingly dethroning M28. However, M28 lost on 2 20km maps (setons and burial mounds), contrary to what I'd have expected. The main change added was improving M28's ability to handle an early strat if it has asfs.
- 8 changes based on replay comparisons from Radde between M28 and M27 on Xander, including making land experimentals more likely to move towards an enemy instead of attack-move, and fixed some issues with hos mongoose and MAA would retreat from units (the former being too cautious in some cases due to a bug, the latter ignoring nearby tanks due to a bug in some cases)
- 5 changes from 2 mapgens where Radde and Mhad beat 1.5 resource M28s, mainly trying to make M28 slightly more resistant to air snipes
- 9 changes from an M28 vs 1.2 RNG game on twin rivers (normally something M28 won but this time it lost after an epic 2 hour game due to poor fatboy usage) - the main change is making M28 retreat fatboys much sooner so they're less likely to be caught up to and killed, and increasing the extent to which M28 will eco when behind on eco.
- 11 changes for Saver's FA mission 1 of the campaign to make M28 more aggressive with navy, have a better unit mix, send GCs at the enemy start position (instead of keeping them on patrol), and being more likely to build T3 arti/similar
- 4 changes from a replay by Bart on gap of rohan, mainly fixing issues with overbuilding T2 arti and SAMs
Trophy awards
- Radde trophy - Awarded to Radde and Mhad for beating 1.5 resource M28 on both a mixed land+navy mapgen, and a navy mapgen
- Sladow trophy - Current holder is triangleleaf for beating v57 M28
Acknowledgements
- Radde - sending various replays including a comparison of M27 and M28 to help me identify significant differences in their approaches that make M27 seem tougher for a human player
- Saver - Replay against M28 as a hostile AI in campaign which highlighted a number of campaign specific issues
- Bart - Replay in a norush gap of rohan high eco game which highlighted some flaws with M28's arti and SAM logic
M27AI and M28AI developer; Devlogs and more general AI development guide:
https://forum.faforever.com/topic/2373/ai-development-guide-and-m27ai-v71-devlog
https://forum.faforever.com/topic/5331/m28ai-devlog-v112
@klecks said in M28AI Devlog (v59):
The game suddenly freezes completely after a certain time. Without error message and without replay. Whether with 2 or 4 players against an M28AI. You then have to end the game via the Task Manager. However, this only happens when you have reached the maximum of 1000 units (i.e. when you go on the attack). And it happens in every game. I have also noticed that the speed sometimes drops to -6.
We play with the following mods:
- All Faction Quantum Gate
- M28AI
- Nuclear Repulsor Shields
- Total Mayhem
- Total Mayhem Lite
- 5x Build Range
Are you aware of any incompatibility with any of these mods?
What you're talking about is probably related to memory. Maybe next time you can take a look at the memory usage for high unit quantities (via task manager).
@klecks FAF is prone to crashing if you reach over 2GB of memory usage; i.e. if you have taskmanager open and you see it approaching 2gb, then it's much more likely to crash.
The memory usage is increased significantly by unit mods (along with various other factors), and has been known to crash with AI (including but not limited to M28) since AI also use up more memory.
M27AI and M28AI developer; Devlogs and more general AI development guide:
https://forum.faforever.com/topic/2373/ai-development-guide-and-m27ai-v71-devlog
https://forum.faforever.com/topic/5331/m28ai-devlog-v112
@maudlin27 said in M28AI Devlog (v58):
AF is prone to crashing if you reach over 2GB of memory usage; i.e. if you have taskmanager open and you see it approaching 2gb, then it's much more likely to crash.
The memory usage is increased significantly by unit mods (along with various other factors), and has been known to crash with AI (including but not limited to M28) since AI also use up more memory.
OK, I understand. This is a screenshot from Task Manager when the game was frozen (two AI und only me). So the prolem with memory depends on the 32bit game
@maudlin27
Hello unfortunately I didn't write down any timestamps. I can only make the following statements about the PC used, the three of us use similar components: processor i7-8700 3.19GHz, 16 - 32GB RAM, Geforce 1080GTX and higher. I will also look at the replay to provide possible timestamps. Thank you again for your great work.
@maudlin27 Replay timestamp https://replay.faforever.com/21362409
min 43: -3, min 44: -4, min 45.20: -4, 45.40: -4, min 46.50: -5. Unfortunately I couldn't find -6. However, I only watched the end of the first section. I hope the data can be of some help to you.
@Klecks Dont use total mayhem lite (crashing/freezing random all the time), or total mayhem with this shieldmod. Better use some other mods if you want this t4 shields~ U can check my mod list some posts above.
