select - should I use FD_SETSIZE or max fd

Hello group,

The man pages say that the first parameter to select (nfds) should be "an integer one more than the maximum of any file descriptor in any of the sets", but I often see the code that uses FD_SETSIZE like this:

Code:

---

  ret = select(FD_SETSIZE, ...);

---

instead of calculated max value like this:

Code:

---

  int nfds = 0;
  ...
  nfds = max(nfds, fd1);
  nfds = max(nfds, fd2);
  ..
  ret = select(nfds, ...);

---

Does this code (with FD_SETSIZE) work, or does it just accidentally work?

example with max(): http://www.linuxmanpages.com/man2/select_tut.2.php
example with FD_SETSIZE: http://www.delorie.com/gnu/docs/glibc/libc_248.html

Which one is correct?

Thanks,
Tvrtko

FD_SETSIZE seems like a lazy programmer cop-out which would cause extra work for the machine.

On my machine FD_SETSIZE equals 1024. What if file descriptor is greater than 1024 and I use FD_SETSIZE? It doesn't seem as lazy anymore, but as incorrect.

> What if file descriptor is greater than 1024 and I use FD_SETSIZE?
It can't be larger than FD_SETSIZE. The fd_set must be capable of representing every valid descriptor.

So if you're only checking a few descriptors, you're checking over 1000 other descriptors which can never satisfy the select call, but you still make the OS loop through them.

Thanks. I checked this by opening more than 1024 files and it failed after it returned file descriptor 1023. Now it makes sense because I assumed existence of descriptors greater than 1024 and generally random numbers but as it turns out it is not the case: "The file descriptor returned by a successful call [to open] will be the lowest-numbered file descriptor not currently open for the process."