Duke4.net Forums: Eduke32 Randomly Crashes - Duke4.net Forums

Jump to content

Hide message Show message
Welcome to the Duke4.net Forums!

Register an account now to get access to all board features. After you've registered and logged in, you'll be able to create topics, post replies, send and receive private messages, disable the viewing of ads and more!

  • 2 Pages +
  • 1
  • 2
  • You cannot start a new topic
  • You cannot reply to this topic

Eduke32 Randomly Crashes  "I have no idea why"

User is offline   Rahu X 

  • 0

#1

Just recently got Eduke32 so I can play Duke Nukem 3D again on Windows 7 while I'm waiting for DNF. Everything works fine for awhile, but around 30-45 min into the game, it randomly crashes.

My OS is Windows 7 Home Premium 64-bit
GPU is a ATI Radeon HD 5770
CPU is a AMD Phenom II X4 965 @ 3.4ghz
12GB of RAM

Using the latest snapshot of Eduke32, but this also occurs on the public build. I just upgraded to the latest snapshot to see if it would help.

I'm also using the public build of HRP.

This is the error it gives me in my log file:

Fatal Signal caught: SIGSEGV. Bailing out.

 2438: 31337 } 
 2439:  1293 ends  0 5 1 1 1 0 532415296 0 0
 2449:  1297 state  532415120
 2451:  1298 enda  0 532415296 0 0
 2456:  1301 state  532415120
 2458:  1302 enda  0 0 8 0
 2463:  1305 ifpdistl  3084 532415420
 2466:  1306 ifrnd  24 532415420
 2469:  1307 spawn  661
 2471:  1308 enda  0 0 0 0 0 1
current actor: 425 (1255)
g_errorLineNum: 1302, g_tw: 4


If you need the whole thing, I can post that as well, though it's pretty long.

Any ideas?

This post has been edited by Rahu X: 08 September 2010 - 12:09 PM

0

Buy Duke3D from GOG.com!

User is offline   Trooper Dan 

  • Duke Plus Developer
  • 742

#2

Post the whole log as an attachment or use pastebin. The beginning of the log contains important information, such as which revision of eduke you are using, your OS, and which CONs you are using.
0

User is offline   Rahu X 

  • 0

#3

View PostDeeperThought, on Sep 8 2010, 03:20 PM, said:

Post the whole log as an attachment or use pastebin. The beginning of the log contains important information, such as which revision of eduke you are using, your OS, and which CONs you are using.

Here you go.
0

User is offline   Trooper Dan 

  • Duke Plus Developer
  • 742

#4

It could be an ATI compatibility issue with Polymer, but I don't know. When it crashes again, see if that last part (the code listing following "Fatal Signal caught: SIGSEGV. Bailing out.") is the same.
0

User is offline   Rahu X 

  • 0

#5

View PostDeeperThought, on Sep 8 2010, 04:21 PM, said:

When it crashes again, see if that last part (the code listing following "Fatal Signal caught: SIGSEGV. Bailing out.") is the same.


It seems mostly the same. Some numbers are different, but basically the exact same result.

Fatal Signal caught: SIGSEGV. Bailing out.

 2438: 31337 } 
 2439:  1293 ends  0 5 1 1 1 0 538893472 0 0
 2449:  1297 state  538893296
 2451:  1298 enda  0 538893472 0 0
 2456:  1301 state  538893296
 2458:  1302 enda  0 0 8 0
 2463:  1305 ifpdistl  3084 538893596
 2466:  1306 ifrnd  24 538893596
 2469:  1307 spawn  661
 2471:  1308 enda  0 0 0 0 0 1
current actor: 425 (1255)
g_errorLineNum: 1302, g_tw: 4


EDIT: Yep. It seems to be a problem associated with Polymer. Tried playing the game without it and it ran fine.

Too bad Polymer doesn't seem to like ATI that much... :lol:

This post has been edited by Rahu X: 08 September 2010 - 05:33 PM

0

User is offline   Chip 

  • 17

#6

Oh, so crashing after long play is not a common problem? I honestly thought that Polymer was still suffering from memory leaks or the sort and that everyone had it.



