Chinaunix

标题: [HP-UX 11.0]Death of the server process [打印本页]

---

作者: grace_limi    时间: 2003-02-25 09:31
标题: [HP-UX 11.0]Death of the server process
[HP-UX 11.00]
Working in HP-UX 11.00 server, recently we encounter a strange problem that a process always dies surprisingly. Look at the following C codes in the process:

   signal( SIGALRM, timesup );      /* Setup a timer */
   alarm( 5 );
   if ((n = send( wall.sockfd, messages, len, 0 )) != len )
      printf ( \"%s: Cannot send to Wall %d - ERROR %d\\n\", FUNC, i+1, errno );
   else
      printf ( \"%s: Wall %d!\\n\", __FILE__, i+1 );
   alarm( 0 );

The codes are to send a simple message from our HP-UX server to a remote PC. The signal handler of SIGALRM timesup() is defined in the main() module.

To our surprise, as soon as the remote PC is restarted and the server process is about to \"send\", the entire process will die. However, in line with the source codes, the process will print the following log info:
     printf ( \"%s: Cannot send to Wall %d - ERROR %d\\n\", __FILE__, i+1, errno );
other than die.

Somebody estimates that maybe some fatal signals, e.g., SIGFPE, SIGSEGV, _SIGPIPE and so on were being delivered to the process at that moment. Nevertheless, I can not image there are those signals generated as soon as the remote PC is restarted.

Which expert can give me any idea on this issue?

---

欢迎光临 Chinaunix (http://bbs.chinaunix.net/)

Powered by Discuz! X3.2