Ansonsten frohe Festtage
m28ai seemed to crash twice when we used it, 2v2ai 3v1ai
33 min first game, 44 min second game
mods used / yes
https://replay.faforever.com/21423105 - 1st game
https://replay.faforever.com/21423406 - 2nd game
debug: Current gametime: 00:32:30
warning: false M28ERROR Count=54: GameTime 1951: Dont have a valid upgrade ID; UnitID=bsb0002
warning: stack traceback:
warning: [C]: in function assert' warning: [C]: in function
pcall'
warning: ...r forged alliance\mods\m28ai\lua\ai\m28utilities.lua(55): in function ErrorHandler' warning: ...der forged alliance\mods\m28ai\lua\ai\m28economy.lua(137): in function
UpgradeUnit'
warning: ...mander forged alliance\mods\m28ai\lua\ai\m28team.lua(2659): in function GetAnyMexOrFactoryToUpgrade' warning: ...mander forged alliance\mods\m28ai\lua\ai\m28team.lua(2880): in function
ConsiderNormalUpgrades'
warning: ...mander forged alliance\mods\m28ai\lua\ai\m28team.lua(2930): in function <...mander forged alliance\mods\m28ai\lua\ai\m28team.lua:2908>
warning: false M28ERROR Count=3: GameTime 1953: After filtering to a faction we have no available engineers - this shouldnt be possible
warning: stack traceback:
warning: [C]: in function assert' warning: [C]: in function
pcall'
warning: ...r forged alliance\mods\m28ai\lua\ai\m28utilities.lua(55): in function ErrorHandler' warning: ...er forged alliance\mods\m28ai\lua\ai\m28engineer.lua(4398): in function
FilterEngineersOfTechAndEngiCountForFaction'
warning: ...er forged alliance\mods\m28ai\lua\ai\m28engineer.lua(5564): in function ConsiderActionToAssign' warning: ...er forged alliance\mods\m28ai\lua\ai\m28engineer.lua(6420): in function
HaveActionToAssign'
warning: ...er forged alliance\mods\m28ai\lua\ai\m28engineer.lua(6681): in function ConsiderCoreBaseLandZoneEngineerAssignment' warning: ...er forged alliance\mods\m28ai\lua\ai\m28engineer.lua(11595): in function
ConsiderLandOrWaterZoneEngineerAssignment'
warning: ...mander forged alliance\mods\m28ai\lua\ai\m28land.lua(5857): in function <...mander forged alliance\mods\m28ai\lua\ai\m28land.lua:5349>
warning: false M28Warning: Count=1: GameTime 1956: No plateau group for iSegmentX=336; iSegmentZ=249; Plateau group of segment midpoint=2751; Plateau Group of tPosition=2751; This can happen on larger maps with long cliffs with narrow paths that small units can travel partially along. Enable logs in the function GetUnitPlateauAndLandZoneOverride for more details
warning: stack traceback:
warning: [C]: in function assert' warning: [C]: in function
pcall'
warning: ...r forged alliance\mods\m28ai\lua\ai\m28utilities.lua(55): in function ErrorHandler' warning: ...mmander forged alliance\mods\m28ai\lua\ai\m28map.lua(571): in function
GetPlateauAndLandZoneReferenceFromPosition'
warning: ...mander forged alliance\mods\m28ai\lua\ai\m28land.lua(317): in function UpdateUnitPositionsAndLandZone' warning: ...mander forged alliance\mods\m28ai\lua\ai\m28land.lua(6067): in function <...mander forged alliance\mods\m28ai\lua\ai\m28land.lua:6007> warning: false M28Warning: Count=8192: GameTime 1959: We are unable to build at a predefined shield location, will check for small blocking M28 buildings and destroy them warning: stack traceback: warning: [C]: in function
assert'
warning: [C]: in function pcall' warning: ...r forged alliance\mods\m28ai\lua\ai\m28utilities.lua(55): in function
ErrorHandler'
warning: ...er forged alliance\mods\m28ai\lua\ai\m28engineer.lua(4648): in function ConstructNewShield' warning: ...er forged alliance\mods\m28ai\lua\ai\m28engineer.lua(4748): in function <...er forged alliance\mods\m28ai\lua\ai\m28engineer.lua:4406> warning: false M28Warning: Count=1: GameTime 1959: Adding unit ID brot3bt2 to table of temp other units - either the unit is a land unit really close to water that we incorrectly think is in water, or it is an amphibious/hover unit that has incorrect caterisation; if the unit runs out of orders then will send it to the rally point warning: stack traceback: warning: [C]: in function
