[sanitizer] Use pthread_threadid_np as thread ID on OS X

[kubamracek]

Summary

Let's use pthread_threadid_np which returns a more reasonable ID than pthread_self (which is actually a stack pointer). The numbers from pthread_threadid_np are already used in other tools, e.g. in LLDB, and often appear in logs, so it's much more useful than pthread_self.

[dvyukov]

lib/sanitizer_common/sanitizer_mac.cc

233 ↗

(On Diff #53204)

Is it OK that the value is potentially truncated?

[kubamracek]

lib/sanitizer_common/sanitizer_mac.cc

233 ↗

(On Diff #53204)

Hm. It's really just an incrementing thread counter in the kernel. Truncating this can only be an issue after the system created 2^32 threads. This sounds unlikely.

On the other hand, would it be a problem to change GetTid to return u64?

[dvyukov]

lib/sanitizer_common/sanitizer_mac.cc

233 ↗

(On Diff #53204)

Yeah, it's probably unlikely on a 32-bit system. And I think we use it only for reporting, so a truncated id should not break anything. Up to you.

compiler-rt/trunk/lib/sanitizer_common/sanitizer_mac.cc

bool FileExists(const char *filename) {
   struct stat st;
   if (stat(filename, &st))
     return false;
   // Sanity check: filename is a regular file.
   return S_ISREG(st.st_mode);
 }
 
 uptr GetTid() {
-  return reinterpret_cast<uptr>(pthread_self());
+  // FIXME: This can potentially get truncated on 32-bit, where uptr is 4 bytes.
+  uint64_t tid;
+  pthread_threadid_np(nullptr, &tid);
+  return tid;
 }
 
 void GetThreadStackTopAndBottom(bool at_initialization, uptr *stack_top,
                                 uptr *stack_bottom) {
   CHECK(stack_top);
   CHECK(stack_bottom);
   uptr stacksize = pthread_get_stacksize_np(pthread_self());
   // pthread_get_stacksize_np() returns an incorrect stack size for the main