Introduced in 10.5.0-devel by this commit.

50433ea526942dbeac75cc415becadea157e5d33 is the first bad commit
commit 50433ea526942dbeac75cc415becadea157e5d33
Author: Marek Olšák <marek.olsak@amd.com>
Date:   Sat Jan 31 17:15:16 2015 +0100

    gallium/util: add u_bit_scan64
    
    Same as u_bit_scan, but for uint64_t.
    
    Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
    Reviewed-by: Brian Paul <brianp@vmware.com>

:040000 040000 4eb1909ac455e690d7de4dbea87cb21b8678ed4b a1d1419d9af472fc2b366764bf8b2a772d11849d M      src
bisect run success

Hi Vinson,
Is that with gcc?  Which compiler version?

I'll attach a patch proposal for you to try...

Created attachment 113670 [details] [review]
proposed patch

If you're using gcc, maybe a gcc version check would be better.
This patch also assumes the radeonsi driver isn't used on NetBSD (the only user of u_bit_scan64() at this time.)

$ gcc --version
gcc (NetBSD nb2 20110806) 4.5.3
Copyright (C) 2010 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.


attachment 113670 [details] [review] fixes the NetBSD build.

Tested-by: Vinson Lee <vlee@freedesktop.org>

We now have:

#ifdef HAVE___BUILTIN_FFSLL
#define ffsll __builtin_ffsll
#elif defined(_MSC_VER) && (_M_AMD64 || _M_ARM || _M_IA64)
static inline int
ffsll(long long int i)
{
   unsigned long index;
   if (_BitScanForward64(&index, i))
      return index + 1;
   else
      return 0;
}
#else
extern int
ffsll(long long int val);
#endif

So I'm assuming this was fixed long ago and closing.