[fixed] r4001, Win7-64: Fails with ZwGetContextThread!ntdll (null)(0) +18

[nig]

may be its all the same, my personal "niggybug" on windows.

its this "ZwGetContextThread!ntdll (null)(0) +18" again. This time the headless (urg) running 3.7.2 rev 4001 died somehow in the final of a long big 4 player game.

Errorlog

[2013-01-10 23:20:01] Runtime Error information:
======================================================
Unknown error!
Stack Trace:
00000000 00000000 00000000 d182862d ZwGetContextThread!ntdll (null)(0) +18
00000000 00000000 00000000 d182862d ZwGetContextThread!ntdll (null)(0) +18

everything was fine (may be controller diconnected, because it was the final undertaking on the screen), and suddenly my windows client (other machine but running win7 64 3.7.2. r4001 too) displayed:
"server disconnected"

screen of server: http://megaglest.pzt.me/4i3v

logs of server were destroyed by server restart through batch as I closed the "Megaglest.exe funktionoiert nicht mehr" - Window. So exit-option of headless-mode was still alive, although windows reportet died application.

shall we put all this "ZwGetContextThread!ntdll (null)(0) +18" on windows in one thread or is this a very common errormessage?

yes, i will try 4005

Greets

[nig]

megaglest v3.7.2-dev
Compiled using: VC++: 1600 on: Jan 11 2013 05:33:05 platform: Windows endianness : little SVN: [Rev: 4006] - using STREFLOP [SSE] - [no-denormals]

did the same (?) again today. This time i wasnt in there - but this time i have logs and console-output at: http://www.fastshare.org/download/debug1.zip

Greets

[ediit: forget headless log, crash seem to happen early in game:]

=========================================
Headless Server Current Game information:
=========================================
Network connection for index: 1
------------------------------
Connected: 1
Handshake received: 1
Connected at: 2013-01-11 19:38:00
Connection duration: 0 hours 1 minutes 0 seconds.
Player Index: 1
IP Address: 201.8.172.63
Player name: higo
Language: portuguese
Game Version: v3.7.1-VC++: 1600-Rev: 3948
Session id: 18014
Socket id: 416
Total Slot Count: 1
=========================================

[softcoder]

Please try svn rev#: 4012 to see if we get better stack traces for the windows build.

[nig]

huuh 4034 did it again on win7 64.

Iwas playing a scenario all alone...long...  near the end of the game there it is again: error message and the offer to quit. No collapse, no freeze.

logs and screens: http://www.fastshare.org/download/debug4.zip

Greets

[Omega]

huuh 4034 did it again on win7 64.

Iwas playing a scenario all alone...long...  near the end of the game there it is again: error message and the offer to quit. No collapse, no freeze.

logs and screens: http://www.fastshare.org/download/debug4.zip

Greets

Can you please not use that site to host your downloads? It prohibits downloading with an adblocker enabled, which is a complete no go for me. Try MediaFire instead.

[nig]

Mediafire requires an registration with valid email-adress and declares to be ad-supported. Thats an nogo for me. may be i find another one.

[Omega]

Mediafire requires an registration with valid email-adress and declares to be ad-supported. Thats an nogo for me. may be i find another one.

Let's be honest here, if it's a free host, it's probably ad supported. Google Search? Youtube? Facebook? Ads. The fact it needs an email sign-in is a bit unfortunate (since it didn't always need one), but on the plus side, that makes it easy to manage your files. You could alternatively try using Google Drive or Dropbox to share files (set it public and copy the public link). You'll still need an account, though. I'm not really familiar with any hosts that aren't evil and don't require an account.

[nig]

now we are complete OT again.

Google Search? Youtube? Facebook?

You are very familiar with the most evil hosts.

Ive got no prblems with Ads. I am not using an Ad-Blocker. Thats you! If Ads are okay, that system can only work if the users dont betray the hosters about the views and clicks on ads.

Internet should be usable without borders, should respect privacy and should be allowed to provide services for free, for share and for fun.
All this is in danger to disappear.

lets keep internet free for everyone, no matter who you are. Registration is the first border . Dont  build borders in the net. Weve got enough of them in the real world.

Greets

[nig]

4036 on Win7 64 today. I hosted a game and it broke with fatal error [main] after about half an hour or so.
game broke immediatly after it reported an frame synchronisation error for one player (see screens). Sorry, no debug.log.

network, console and screens...:
http://www.datafilehost.com/download-b69b59ce.html

Greets

[nig]

r4040 did it again.

logs, screen and console at http://www.datafilehost.com/download-2075b3fd.html

greets

[softcoder]

Try 4041

[nig]

4040 did it reproducable on two different machines, playing scenario Anarchy.

But: I had today several long network games hosted on headless linux 371 without any error.

