Spring RTS Engine

Hi,

I got a pretty bad problem which causes my spring to freeze in midgame in 4 of 5 games. It happens without any obivous reason, the game just freezes, Vista tells me "the program doesnt responds anymore" and I have to close it via the taskManager. While that freeze spring uses one cpu core at full load. Last time I was only microing some units around when it happend. This can happen after 5 minutes but can also happen after 60 minutes. I tried all in last few days, I still have no idea what its coming from. I had this problem some times in the past but at the moment I can barely play a complete game of spring
My first guess was something with my graphics card but even after some testing I wasnt able to find a reason for it.

These are my specs:
Core2Duo E6300 with 6GB RAM
8800GT
Vista 64Bit
Tested with BA65

This is what I tried already:
-Reinstalled spring in a new directory
-Resetted all settings to default via springsettings.exe
-Updated to latest ForceWare 178.24
-Ran an OpenGL stress test which worked fine
-Turned off threaded-optimization in Nvidia-drivers

While testing stuff I found out that my spring also freezes in the same way if I active and deactive IceUI in the F11 exactly three times. When switching it on for the third time it makes my spring freeze exactly the same way it happens ingame. I dont think it is related to my "main problem" but it is the same result of freezing.

The bad thing is I am not able to reproduce the error spontaneous. Even if I play a game and it works it doesnt mean it cant freeze in the next game
So please, any hints are very appreciated, big thanks in advance!

PS. Maybe someone could send me replay of a DSD-BA65 match that lasts at least an hour? I could test if it happens too while watching a replay.

I appended a debugger last time it happened again, but I dont know where I can translate it. Maybe I can find the cause of the problem with it?

This is the stacktrace of the main thread after pausing the process while it was freezed:

should help. lemme see...

Big Thanks! Hm, does it make any sense for you? I looked through the source code but I dont even know how the functions of the stacktrace are related to each other but I didnt see anything I would suspect to cause an endless loop.
The last call seems to be something in the Lua-Code, do you think its probably a problem with a LUA-widget? I usually use alot of them.

this may actually be a loop; try resuming in the debugger and see if stacktraces differ (and where.)

if you download gdb/addr2line and debug symbols (see bibim's site for svn builds), you should be able to trace it yourself. also, there've been reports of windows global shortcut for keyboard layout switching (ctrl+alt or something) as the cause for lockup.

Ok thanks, I was at least able to step through the assembly and to translate some addresses manually.
So I think I figured out that when the freeze occurs I get somehow stuck in the function luaV_execute in lvm.cpp line 402.
I tested against a running spring: it enters this function and leaves it a short time after again (not that surprising). But in my case he never gets out there again. There is an for(;;)-loop inside that function, probably this loops endlessly without getting to a return-statement.
I forced him to get to the "return" by editing the instruction pointer which at least made spring to throw an unhandled exception and close gracefully (again not very surprising).

That function looks like some low-level-lua-stuff and I really have no idea what is happending there. If somehow who knows about it could take a look at it I would be really happy, big thanks in advance! It really bugs me to freeze up nearly every game after playing it halfway

The IceUI stuff was an endless recursion.. It was just my bug and will be fixed in the next IceUI update..

Thanks for the report btw :)

MelTraX wrote:The IceUI stuff was an endless recursion.. It was just my bug and will be fixed in the next IceUI update..

Thanks for the report btw :)

Nice1, but this seems to be another bug because I disabled IceUI (and some other widgets) for testing purpose some time ago and it still happens.

I still encounter this bug nearly every game.
I used a widget that logs all calls to other widgets and when they are completed. But it seems not to be a "normal" user widget which freezes.
That was my last idea, I absolutley dont know whats the cause of this. If a dev would help me out with that issue it would be highly appreciated and eternal gratefulness would be yours.

tip, not all lua errors are caused by lua (remember ground decal crash in 76b1)

So, any idea how I could continue to debug this?
Or even how I could reproduce the bug? It doesnt occur when running replays or spectating other games.

I wiped spring completely from my HDD yesterday and reinstalled everything and resetted all settings. I disabled all features in springsettings and just installed a handful of tiny widgets. I also stopped all running background windows processes. But it still happens.

since a few weeks i also get some freezes in spring, but it happens only once in 10 or 15 games. it doesnt always freeze completely, sometimes it recovers after some time. i also have 8800GT. perhaps its something related to the graphics driver? using a very old one would be a nice try. also in your case its perhaps a hardware problem? perhaps RAM?

very_bad_soldier wrote:Hi,

I got a pretty bad problem which causes my spring to freeze in midgame in 4 of 5 games. It happens without any obivous reason, the game just freezes, Vista tells me "the program doesnt responds anymore" and I have to close it via the taskManager. While that freeze spring uses one cpu core at full load. Last time I was only microing some units around when it happend. This can happen after 5 minutes but can also happen after 60 minutes. I tried all in last few days, I still have no idea what its coming from. I had this problem some times in the past but at the moment I can barely play a complete game of spring
My first guess was something with my graphics card but even after some testing I wasnt able to find a reason for it.