Well in that case, I too have this problem and indeed have an ATI card. The next time I run Eduke32, I'll get a log but when it crashes its a hard crash where not even the music is looped and I must call windows to close down the program. However, windows spends a long time trying to recover from this, like 5+ minutes - as if it had very little available memory to work with, like there was a memory leak that eventually consumed all my RAM.
This crash occurs earlier the more map loading screens you go through. Typically the game will crash and stop loading during the third map loading screen of the run. If I'm running through the game quickly then I can get to the forth map load screen without it crashing..... but the game will crash very soon from that point.

Also, before the game totally stops dead, it will start to "jump" quite a few times a minute (like when you first see an unloaded texture but reputably), when this starts to happen, the game will usually crash with the next 1-2 minutes. At this point I save progress and close down Eduke32 to avoid the soon-to-kill-all-crash. Restarting Eduke32 straight after and loading to my save is fine.



I have yet to try the latest Eduke32 build but this only occurs while running Polymer and each time, I have ran the HRP. I guess I should try Polymer without the HRP and see if it lasts longer?
Oh and when the game fully crashes, it records no error nor does it try to bail out - the log is left unfinished.

This post has been edited by Chip: 10 September 2010 - 05:23 AM

0

User is offline   SkullHacker 

  • 0

#7

Same here :)

Asus g73 laptop with Radeon mobility HD5870. Crashes randomly during gameplay, And in Mapster32 when setrendermode 4 is used it crashes quite a lot (at least once a minute!), especially in areas with lots of point/spot lights...
makes mapping with the new polymer lights almost impossible...

I don't get the 'jumping' you described though. Just a complete freeze, and then windows telling me eduke32/mapster32 has stopped working... and the log always shows something like this:

Fatal Signal caught: SIGSEGV. Bailing out.

6167: 3144 enda 140 0 0 0
6172: 3145 enda 150 0 0 0
6177: 3146 enda 50 0 0 0
6182: 3147 enda 0 0 0 0
6187: 3148 enda 15 0 0 0
6192: 3151 ifcount 18 75404636
6195: 3152 killit
6196: 3153 else 75404712
6198: 3154 ifcount 9 75404712
6201: 31337 {
6202: 3156 sizeto 0 0
6205: 3157 sizeto 0
current actor: 29 (1646)
g_errorLineNum: 3148, g_tw: 4

This post has been edited by SkullHacker: 11 September 2010 - 10:27 AM

0

User is offline   jungler 

  • 0

#8

same problem here :) after few minutes of play game freez totaly. I have ATI HD5650. I needed to turn polymer off and it works now. Too bad that i can not use polymer with my graphics card :)

I3 330M 2,13 GHz
ATI mobility radeon HD5650
4GB RAM
Win 7 64bit

This post has been edited by jungler: 14 September 2010 - 02:56 AM

0

User is offline   RazorZ 

  • 1

#9

Unfortunately my Eduke32 crashes, too. Not because of Polymer or HRP, I canīt use then (slowdown), but because when I turn on the cheat "Show Map ON", take a look at the map (unless I donīt do it right away - but itīs quite random, though) and the game crashes. It has happened to me frequently since 2008-2009 (earlier versions worked just fine). :)

Any ideas? :)

Mapper from Finland.
My maps:
Released:Resistance Rocks, Demolition Plant, Mincer_City
Unreleased (WIP): Dead Docks, Laboratory Lancer
0

User is offline   jungler 

  • 0

#10

maybee do not use cheats? :-D

This post has been edited by jungler: 14 September 2010 - 05:40 AM

0

User is offline   TerminX 

  • EDuke32 Project Manager
  • 2,243

  #11

View PostRazorZ, on Sep 14 2010, 04:57 AM, said:

Unfortunately my Eduke32 crashes, too. Not because of Polymer or HRP, I canīt use then (slowdown), but because when I turn on the cheat "Show Map ON", take a look at the map (unless I donīt do it right away - but itīs quite random, though) and the game crashes. It has happened to me frequently since 2008-2009 (earlier versions worked just fine). :)

Any ideas? :)

Sure, provide the map that causes the crash in map mode and I'll fix it.
0

User is offline   trencheel303 

  • 6

#12

Mine does the same, anything from 20 minutes to an hour into a game, for no apparent reason whatsoever. It doesn't bother me because a) I have a very short attention span and most of the time, I quit out myself before it crashes and :P it probably doesn't help that I use really odd hardware.
0

User is offline   Alithinos 

  • 21

#13

I too have an ATI graphics card but it works brilliant at Windows XP 32bit,with only 2gbs of ram.

Let me ask you a question...

Did you had another version of Windows before 7,and you installed the game at other version and then used the upgrade option that says it will not erase everything from your hard drive ?
If you installed a program at a windows version and then you upgraded to another one without formatting your disc there is a high possibility that your programs will become problematic and crash no matter what microsoft says about "safe transfer".

If that is a yes,then you will have to uninstall and reinstall everything Duke Nukem 3d related.
If it's a no,then there is something else happening.

But please be aware that Windows 7 is a new system and neither Microsoft have made enough patches (service packs) and perhaps the eDuke32 and HRP modders don't have yet enough expirience about how it works under Windows 7.

I think it's not a hardware problem,and it's just that eDuke32 doesn't communicate as it should with Windows.
:P

This post has been edited by Alithinos: 22 September 2010 - 04:41 PM

0

User is offline   Awesomebob 

  • 0

#14

View PostAlithinos, on Sep 22 2010, 05:39 PM, said:

I too have an ATI graphics card but it works brilliant at Windows XP 32bit,with only 2gbs of ram.

Let me ask you a question...

Did you had another version of Windows before 7,and you installed the game at other version and then used the upgrade option that says it will not erase everything from your hard drive ?
If you installed a program at a windows version and then you upgraded to another one without formatting your disc there is a high possibility that your programs will become problematic and crash no matter what microsoft says about "safe transfer".

If that is a yes,then you will have to uninstall and reinstall everything Duke Nukem 3d related.
If it's a no,then there is something else happening.

But please be aware that Windows 7 is a new system and neither Microsoft have made enough patches (service packs) and perhaps the eDuke32 and HRP modders don't have yet enough expirience about how it works under Windows 7.

I think it's not a hardware problem,and it's just that eDuke32 doesn't communicate as it should with Windows.
:P


Then it should be fine in compatibility mode, and it's not.

Posted Image
0

User is offline   tinybob 

  • 0

#15

I know this is probably just beating a dead horse, but my eduke32 also crashes. I can make it almost completely through hollywood holocaust, going slow, when it exits. but it doesnt give me an error report, at least not that i can tell, it just closes.
0

User is offline   The Dukenator 

  • 1

#16

View Posttinybob, on Oct 15 2010, 02:38 AM, said:

I know this is probably just beating a dead horse, but my eduke32 also crashes. I can make it almost completely through hollywood holocaust, going slow, when it exits. but it doesnt give me an error report, at least not that i can tell, it just closes.


Post your Eduke32.log.

The Dukenator's Raptr
These are English forums; anything the moderators can't read is eligible for deletion. - TerminX
For a second there, I thought you was Basky - Ivan
0

User is offline   tinybob 

  • 0

#17

View PostThe Dukenator, on Oct 15 2010, 01:05 AM, said:

Post your Eduke32.log.


