the program I am running have no problem performing socket operations when it first started up. After running the program and be doing task for prolong time, the program exits.
using Strace it state that the program quits on EBADF, I can see why its getting a EBADF. why its doing a close on a incomplete read (recvfrom)? I am using Linux Datagram.


there isn't a core dump file that I can use GDB to bebug.
I did run valgrind to check for memory problems.

what are the suggested ways to prevent this from happening?

It is not necessarily possible to prevent it from happening - that depends on what exactly causes the close.

If you want a coredump, you need to specify the maximum size in the shell before running the program:
ulimit -c 1k_byte_blocks_reserved_for_coredump

You don't need a coredump if you use GDB to debug it from the start.
ta0kira

I did set ulimit -c unlimited and I did not get any core dump.
I am running the code in a embedded system. Its is possible to do a remote GDB to debug the issue. problem is this issue is 90% reproducible.

anyone has ideas why its doing a close on a unfinished read?

1. why 'unlimited' rather than an actual number?

2. if the system is set up for remote debugging then you can use remote debugging; there are some extra things to do though, it doesn't happen magically.

3. no idea; the information you provided really doesn't tell us much so you'll have to trace through the program and look at what things might cause the failure and test which case did cause the failure. In general, when dealing with sockets you can expect them to close at any time; 'SIGPIPE' can also be generated. It is even possible for a memory management problem in your code to cause the failure. It really is up to you to debug this because it is not a problem which can be resolved with a fixed set of instructions.

+1

OP: Please list the code around, and including, the area where the recv() is being used. Perhaps it is possible that you are somehow smashing the stack where your file descriptor resides as you are reading data.

1) don't know what number to put, so had it unlimited. unlimited is a bad thing in linux?

2) I understand that perfectly, no hoping any thing to happen magically in this world. Not asking it to happen magically too! ^^;

3) Problem is I don't know why it happen. if a SIGPIPE was to happen will STRACE capture it? I am not aware that sockets are unreliable and I have expect it to close anytime unexpectly. I will take care of it. I am not asking guru here to help me debug the problem or magically resolve my issue. I should have phrase it better, I am asking if ppl experience similar incident what approach is taken to resolve it.
yes it might be possible its a memory issue. I will spend time adding self ptr to ensure no memory are trash.

thanx for all the tips.