[SOLVED] Client crash while playing on Linux server with 267 build.

METROID

The owner of the Russian servers asked me to report this. You are free to request any additional info.

After the server crashes the server's console shows errors:
"Error: Exceeded maximum number of snapshots".
Here is a video of the bug reproduction on the standard settings of client and server:


Client: everyone on the server is affected.
Server: Ubuntu Server 14.04.1 LTS (64 bit), Intel Core i5-2550K, RAM 8 GB.

Crash circumstances:
1. Immediately after connecting to server (sometimes).
2. Immediately after choosing the team, or later after changing the team. Anyone can crash over the time, the chance of crash is rising upon changing the team. I never get a crash when choose the Alien's team and very often when choose Marine's team. Sometimes before I get the crash there is a small game freeze and a yellow plug shows up. The chance of the crash increases over the map change.
3. Before installing the 64-bit version on the server machine, I managed to crash server two times with "Segmentation fault (core dumped)" errors in the console.


You can test all by yourself:
Server name: CRASH
Server ip: 185.44.69.242:27015
Password: 1


=====================
MY PLOG
BIG tech_support.rar (~400Mb), taken while ns2.exe was running (game crashed when I tried to join Marines).

rkfg

Here's the client backtrace:

Program received signal SIGSEGV, Segmentation fault.
0x080753c4 in Read<unsigned char> (this=0xff8a7a84) at ../Source/Engine/BinaryReader.inl:31
31 ../Source/Engine/BinaryReader.inl: No such file or directory.
(gdb) bt
#0 0x080753c4 in Read<unsigned char> (this=0xff8a7a84) at ../Source/Engine/BinaryReader.inl:31
#1 ReadUInt8 (this=0xff8a7a84) at ../Source/Engine/BinaryReader.inl:79
#2 M4::ClientGame::OnPacket (this=0xea448008, packet=...) at ../Source/Spark_Client/ClientGame.cpp:1442
#3 0x08075868 in M4::ClientGame::ServiceNetwork (this=0xea448008) at ../Source/Spark_Client/ClientGame.cpp:1403
#4 0x080793a3 in M4::ClientGame::Update (this=0xea448008, deltaTime=0.0050111990003642859) at ../Source/Spark_Client/ClientGame.cpp:832
#5 0x0806d356 in M4::ClientEngine::Update (this=0xee46a170, deltaTime=0.0050111990003642859) at ../Source/Spark_Client/ClientEngine.cpp:168
#6 0x080c5574 in M4::BaseEngine::RunFrame (this=0xee46a170) at ../Source/Spark_Server/BaseEngine.cpp:201
#7 0x080c55a0 in M4::BaseEngine::Run (this=0xee46a170) at ../Source/Spark_Server/BaseEngine.cpp:165
#8 0x0806b663 in main (argc=<error reading variable: Cannot access memory at address 0x0>, argv=<error reading variable: Cannot access memory at address 0x4>)
at ../Source/ShooterExe/Main.cpp:82

It's not Linux specific, Windows clients crash as well.

GhoulofGSG9

calling @matso into this.

Thanks guys for the research and very detailed post

METROID

@rkfg, it's about crashes that clients are experiencing on Linux server, they can use Windows or Linux clients. Also Linux server sometimes crashes as well.

rkfg

@METROID‌, I know, just wanted to clarify since my backtrace is for the linux client. Also, I want to point out that it wasn't an issue on build 266. At first we thought it's client that's broken but then tried other servers and everything was smooth (except, you know, tips and texture memory settings that crash everyone). Seems like something with protocol, and it's very low-level.

METROID

Updated my first post by adding Plog and Tech_support files.

IronHorse

Thanks!

matso

Found the reason for the crashes; fixed.

Thanks for the info!

METROID

O great wizard @matso, thank you!!! But... we can expect the fix in 268 build or what...

matso

Yea, we debated going for a hotfix - but a hotfix takes about 4 working days to do, which is a significant amount of time compared to when we hope to release 268.

So 268 it is.

METROID