Problems with sim_sock in forthcoming FreeBSD 15

[rms47]

Bob Eager wrote that an upcoming version of FreeBSD produces this compilation error:

sim_sock.c:1257:16: error: use of undeclared identifier 'SD_BOTH'
1257 | shutdown(sock, SD_BOTH);
| ^
1 error generated.

The problem stems from this macro definition near the top of "sim_sock.c":

#if defined(SHUT_RDWR) && !defined(SD_BOTH)
#define SD_BOTH SHUT_RDWR
#endif

He's chased it down to a change in how SD_RDWR is defined. He writes:

This is a build on FreeBSD 15 (still in development), so I had to spin
up a VM with that running to see the problem.

In FreeBSD 15, they have replaced (in socket.h), this:

---

/*

• howto arguments for shutdown(2), specified by Posix.1g.
  /
  #define SHUT_RD 0 / shut down the reading side /
  #define SHUT_WR 1 / shut down the writing side /
  #define SHUT_RDWR 2 / shut down both sides */

---

with this:

---

/*

• howto arguments for shutdown(2), specified by Posix.1g.
  /
  enum shutdown_how {
  SHUT_RD = 0, / shut down the reading side /
  SHUT_WR, / shut down the writing side /
  SHUT_RDWR / shut down both sides */
  };

---

So the test for 'defined' doesn't work any more.

Conditional hacking is beyond my pay grade, but I do know that enums won't work in #if defined statements.

[pkoning2]

I wonder what they thought they were doing.

In any case, the simplest workaround would be to remove the first clause, so it reads simply #if !defined(SD_BOTH).

[rms47]

Yes, I agree. SD_BOTH is in Win32, SD_RDWR in Posix, so that about covers the gamut.

[bscottm]

I wonder what they thought they were doing.

Type safety is the usual reason for replacing a list of #define-s with an enum. Static analyzers catch more issues with enum-s than with #defines-s.