Determine a crash?

[Corthezz]

Hello,
I am working on a bot project for the WoW version 2.4.3 right now. From time to time the client is crashing without any way to prevent it.
Restarting wow.exe and picking up the previous state isnt a problem however I am stil thinking about the best possible way to detect the crash.

Right now I think about throwing in a state which looks for the WowError.exe but then again I would need to match the right error process to the right WoW instance (incase of multiple bots running).
Does anyone have any better way or a way to match the error process to the right WoW instance?

Thanks in advance for any helpful reply!

[Cypher]

Are you injected or 'external'?

EDIT: Also, if it's your bot which is causing the crash, the first thing I would do is fix the bug!

[Corthezz]

Are you injected or 'external'?

EDIT: Also, if it's your bot which is causing the crash, the first thing I would do is fix the bug!

Its external however I inject ASM using fasm.managed to detour and call some function. The problem is also happening from time to time without the bot being attached.
Right now I solved it with just closing all ErrorWow.exe processes and afterwards checking if a process with the current pid we are attached to exists. If not -> restart the WoW client and reattach to the new one.

[namreeb]

Are you saying that currently the client is crashing but you are not getting an error dump with a stack trace etc? If you are getting one, the stack trace would be helpful.

[Robske]

He mentioned using WowError to detect the crash, pretty sure an exception detail and stack-trace is available there.
As other people in the thread said, fix the bug instead of sweeping it under the rug.

[Cypher]

Its external however I inject ASM using fasm.managed to detour and call some function. The problem is also happening from time to time without the bot being attached.
Right now I solved it with just closing all ErrorWow.exe processes and afterwards checking if a process with the current pid we are attached to exists. If not -> restart the WoW client and reattach to the new one.

If you're injecting code you're not really external.

Anyway, you could probably detect this by attaching as a debugger if 'external' or using an unhandled exception filter if injected, but simply fixing the bug is definitely the best way forward.

I'd suggest using procdump or a debugger in order to catch the next crash, so you can actually investigate it. Using procdump would probably be better because then you can post the dump here for others to take a look (if anyone has the time).

It would probably also help if you told us what your injected thread(s) are doing, because there are some things which really need to be done from the main thread in order to work reliably (and will cause spurious crashes otherwise).

[Jadd]

The problem is also happening from time to time without the bot being attached.

Am I the only one that read this? Doesn't sound like a bug on his end.

[namreeb]

I read it, but I wasn't sure what it means. It sounds like perhaps he is injecting some code and running some out of process code. Perhaps he has 'attached' and made changes, then 'detached'.

[Corthezz]

What i actually do on attach is: Detour the EndScene to my codecave which executes all common functions I need (OnRightClick, AutoLoot, CTM etc) and beside from that I use ReadProcessMemory and WriteProcessmemory.

The error is also happening from day to day when the bot isnt even attached to the client (= Endscene not modified). It says something about NetClient. Sadly I dont have a crash log right now.

[Cypher]

I read it, but I wasn't sure what it means. It sounds like perhaps he is injecting some code and running some out of process code. Perhaps he has 'attached' and made changes, then 'detached'.

Yeah, I assumed the same thing.

What i actually do on attach is: Detour the EndScene to my codecave which executes all common functions I need (OnRightClick, AutoLoot, CTM etc) and beside from that I use ReadProcessMemory and WriteProcessmemory.

The error is also happening from day to day when the bot isnt even attached to the client (= Endscene not modified). It says something about NetClient. Sadly I dont have a crash log right now.

When you say that it's not attached, do you mean:
1. You attach, do some stuff, then detach.
2. You never attach.
3. Something else.

It's important to be clear on that point.

As far as getting a crash log, using procdump to get an actual dump file would be best, but even the wowerror crash log would probably help (though like I said, I'm not positive how much help we'll be able to be, it depends on the nature of the crash).

[Corthezz]

Yeah, I assumed the same thing.



When you say that it's not attached, do you mean:
1. You attach, do some stuff, then detach.
2. You never attach.
3. Something else.

It's important to be clear on that point.

As far as getting a crash log, using procdump to get an actual dump file would be best, but even the wowerror crash log would probably help (though like I said, I'm not positive how much help we'll be able to be, it depends on the nature of the crash).

Sure .
When I talk about attach I refer to my detours being applied and my memory class being set to write to the selected WoW process. I maybe should try to express a bit more clear here, sorry.

Well more about the problem: It happens on an altered (detours applied) aswell an unaltered WoW process. Sadly I cant provide any crash dump since this error is happening only like 1 time every 24 hours (but we all strive to perfection dont we ).
As soon as it happens again I will def. update the thread with a good log.

What I did for the moment:
Implemented a state checking if WowError.exe exists and if it does retrieve its parent pid. If the parent pid matches to the WoW instance my bot is working with I close the corresponding WowError.exe aswell the client itself and continue with opening a new client, attaching my detours etc.

That one was a good help: c# - How to get parent process in .NET in managed way - Stack Overflow
http://stackoverflow.com/questions/2...my-application

[Master674]

Sounds like you're writing to memory while it gets executed by a thread.
Make sure you're using threadsafe detours or only attach once and never detach again and if it crashes then meh... Let it crash and start again.

[Cypher]

Sure .
When I talk about attach I refer to my detours being applied and my memory class being set to write to the selected WoW process. I maybe should try to express a bit more clear here, sorry.

Well more about the problem: It happens on an altered (detours applied) aswell an unaltered WoW process. Sadly I cant provide any crash dump since this error is happening only like 1 time every 24 hours (but we all strive to perfection dont we ).
As soon as it happens again I will def. update the thread with a good log.

What I did for the moment:
Implemented a state checking if WowError.exe exists and if it does retrieve its parent pid. If the parent pid matches to the WoW instance my bot is working with I close the corresponding WowError.exe aswell the client itself and continue with opening a new client, attaching my detours etc.

That one was a good help: c# - How to get parent process in .NET in managed way - Stack Overflow
c# - How can I get the PID of the parent process of my application - Stack Overflow

So if you just play the game normally and never attach your bot, it will still crash sometimes? Or does it only happen when you've attatched before, then detatched? (Sorry to keep harping on this, but it's still unclear.)

[Jadd]

You can get crash logs from <WoW Directory>\Errors\.

[Corthezz]

Also when never attached to the client before

You can get crash logs from <WoW Directory>\Errors\.

I will look later.