Greets

[nig]

may be interesting:

the error in 4040 is reproducable : Just start scenario "Anarchy" and do completely nothing. After 15 min the error appears.
Debug is off.
Funny: If I change the tileset in "Anarchy.xml" to winter, everything is fine. An overflow from blocked workers or just an issue with openGL?
If I change back to "winter-forest" error appears again.

greets

[softcoder]

Give rev 4051 a try

[nig]

4060 : Anarchy crashed in the winter_forest again. This time I moved a bit.

logs: http://www.datafilehost.com/download-e895078c.html and greets

[titi]

For info:
nig has these problem on two different computers, both with nvidia cards and windows. One is a notebook (Nvidia 2000m ?), the other one a Desktop computer (Geforce 580 GTX ).
It just happens in 3.7.2-dev, NOT in 3.7.1 . He also said that the scenario behaves different  in 3.7.2-dev. This made me think it might have soemthing to do with the latest AI changes made in svn ?

It looks like the Problem has nothing to do with selection problems and happens with both, selectBuf amd colorPicking.

[softcoder]

Try 4061. I tried the same scenario but it does not crash for me. I added some more logging so lets see if we can track down this bug.

[tomreyn]

I forgot to report that I also tried to reproduce this on r4060 yesterday on a Windows 7 (32 bit) VM. I just started the Anarchy scenario and kept it running (in the background, without actually playing) for 25 minutes, but it did not crash. In case this issue is video related this test scenario does not help much, though (due to video drivers actually running on the Linux VirtualBox host).

[Coldfusionstorm]

What AI changes?, i noticed some revisions ago that the AI got much smarter, In my game it was acually able to build units. (something wich it have never done before), but this behavior recently changed.(so that the AI is now dumb again   ).

[nig]

r4064: Its still there.
I fumbled around with fog of war, changing teams to have a look to see what is happening in the other teams. But every thing i do to just see what all the units are doing is changing the game play and the error disappears.

So Im am nearly convinced its just a silly event in the game play which causes that error. The main thing is not, that anarchy is crashing when you do nothing for 15 minutes (or just one at speedloops 16). The thing is, that this event happens in other games and is probably triggering the same error again.

interesting is the dependency on tilesets. does the tileset matter for the decisions of AI?. In Anarchy it has to be a movement of AI that triggers that error - but only in tilset winter_forest.

Greets

I remember I had a similar error long time ago with "my island". I could reproduce that error in that time by doing the same very defensiv tactic (try to be not to be seen too early).
If you do this in those days there will come an initiate and will discover your base. After a short time there was then the error appearing regulary.

So it might be there for long time - but just happening rarly

[nig]

figured out, that you can look at the bug in a normal game, which settings are the same llike the sceanrio:


Download this image: http://www.datafilehost.com/download-003b6a36.html

I tried to find something unusual like big crowds or something blocked or whatever.
The error occures in the second attack of blue magic against yellow magic. yellow has lost its castle already by the green tech, and one battlemage of blue is attacking the final summoner guild, followed by some blue collegs. Game crashes befor they arrive and before summoner guild is finally down.

found nothing unusual until now, but:

Error does not occure when i load a saved game. I did the saving before and the game crashed like always. But not if i reload that save.
Error is independent of resolution of screen, but depends on the position of camera:
If I look on that fight described above at the yellow base, the error does not occure.
If  i look at green or blue base, the error occores.

That is all real magic

Edit (tomreyn): Image downloaded from file storage host, uploaded to image hosting site + embedded
Subsequent edit by Omega: The image host used by Tomreyn does not allow direct linking. Reuploaded to Imgur.

[titi]

In linux I get no crash and there is not really a second attack of blue attacking yellow.
If blue attacks yellow the second time only oone building is left amd green is in blues "town" .

Please go to options and set selectionType to "selectBuf" and try it again. Does this also happen with 
FastSpeedLoops=8
in glestuser.ini ?

[nig]

In linux I get no crash and there is not really a second attack of blue attacking yellow.
If blue attacks yellow the second time only oone building is left amd green is in blues "town" .

Yes, thats what I tried to describe. Same happens to me. The crash happens when the blue ghoul reaches the scene. A few secs before the last summoner guild of yellow dissappears

happens too at fastspeed=8, selectBuf was selected all the time

Greets

[softcoder]

I cannot reproduce this on Linux using the steps given above.

*Update: please show us debug.log as i log more information which may help find the bug.

Thanks

[nig]

logs of 4065, logged to HDD at speedloops=8, colourpicking:
http://www.datafilehost.com/download-65b75b46.html

logs of 4065, logged to HDD at speedloops=8, selectbufpicking:
http://www.datafilehost.com/download-70611b81.html

Greets