PRPnet server bugs and barfing problem

[mdettweiler]

Originally Posted by MyDogBuster

Max, you might want to check to see if some of the candidates file is missing right around where Lennart's returned tests failed. Also look in the completed work file. They have to be somewhere. I think you'll find the server was looping because some of the file is gone.

Hmm...interesting, I hadn't thought of that. Okay, I'll check it and report back here with my findings.

[rogue]

Originally Posted by mdettweiler

Oddly enough, on numerous occasions in the log excerpt above, it seems that various clients connected to the server, grabbed the greeting, and exited. Mark, do you know why any clients would want to do that?

Also, there are a lot of candidates being returned that aren't found on the server; I'm not sure if this is something odd, or if there just happened to be a number of expired tests right around the time when I checked the server.

At any rate, though, at 13:36:53 GMT it received a connection from a client, then promptly froze. It remained that way until over an hour later when I had to -SIGKILL the server.

I have had spurious reports of the server expiring tests early, but haven't been able to confirm any of them. I haven't found anything yet in the code that looks suspicious.

I do not know why clients would just grab the greeting and exit. Normally they would get work first before getting the greeting.

It would be helpful if someone on Windows could reproduce the issue where the server is hanging as they could use Visual Studio to attach to the process and trace it through.

[mdettweiler]

Okay, I've checked the server, and I don't see any particularly noticeable gap in the prpserver.candidates file. Doesn't look like that's the problem in this particular case.

Regarding debugging the problem with Visual Studio: would I need to attach it to the server, or to the client? The server is running on Linux, so that would be rather difficult; and as for the client, there's no way of knowing which of the lots of clients on the server would be the culprit, so that would be difficult as well (and besides, I think most of Lennart's clients are on Linux).

Ian, are you running your server on Windows by chance?

[MyDogBuster]

Ian, are you running your server on Windows by chance?

I am, but I don't have Visual Studio.

Besides, my server has been running flawlessly for sometime now. Zero problems, and I have never seen any of my clients just sign on and only get the greeting. Sounds like a Linux problem to me.

Max, can you check to see if those candidates received from Lennart were in the original file loaded into the system. If so, then they either have to still be there, or they are in the completed work file. You should also be able to trace back in the log to see when those tests were sent to Lennart. Knowing all that will shed a whole bunch of light on the subject.

[rogue]

Originally Posted by mdettweiler

Okay, I've checked the server, and I don't see any particularly noticeable gap in the prpserver.candidates file. Doesn't look like that's the problem in this particular case.

Regarding debugging the problem with Visual Studio: would I need to attach it to the server, or to the client? The server is running on Linux, so that would be rather difficult; and as for the client, there's no way of knowing which of the lots of clients on the server would be the culprit, so that would be difficult as well (and besides, I think most of Lennart's clients are on Linux).

The one that seems to be stuck.

[mdettweiler]

Originally Posted by MyDogBuster

I am, but I don't have Visual Studio.

Besides, my server has been running flawlessly for sometime now. Zero problems, and I have never seen any of my clients just sign on and only get the greeting. Sounds like a Linux problem to me.

Max, can you check to see if those candidates received from Lennart were in the original file loaded into the system. If so, then they either have to still be there, or they are in the completed work file. You should also be able to trace back in the log to see when those tests were sent to Lennart. Knowing all that will shed a whole bunch of light on the subject.

All right, I'll check that.

Originally Posted by rogue

The one that seems to be stuck.

That would be the server (though the client could also be stuck, for all I know; of course, I have no way of knowing which of Lennart's machines it would be). BTW, would gdb work for debugging the Linux version instead of Visual Studio?

[Lennart]

Originally Posted by rogue

The one that seems to be stuck.

Last 4 days i have only had my clients running on Linux.

I have ~80 clients on Linux and 14 on Windows but no windows last 4 days.

Lennart

[gd_barnes]

Originally Posted by mdettweiler

BTW, would gdb work for debugging the Linux version instead of Visual Studio?

gdb?

Those are my initials. No, I wouldn't work for debugging the Linux version of Visual Studio.



Acronyms never cease to amaze me.

[rogue]

Originally Posted by mdettweiler

All right, I'll check that.

That would be the server (though the client could also be stuck, for all I know; of course, I have no way of knowing which of Lennart's machines it would be). BTW, would gdb work for debugging the Linux version instead of Visual Studio?

If you can attach gdb to a process, then yes. I believe that you can also force a core dump with the kill command. You can then use gdb on the core file that is generated.

[mdettweiler]

Originally Posted by rogue

If you can attach gdb to a process, then yes. I believe that you can also force a core dump with the kill command. You can then use gdb on the core file that is generated.

Could you possibly provide instructions for doing this? I don't know anything about how to use gdb, though I remember you had me do it once before for either a Phrot or PRPnet client executable (I forget exactly which).

@Gary: LOL. I believe it stands for "GNU Debugger", to go with gcc (GNU C Compiler) and g++ (GNU C++ Compiler).

[rogue]

Originally Posted by mdettweiler

Could you possibly provide instructions for doing this? I don't know anything about how to use gdb, though I remember you had me do it once before for either a Phrot or PRPnet client executable (I forget exactly which).

@Gary: LOL. I believe it stands for "GNU Debugger", to go with gcc (GNU C Compiler) and g++ (GNU C++ Compiler).

I believe that a kill -6 will create a core dump when it kills the process. I suggest you read the documentation for your OS or use "man kill" from the command prompt to see if it tells you.