These are my specs:
Core2Duo E6300 with 6GB RAM
8800GT
Vista 64Bit
Tested with BA65

This is what I tried already:
-Reinstalled spring in a new directory
-Resetted all settings to default via springsettings.exe
-Updated to latest ForceWare 178.24
-Ran an OpenGL stress test which worked fine
-Turned off threaded-optimization in Nvidia-drivers

While testing stuff I found out that my spring also freezes in the same way if I active and deactive IceUI in the F11 exactly three times. When switching it on for the third time it makes my spring freeze exactly the same way it happens ingame. I dont think it is related to my "main problem" but it is the same result of freezing.

The bad thing is I am not able to reproduce the error spontaneous. Even if I play a game and it works it doesnt mean it cant freeze in the next game
So please, any hints are very appreciated, big thanks in advance!

PS. Maybe someone could send me replay of a DSD-BA65 match that lasts at least an hour? I could test if it happens too while watching a replay.

Wernt there more people with that problem? like fang?

every time I start spring, without fail, it will sporadically hit a cpu-intensive loop and freeze (delayed response from myself for frame xx) for a bit, and barely be responsive.
pressing B reveals: draw interface has taken copious amounts of my time vs all of the other calls...

aegis wrote:every time I start spring, without fail, it will sporadically hit a cpu-intensive loop and freeze (delayed response from myself for frame xx) for a bit, and barely be responsive.
pressing B reveals: draw interface has taken copious amounts of my time vs all of the other calls...

I had that with 0.77b3 too at some point, but then it suddenly disappeared.

very_bad_soldier wrote:Hi,

I got a pretty bad problem which causes my spring to freeze in midgame in 4 of 5 games. It happens without any obivous reason, the game just freezesF

°

I have THE SAME problem - it never ever happend in 76 and i did not change a thing since then. Will try to digg which thread causes this crap.

BTW my PC is super stable tested with prime95 , ati artefacet finder and video card stability test.

Spec:
E8400
NV 9500GT
win xp
ba65, many widgets, iceui

It just froze again. Here is stack log from process explorer:

ntkrnlpa.exe!KiUnexpectedInterrupt+0x8d
ntkrnlpa.exe!PsDereferencePrimaryToken+0x362
ntkrnlpa.exe!KiDeliverApc+0xb3
hal.dll+0x2c35
spring.exe+0x40159d
spring.exe+0x417120
spring.exe+0x40ff21
spring.exe+0x416962
spring.exe+0x416a0f
spring.exe+0x40fd9d
spring.exe+0x2f508f
spring.exe+0x2f51ae
spring.exe+0x30b17e
spring.exe+0x1875b5
spring.exe+0x5223ff
spring.exe+0x67f341
spring.exe+0x1c4d14
spring.exe+0x1b77bd
spring.exe+0x1b79b9
spring.exe+0x634c01
spring.exe+0x10a7
spring.exe+0x1123
kernel32.dll!RegisterWaitForInputIdle+0x49

Also from process explroers cpu history bar i can see that after freeze game uses only CPU (green graph) and no more kernel fucntios like GFX (red graph). It only happens when i actually play, if i leave spring in single player for hours it runs fine. What else to try?

After freeze i have to kill spring.exe.

1 Q:

can lua cause freeze ? Can widget?

Widgets can freeze and crash spring, or at least I've managed to do accidentally do it when developing stuff...

I used a modified version of Meltrax's widget profiler to test for widget freezes. Instead of profiling, it logs the last called LUA function to a file on harddisk ("luacalls.dbg" in spring-folder). Also it logs when the function gets left again.

Something like this means the function DrawScreen from the widget DGunLimit was entered. ("C" = called). When you see a CL-line (closed) it means the function was left again. This is the normal case. If your spring freezes then look into the file. When you see in the file just the line: Then it means the function was entered but never left again -> you probably freezed inside that function. This method doesnt work to 100% since it can also mean the lua engine was still runnning and you just terminated it when it was inside a LUA function but I think that isnt very likely.

But in my case it was not a freezed widget since I could see both C and CL-line in the logfile.

BTW: I installed a quite old Nvidia-driver and I am testing at the moment, nice idea, thx!

EDIT: Cause of very frequent file writes the widget produces a noticeable lag every few seconds. But I could not get around that sadly...