Programming

catch signal SIGPIPE ,print errno but it's value equal to 2(ENOENT)

#define ENOENT 2 /* No such file or directory */

is it should be EPIPE ?
#define EPIPE 32 /* Broken pipe */

Thanks !

1) Any system calls between the SIGPIPE signal and what you use to print the message can change the value of errno.

2) I'm not sure it's valid yet inside the signal handler itself, either, errno may be set after it!

Usually, for sigpipe, you have an empty do-nothing signal handler, and handle EPIPE when the next read/write operation fails.

My signal handler just simple as this . no change of errno value .

Code:

signal_handl(int ){
    printf("errno %d ",errno);
}

I suppose errno's just not being set until the signal handler returns, then. Which makes sense when I think about it since the system call getting SIGPIPE hasn't returned yet either!

Never call printf in a signal handler by the way, it's not signal-safe. This means it will work much of the time but will crash in odd unpredictable circumstances for no reason your debugger will be able to guess. fprintf(stderr, "..."); is nonbuffered, hence a bit safer, but still not guaranteed! Use lower system calls like write().

Actually, to be (overly?) pedantic, it almost certainly won't crash but it will deadlock your app if the signal is delivered to a thread that holds any lock that the printf() call might need, such as for heap allocations, which printf() tends to do.

And fprintf( stderr, ... ) is just as likely to need those locks, so it's not any safer.

Although for practical purposes, in this case it's pretty difficult to get a SIGFPE delivered to a thread that's already deep in libc doing things like malloc() or free().

So it's nothing to do with the file-handles, and everything to do with memory allocation? That's unfortunate. You couldn't even call sprintf then...

Here is a really good detailed explanation of aynchronous-safe from cert.org in the secure coding section:
https://www.securecoding.cert.org/co...ignal+handlers