The test failed in the build 632 (Wed Jul 26 10:47:01 2017) for the first time.

http://buildbot.python.org/all/builders/AMD64%20FreeBSD%20CURRENT%20Debug%203.x/builds/632/steps/test/logs/stdio

== CPython 3.7.0a0 (heads/master:ede9084476, Jul 26 2017, 20:49:33) [GCC 4.2.1 Compatible FreeBSD Clang 5.0.0 (trunk 308421)]
== FreeBSD-12.0-CURRENT-amd64-64bit-ELF little-endian
== hash algorithm: siphash24 64bit
== cwd: /usr/home/buildbot/python/3.x.koobs-freebsd-current/build/build/test_python_23542
== CPU count: 2
== encodings: locale=US-ASCII, FS=ascii
Testing with flags: sys.flags(debug=0, inspect=0, interactive=0, optimize=0, dont_write_bytecode=0, no_user_site=0, no_site=0, ignore_environment=1, verbose=0, bytes_warning=2, quiet=0, hash_randomization=1, isolated=0)
Using random seed 9341910
...
======================================================================
FAIL: test_makedev (test.test_posix.PosixTester)
----------------------------------------------------------------------

Traceback (most recent call last):
  File "/usr/home/buildbot/python/3.x.koobs-freebsd-current/build/Lib/test/test_posix.py", line 543, in test_makedev
    self.assertEqual(posix.makedev(major, minor), dev)
AssertionError: 1292118443 != 954774858155

The test fails on Debug but also Non-Debug buildbots, master and 3.6 branches. It looks more like a change on the buildbot, maybe a FreeBSD upgrade?

I'm able to reproduce the bug on koobs's FreeBSD CURRENT:

• stat() returns st_dev = 0xde4d0429ab
• major(0xde4d0429ab) returns 0x29
• minor(0xde4d0429ab) returns 0x4d0400ab

minor() truncates most significant bits. major/minor are defined in sys/types.h:
https://github.com/freebsd/freebsd/blob/master/sys/sys/types.h#L372

#define	major(x)	((int)(((u_int)(x) >> 8)&0xff))	/* major number */
#define	minor(x) ((int)((x)&0xffff00ff)) /* minor number */

The definition of minor() confirms that most significant bits are truncated by "& 0xffff00ff".

I'm surprised that stat().st_dev returns a device larger than INT_MAX: 0xde4d0429ab.

os.stat() seems correct, st_dev and st_ino are the same than the system command "stat":

CURRENT-amd64% ./python -c 'import os; st=os.stat("setup.py"); print(st)'
os.stat_result(st_mode=33188, st_ino=2384528, st_dev=954774858155, st_nlink=1, st_uid=1003, st_gid=1003, st_size=99944, st_atime=1501162007, st_mtime=1501162007, st_ctime=1501162007)

CURRENT-amd64% stat setup.py
954774858155 2384528 -rw-r--r-- 1 haypo haypo 18446744073709551615 99944 "Jul 27 23:26:47 2017" "Jul 27 23:26:47 2017" "Jul 27 23:26:47 2017" "Jul 27 23:26:47 2017" 100352 79 0x800 setup.py

The used filesystem for /home is ZFS:

CURRENT-amd64% df .
Filesystem 1K-blocks Used Avail Capacity Mounted on
storage/home 17041752 11205517 5836235 66% /usr/home

CURRENT-amd64% mount|grep home
storage/home on /usr/home (zfs, local, noatime, nfsv4acls)

The problem is more the major(), minor() and makedev() commands:

CURRENT-amd64% ./python       
Python 3.7.0a0 (heads/master:d5ed47dea2, Jul 28 2017, 00:10:54) 
[GCC 4.2.1 Compatible FreeBSD Clang 5.0.0 (trunk 308421)] on freebsd12
Type "help", "copyright", "credits" or "license" for more information.
>>> import os
>>> dev=954774858155; os.makedev(os.major(dev), os.minor(dev))
1292118443

"makedev, major, minor -- device number conversion" doc on FreeBSD CURRENT:

https://www.freebsd.org/cgi/man.cgi?query=makedev&apropos=0&sektion=0&manpath=FreeBSD+12-current&arch=default&format=html

SYNOPSIS
     #include <sys/types.h>

 dev_t
 makedev(int major,	int minor);

     int
     major(dev_t dev);

     int
     minor(dev_t dev);

DESCRIPTION
The makedev() macro allows a unique device number to be generated based
on its major and minor number. The major() and minor() macros can be
used to obtain the original numbers from the device number dev.

 In	previous implementations of FreeBSD all	block and character devices
 were uniquely identified by a pair	of major and minor numbers.  The major
 number referred to	a certain device class (e.g. disks, TTYs) while	the
 minor number identified an	instance within	the device class.  Later ver-
 sions of FreeBSD automatically generate a unique device number for	each
 character device visible in /dev/.	 These numbers are not divided in
 device classes.

 On	FreeBSD	these macros are only used by utilities	that need to exchange
 numbers with other	operating systems that may use different encodings for
 dev_t, but	also applications that present these numbers to	the user in a
 more conventional way.

RETURN VALUES
The major() macro returns a device major number that has a value between
0 and 255. The minor() macro returns a device minor number whose value
can span the complete range of an int.

Confirmation that minor() truncates high bits:

CURRENT-amd64% ./python -c 'import os; print(hex(os.minor(0xaabbccddff)))'
0xbbcc00ff

At May 23, the dev_t type changed from 32 bits to 64 bits on FreeBSD in the kernel, but minor() wasn't updated. I reported a bug to FreeBSD:
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=221048

New changeset 12953ff by Victor Stinner in branch 'master':
bpo-31044: Skip test_posix.test_makedev() on FreeBSD (bpo-2915)
12953ff

New changeset 54cb340 by Victor Stinner in branch '3.6':
bpo-31044: Skip test_posix.test_makedev() on FreeBSD (bpo-2915) (bpo-2917)
54cb340

New changeset c2f7fb6 by Victor Stinner in branch '2.7':
[2.7] bpo-31044: Skip test_posix.test_makedev() on FreeBSD (bpo-2915) (bpo-2918)
c2f7fb6

New changeset b5d702e by Victor Stinner in branch 'master':
bpo-31044, test_posix: Reenable makedev() tests on FreeBSD (bpo-7449)
b5d702e

I enabled again the test in the master branch, the FreeBSD kernel bug has been fixed 10 months ago:

• https://svnweb.freebsd.org/base?view=revision&revision=321920
• https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=221048

Technically, we could reenable the test in 3.7 as well, but I prefer to leave it disabled since I'm not 100% sure that all FreeBSD kernels in the wild are fine, and the code is already well tested by other platforms.