assert'
warning: [C]: in function pcall' warning: ...r forged alliance\mods\m28ai\lua\ai\m28utilities.lua(55): in function
ErrorHandler'
warning: ...mander forged alliance\mods\m28ai\lua\ai\m28navy.lua(1488): in function <...mander forged alliance\mods\m28ai\lua\ai\m28navy.lua:1331>
info: OnDecayed: Time=1961.3000488281; self.UnitId=brnt3perses
debug: Loading module: /projectiles/adfreactoncannon01/adfreactoncannon01_script.lua
debug: Loading module: /projectiles/laserbotterran01/laserbotterran01_script.lua
warning: false M28ERROR Count=81: GameTime 1975: Dont have a valid upgrade ID; UnitID=bsb0002
warning: stack traceback:
warning: [C]: in function assert' warning: [C]: in function
pcall'
warning: ...r forged alliance\mods\m28ai\lua\ai\m28utilities.lua(55): in function ErrorHandler' warning: ...der forged alliance\mods\m28ai\lua\ai\m28economy.lua(137): in function
UpgradeUnit'
warning: ...mander forged alliance\mods\m28ai\lua\ai\m28team.lua(2659): in function GetAnyMexOrFactoryToUpgrade' warning: ...mander forged alliance\mods\m28ai\lua\ai\m28team.lua(2880): in function
ConsiderNormalUpgrades'
warning: ...mander forged alliance\mods\m28ai\lua\ai\m28team.lua(2930): in function <...mander forged alliance\mods\m28ai\lua\ai\m28team.lua:2908>
debug: Loading module: /mods/totalmayhem/projectiles/brot3shbmempproj/brot3shbmempproj_script.lua
info: OnDecayed: Time=1979.8000488281; self.UnitId=xrb9304
debug: Current gametime: 00:33:00
debug: Loading module: /mods/blackopsfaf-unleashed/units/bsb5205/bsb5205_script.lua
debug: Loading module: /mods/totalmayhem/projectiles/brpt1extank2proj/brpt1extank2proj_script.lua
debug: Loading module: /projectiles/tdfplasmaheavy03/tdfplasmaheavy03_script.lua
debug: Loading module: /units/xsl0303/xsl0303_script.lua
debug: Loading module: /mods/blackopsfaf-unleashed/units/bab2306/bab2306_script.lua
debug: Current gametime: 00:33:30
info: Minimized true
info: Minimized false
debug: Current gametime: 00:44:00
info: CWldSession::DoBeat() unknown entity id (0x2ff0029a) supplied in a pose update.
warning: Error running OnNotAdjacentTo script in Entity urb1301 at 6d934808: ...\gamedata\lua.nx2\lua\sim\adjacencybufffunctions.lua(18): attempt to call method DestroyAdjacentEffects' (a nil value) stack traceback: ...\gamedata\lua.nx2\lua\sim\adjacencybufffunctions.lua(18): in function
OnBuffRemove'
...gramdata\faforever\gamedata\lua.nx2\lua\sim\buff.lua(687): in function RemoveBuff' ...ver\gamedata\lua.nx2\lua\sim\units\structureunit.lua(810): in function <...ver\gamedata\lua.nx2\lua\sim\units\structureunit.lua:787> [C]: in function
Kill'
...gramdata\faforever\gamedata\lua.nx2\lua\sim\unit.lua(1320): in function DoTakeDamage' ...ver\gamedata\lua.nx2\lua\sim\units\structureunit.lua(664): in function
DoTakeDamage'
...gramdata\faforever\gamedata\lua.nx2\lua\sim\unit.lua(1286): in function OnDamage' ...gramdata\faforever\gamedata\lua.nx2\lua\sim\unit.lua(5751): in function <...gramdata\faforever\gamedata\lua.nx2\lua\sim\unit.lua:5750> [C]: ? ...ta\faforever\gamedata\lua.nx2\lua\sim\projectile.lua(623): in function
DoDamage'
...ta\faforever\gamedata\lua.nx2\lua\sim\projectile.lua(364): in function `OnImpact'
...uclearshell01\tifhetacticalnuclearshell01_script.lua(12): in function <...uclearshell01\tifhetacticalnuclearshell01_script.lua:11>
info: Minimized true
info: Minimized false
@thewatcheral said in M28AI Devlog (v59):
21423105
Thanks, looks like there's an incompatibility with the experimental shield mod that triggered when the experimental shield was killed (and is highly likely to have caused the crash), I'll see if I can resolve in the next release.
M27AI and M28AI developer; Devlogs and more general AI development guide:
https://forum.faforever.com/topic/2373/ai-development-guide-and-m27ai-v71-devlog
https://forum.faforever.com/topic/5331/m28ai-devlog-v112