Python 2.7.6 (default, Mar 22 2014, 22:59:56) 
>>> import os, stat
>>> os.stat('/')
posix.stat_result(st_mode=16877, st_ino=2, st_dev=64513L, st_nlink=29, st_uid=0, st_gid=0, st_size=4096, st_atime=1425341751, st_mtime=1424824650, st_ctime=1424824650)
>>> x =os.stat('/')
>>> x.st_mtime
1424824650.653934
>>> x[stat.ST_MTIME]
1424824650
>>> os.stat_result
<type 'posix.stat_result'>


I have also confirmed the same behavior in 3.4.2.

This may be intentional.  Tuple [stat.ST_XXX] access to the stat return value is the old legacy way to access these prior to Python 2.2 when the stat_result object was introduced.  At that point they were likely all ints.

The os.stat_float_times() API exists to change the behavior of stat_result objects to return ints instead of floats by default, but the behavior of the tuple indexed values is always ints.

We need to explicitly document this behavior difference.  Changing the legacy tuple indexing behavior to return a float would likely break code.

Why file this?  We just ran into a bug due to code comparing a [stat.ST_MTIME] value to a stored stat_result.st_mtime thus continually reporting a difference due to the precision difference between the two but the numeric comparison doing its job regardless.

It's already documented:

https://docs.python.org/dev/library/os.html#os.stat_result

"For compatibility with older Python versions, accessing stat_result as a tuple always returns integers."

I missed that because i was looking for it to be called out under 2.7 os.stat() docs rather than under 2.7's os.stat_float_times() which is a method nobody is likely to read the documentation for as floats have been the default since 2.5.

The 2.7 docs are much less organized around this than the 3.x.

Adding that note in the same paragraph on 2.7's os.stat() docs would be sufficient.

"For backward compatibility, the return value of stat() is also accessible as a tuple of at least 10 integers" implies it, but doesn't explicitly call out the integer [ST_XXX] vs float .st_xxx difference as adding that sentence does.

Python 2.7 is no longer supported.