[Support] Reimplement getMainExecutable() using sysctl on NetBSD

[mgorny]

Summary

Use sysctl() to implement getMainExecutable() on NetBSD, rather than
trying to guess the correct path from argv[0]. This is one
of the fixes to recent clang-check-mac-libcxx-fixed-compilation-db.cpp
test failure on NetBSD.

This has been historically done on both FreeBSD and NetBSD in r303015,
and reverted in r303285 due to buggy implementation on FreeBSD.
However, FWIK the NetBSD implementation does not suffer from the same
bugs and is more reliable than playing with argv[0].

[krytarowski]

lib/Support/Unix/Path.inc

178 ↗

(On Diff #182696)

Please include a comment with a rationale (hardlink issue) that this cannot be used on FreeBSD.

[mgorny]

Added FreeBSD warning comment.

[dim]

I can't vouch for the NetBSD part, but this definitely seems like a good idea to me. I'm unsure if the sysctl is already fixed on the FreeBSD side. @emaste any idea?

[joerg]

I'd really prefer to keep the argv[0] code as is. I'm not sure what that test case is supposed to do, but it seems quite questionable as "check" is not a valid language frontend nor a version suffix. It should not work.

[krytarowski]

In D56975#1364959, @joerg wrote:

I'd really prefer to keep the argv[0] code as is. I'm not sure what that test case is supposed to do, but it seems quite questionable as "check" is not a valid language frontend nor a version suffix. It should not work.

Related D56976. A broken program/test does not pass argv0.

[devnexen]

In D56975#1364533, @dim wrote:

I can't vouch for the NetBSD part, but this definitely seems like a good idea to me. I'm unsure if the sysctl is already fixed on the FreeBSD side. @emaste any idea?

Does not seem to be the case IMHO.

[mgorny]

Reverted in r355302. It seems that it is unreliable and sometimes evaluates to empty path, breaking clang.

[mgorny]

I'm not going to pursue this further.

llvm/trunk/lib/Support/Unix/Path.inc

 #endif
 
 #if defined(__NetBSD__) || defined(__DragonFly__) || defined(__GNU__)
 #define STATVFS_F_FLAG(vfs) (vfs).f_flag
 #else
 #define STATVFS_F_FLAG(vfs) (vfs).f_flags
 #endif
 
+#if defined(__NetBSD__)
+#include <sys/param.h>
+#include <sys/sysctl.h>
+#endif
+
 using namespace llvm;
 
 namespace llvm {
 namespace sys  {
 namespace fs {
 
 const file_t kInvalidFile = -1;
 
-#if defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__) ||     \
+#if defined(__FreeBSD__) || defined(__OpenBSD__) ||     \
     defined(__minix) || defined(__FreeBSD_kernel__) || defined(__linux__) ||   \
     defined(__CYGWIN__) || defined(__DragonFly__) || defined(_AIX) || defined(__GNU__)
 static int
 test_dir(char ret[PATH_MAX], const char *dir, const char *bin)
 {
   struct stat sb;
   char fullpath[PATH_MAX];
 
#if defined(__APPLE__)
   // binaries with symbols.
   char exe_path[MAXPATHLEN];
   uint32_t size = sizeof(exe_path);
   if (_NSGetExecutablePath(exe_path, &size) == 0) {
     char link_path[MAXPATHLEN];
     if (realpath(exe_path, link_path))
       return link_path;
   }
-#elif defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__) ||   \
+#elif defined(__NetBSD__)
+  // NB: sysctl() solution can't be used on FreeBSD since it may return
+  // a wrong path when a file is hardlinked in multiple locations.
+  // See r303285 for an earlier revert.
+
+  char exe_path[PATH_MAX];
+  int mib[4] = {CTL_KERN, KERN_PROC_ARGS, -1, KERN_PROC_PATHNAME};
+  size_t len = sizeof(exe_path);
+
+  if (sysctl(mib, 4, exe_path, &len, nullptr, 0) == 0)
+    return exe_path;
+#elif defined(__FreeBSD__) || defined(__OpenBSD__) ||   \
     defined(__minix) || defined(__DragonFly__) ||                              \
     defined(__FreeBSD_kernel__) || defined(_AIX)
   char exe_path[PATH_MAX];
 
   if (getprogpath(exe_path, argv0) != NULL)
     return exe_path;
 #elif defined(__linux__) || defined(__CYGWIN__)
   char exe_path[MAXPATHLEN];