The logAttached File  eduke32.log (2.07K)
Number of downloads: 143 should be attached. Unless i did something horribly wrong during the attachment process.
And Ive been running it without polymer, if that matters.

and here is a dxdiag if it helps at all: Attached File  DxDiag.txt (41.61K)
Number of downloads: 194
0

User is offline   Trooper Dan 

  • Duke Plus Developer
  • 742

#18

View Posttinybob, on Oct 15 2010, 11:25 PM, said:

The logAttachment eduke32.log should be attached. Unless i did something horribly wrong during the attachment process.
And Ive been running it without polymer, if that matters.

and here is a dxdiag if it helps at all: Attachment DxDiag.txt


Strange. Does it crash in all maps? According to that log, you were playing Toxic Dump, which has two corrupt sprites (not sure if that's relevant), and the game was processing the WATERBUBBLEMAKER actor when it crashed.
0

User is offline   Helixhorned 

  • EDuke32 Engine Maintainer
  • 610

#19

Are you running the synthesis builds? It's good to see that catching crashes works for you (the "bailing out" message), because I can't get it to work on Windows for the life of me. TX/Plagman: could we build debug snapshots in synthesis for the time being? I got a stack trace printer on crash ready, but in Windows it won't do much when compiled as release build, probably due to -fomit-frame-pointer. Optimizing makes the traces much less useful, though.

(o~ Lunatic home: documentation, stand-alone LuaJIT, LunaCON and utilities ~o)

This post has been edited by Helixhorned: 16 October 2010 - 04:43 AM

0

User is offline   tinybob 

  • 0

#20

View PostDeeperThought, on Oct 16 2010, 02:05 AM, said:

Strange. Does it crash in all maps? According to that log, you were playing Toxic Dump, which has two corrupt sprites (not sure if that's relevant), and the game was processing the WATERBUBBLEMAKER actor when it crashed.


On toxic dump it has crashed multiple times in different parts. It has also crashed on every map in the L.A. meltdown series so far. I've gotten around that by saving frequently so that i end up doing the level in smallish chunks. Good news is that i will be getting my usual computer fixed soon, so i should be able to test it out on that and see if i have the same problems.

Helixhorned: i do not know if im running synthesis builds. I literally got duke 3d and eduke about 3 days ago, and havent really learned much about them. I'm for sure not running polymer cause it wouldnt get past the opening title sequence.
0

User is offline   Plagman 

  • VP of Media Operations
  • 290

  #21

View PostHelixhorned, on Oct 16 2010, 05:29 AM, said:

Are you running the synthesis builds? It's good to see that catching crashes works for you (the "bailing out" message), because I can't get it to work on Windows for the life of me. TX/Plagman: could we build debug snapshots in synthesis for the time being? I got a stack trace printer on crash ready, but in Windows it won't do much when compiled as release build, probably due to -fomit-frame-pointer. Optimizing makes the traces much less useful, though.


Yeah, my original plan was to make synthesis build debug snapshots alongside the regular ones and to throw in some code to unroll the stack and print a backtrace from inside the crash handler. I tried to do this first, but for some reason I could never get a full backtrace, the last frame is always the first of the signal handler (even though I double checked I wasn't installing a signal alternate stack). Anyway, this was in sdlayer.c and probably of limited usefulness as linux users are usually more able to use gdb and give out useful info after a crash. I'll give it another try using the debugging tools MS provides (I think you can get a few symbols from a system DLL that will unroll the stack for you), but I'm worried it won't work with apps built from MinGW. I'll try it again sometime.

EDuke32 - "The corrupt doctrine of terror has begun."
0

User is offline   Plagman 

  • VP of Media Operations
  • 290

  #22

HH, I just saw your change and the code you added to sdlayer is pretty much the exact same I had and sure enough, I get the useless output as well. Here's what I get:

./eduke32[0x81c823b]
[0xfeb400]

Where 0x81c823b is the return address of the backtrace() call inside sighandler. Do you get useful output when crashing with that code?

