[fixed]3.8.0-beta1 OOS when playing windows/linux and using egypt

[tomreyn]

We played another game on r4596 today. I was hosting on atibox, Filux (Linux 64 bit), Jammy (Windows 64-bit) were clients. I was on one side of the map with a CPU Mega (on the same team) and  Filux and Jammy formed the second team on the other side of the map. Roughly 10 minutes into the game, the game reported for both Filux and me that Filux (the Linux -64 system, so same platform!) had run into a checksum mismatch. Right at this time both Filux, and less of a second later, Jammy, got disconnected off the server. Jammy had mentioned that the game was sluggish for him from the very start.

My logs
Jammy's logs
Filux's screenshot and logs

[tomreyn]

Now we did the same game setup again, but just Filux vs me (hosting), no Jammy, no CPU. And we got a mismatch:

Code: [Select]

<tomreyn> filux was reported to have disconnected the very first moment our fighters met
<tomreyn> i wonder whether we saw the same  series of events there, filux. what i saw / did was this: i had two fakirs who discovered that you were building a new main building on the top left of the map.
<tomreyn> initially only one of the fakirs attacked this structure which your worker was just working on.
<filux> yes, and exactly in the disconnection moment I started another build on the bottom too
<tomreyn> then i made both fakirs attack your worker directly
<tomreyn> where i had two more units lurking
<filux> in the first disconnection I produced some workers
<filux> then it can be related with main norsemen building
<tomreyn> if you were building very close to the bottom gold spot then my units there should have seen yours
<filux> or fighting :)
My logs
Filux's logs

[tomreyn]

r4598 today, I was hosting on atibox, MuWuM (Windows 7-64) was client, and had a CPU ultra team mate. Around the middle of this Prax game, he ran into a CRC mismatch during the second CPU attack wave (neither MuWuM nor I had "visited" the other side of this Conflict map, yet).

Linux server logs
Windows client logs

[tomreyn]

Another one, same setup as before, just the Windows client was Nig's (again W7-64) this time.

Linux server logs
Windows client log

[tomreyn]

r4601, atibox hosting, Carolinux' Linux system (not sure which architecture) was the client, he ran into a checksum mismatch around mid-game.

Server logs
Client logs

[softcoder]

The link to client logs is not working (404)

[tomreyn]

Hmm, that's right. Two problems there:
1. I provided the wrong URL, the correct one is http://www.sendspace.com/file/rzo4he
2. At the correct URL it doesn't work either: "The file has been deleted by the uploader and it cannot be restored. Please contact the sender and ask them to upload the file again."

We played another test game tonight, though.
I (atibox) was hosting, Lyra (linux-64) and Herbstvisage (Windows 7-64) were clients.

Tomreyn's server logs (includes terminal output)
Lyra's client logs (terminal output)
Herbstvisage's client logs

The checksum mismatch happened about mid-game on this player vs player game on a 3vs3 map. I was playing against the two, there was no CPU.
Lyra recieved a message that she ran into a CRC mismatch. Herbstvisage said he "also had the message" (I'm not sure whether he got a report that he ran into the mismatch or that Lyra did). For the server, a checksum mismatch was reported but I forgot for which player. Currently this is not written to the terminal (might be good to change this) but just reported on screen, so I'm afraid I cannot tell.

[tomreyn]

More games, more logs: We played two games and ran into checksum mismatches both times, both same and cross platform.

We, that's:
tomreyn (atibox) hosting
Lyra (Linux x86_64) - client
Jammy (Linux x86_64) - client
Herbstvisage (Windows 7-64) - client

Game 1:

• tomreyn
• Jammyjamjamman
• lyra
• Herbst

Game 2:

• tomreyn
• Herbst
• Jammyjamjamman
• lyra

[MuwuM]

r4606 today, tomreyn was hosting on atibox, I (MuwuM, Windows 7-64) was client, we where both in one Team and had 2 CPU enemies. At the first CPU attack wave, I ran into a CRC mismatch (neither tomreyn nor I had "visited" the other side of this map, yet).

CRC debug files:

tomreyn (atibox): http://tomreyn.megaglest.org/debugcrcworld_tomreyn_4696_20131015.tar.xz

MuwuM (win7 64 bit): http://glest.muwum.org/debugCRCWorld-2013-10-16.7z

[softcoder]

Updated svn with more info please try again and post the logs

[tomreyn]

Lyra and I just tested on 4609, I was hosting (atibox), she was on Windows 7-64, and ran into a CRC mismatch quite early.
Server logs
Client logs (terminal output)

Thanks!

[tomreyn]

This time, I tested with Derek on r4610.
Server logs
Client logs (shall be added by Derek when he's finished uploading.

[tomreyn]

Lyra and I tried r4612 today. Her Windows client (using the cudgellady snapshots) crashed when connecting to my host (running a full rebuild on atibox). We tried this once, I quit and restarted, and we tried again with the same effect.

Lyra's terminal outputs:
http://pastebin.com/raw.php?i=haz5NbNj
http://pastebin.com/raw.php?i=UynFEw48
http://pastebin.com/raw.php?i=07fWk4wM

Each time the Windows client crashed, two DMP files were written. Here's the stack traces for these:
http://megaglest.nopaste.dk/p63026
http://megaglest.nopaste.dk/p63027
http://megaglest.nopaste.dk/p63028
http://megaglest.nopaste.dk/p63029

[softcoder]

Should be fixed in svn now, sorry about that

[tomreyn]

r4616: CRC mismatch with four players. Linux-64 host (atibox), Lyra on Linux-64, Nig and CyberneticPony on Windows-64.

All logs: http://tomreyn.megaglest.org/debugcrcworld_tomreyn_20131019_1.tar.xz

Before this one, I played two x-platform games with a total of 2 human players each, and there were no problems.

[tomreyn]

I played a couple r4618 games today, some with Typedef (Linux 3.x-64), some with Nig (Windows 7-64). There were no checksum mismatches. However, we had serious problems with FPS dropping remarkably after the first third/half of the game. Both ends were then constantly playing at 1-3 FPS, i.e. unplayable. This happens whether or not we are using CRC checksumming, with different techtrees, different maps, different AI settings, different tilesets, and whether Nig or I were hosting.

Yesterday I took part in three games where this wasn't a problem, and where no mismatches occured either. But those were only 2 player games, like today. So it may be too early to tell whether everything is fixed. And large games are not possible with the current slowdowns.

[titi]

I switched all CRC/logging off now ( I was the server ) and I played 2 games with Jammyjamman ( rev

EnableNetworkGameSynchChecks=0
EnableNetworkGameSynchMonitor=0

It was very smooth and performat (rev 4641)
We just need to play a real big 8 player game now. Then we set this as default and we are nearly ready for a release. Just some translation needed. ( and tiny data changes which I don't want to do now )
 

[titi]

Yeah , I can mark this as fixed !!!  !