EDuke32 - "The corrupt doctrine of terror has begun."
0

User is offline   Helixhorned 

  • EDuke32 Engine Maintainer
  • 610

#23

With the debug builds, I get useful output like this:

./eduke32[0x81c786b]
[0xb76f0400]
./eduke32[0x806ed9c]
./eduke32[0x81c7797]
/lib/libc.so.6(__libc_start_main+0xe7)[0xb7420ce7]
./eduke32[0x804b091]

Or, if linked with -rdynamic and compiled without -fomit-frame-pointers, adorned with names. With release builds, I'm able to get at least to the address of the function responsible for the crash:

./eduke32[0x81974df]
[0xb7868400]
./eduke32[0x819a8a4]

Here, 0x819a8a4 is the second instruction of setvlinebpl as described in the bug tracker. For this to work without frame pointers though, -fasynchronous-unwind-tables is needed.

So, for Linux the situation is fine IMO. I'm having a much harder time achieving the same effect on Windows. Since there's no backtrace() in the MinGW-provided libs, I use code based on this article to get the addresses. So far, so good. However, I'm unable to catch SIGSEGV. I tried everything from setting the handler just before the crash to using Windows' SetUnhandledExceptionFilter instead of signal(), but it just won't catch! And in all probability, if it worked, it would produce the wrong results like for the original poster of the article.

edit: er, wrong url

(o~ Lunatic home: documentation, stand-alone LuaJIT, LunaCON and utilities ~o)

This post has been edited by Helixhorned: 17 October 2010 - 12:16 PM

0

User is offline   Plagman 

  • VP of Media Operations
  • 290

  #24

Weird; how are you triggering the crashes? All my experiments have been with kill -11, so maybe that has something to do with it.

EDuke32 - "The corrupt doctrine of terror has begun."
0

User is offline   Helixhorned 

  • EDuke32 Engine Maintainer
  • 610

#25

Usually with something like *(int*)123=456.

(o~ Lunatic home: documentation, stand-alone LuaJIT, LunaCON and utilities ~o)
0

User is offline   Plagman 

  • VP of Media Operations
  • 290

  #26

Aaah yeah, the output is a lot more meaningful with real crashes as opposed to `kill -11`. Cool! Do you ever get symbol names from the engine, though? I only seem to get symbols from gameland. I wonder if something gets lost in translation when statically linking against libbuild.a, though nm -D successfully seems to list the engine symbols from the resulting binary. I'll try to see if I can get something similar going on in Win32, though I don't have high hopes if you already tried and failed.

I'm also dist-upgrading in order to try to reproduce the asm crash you reported.

EDuke32 - "The corrupt doctrine of terror has begun."
0

User is offline   Plagman 

  • VP of Media Operations
  • 290

  #27

Alright, can you guys try the newest snapshot:

http://dukeworld.duk.../20101018-1715/

It includes code to print out a full backtrace in eduke32.crashlog when a crash happens (assuming you start the included eduke32.debug.exe or mapster32.debug.exe). People who experience crashes, please try it and attach the crashlog here. YMMV.

(code courtesy of http://code.google.c...cktrace-mingw/)

EDuke32 - "The corrupt doctrine of terror has begun."
0

User is online   The Commander 

  • Brown Fuzzy Fruit
  • 1,077

#28

http://pastebin.com/CuarFerP - When quitting my mod, can also happen on level restart or map change.

Posted Image
0

User is offline   Helixhorned 

  • EDuke32 Engine Maintainer
  • 610

#29

View PostPlagman, on Oct 19 2010, 05:43 AM, said:

Do you ever get symbol names from the engine, though? I only seem to get symbols from gameland.

I'm having no problem there. Cool that you got it working. Now the business of fixing crashes should be sped up a little!

(o~ Lunatic home: documentation, stand-alone LuaJIT, LunaCON and utilities ~o)
0

User is offline   Awesomebob 

  • 0

#30

Re-installed my ATI card, so now I can take advantage os this new feature. Here's your log.
LOG

Posted Image
0

Share this topic:


  • 2 Pages +
  • 1
  • 2
  • You cannot start a new topic
  • You cannot reply to this topic


Buy Duke3D from GOG.com!

All trademarks are property of their respective owners and all comments are owned by their posters. © 2004-2014 Duke4.net

Enter your sign in name and password


Sign in options