http://buildbot.python.org/all/builders/PPC64%20AIX%203.x/builds/1138/steps/test/logs/stdio

======================================================================
ERROR: test_mktime (test.test_time.TimeTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/shager/cpython-buildarea/3.x.edelsohn-aix-ppc64/build/Lib/test/test_time.py", line 348, in test_mktime
    self.assertEqual(time.mktime(tt), t)
OverflowError: mktime argument out of range

======================================================================
FAIL: test_ctime (test.test_time.TimeTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/shager/cpython-buildarea/3.x.edelsohn-aix-ppc64/build/Lib/test/test_time.py", line 233, in test_ctime
    self.assertEqual(time.ctime(testval)[20:], str(year))
AssertionError: '1941' != '-100'
- 1941
+ -100

Oh, I missed also this one:



======================================================================
FAIL: test_mktime_error (test.test_time.TimeTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/shager/cpython-buildarea/3.x.edelsohn-aix-ppc64/build/Lib/test/test_time.py", line 365, in test_mktime_error
    self.assertEqual(time.strftime('%Z', tt), tzname)
AssertionError: 'LMT' != 'PST'
- LMT
+ PST

@David: Can you try to check what the minimum accepted timestamp for the time module?

The valid range is 00:00:00 UTC, January 1, 1970 to 03:14:07 UTC, January 19, 2038.

@David Edelsohn: Can you please test attached patch on AIX?

It should fix test_mktime and test_ctime.

Isn't this a duplicate of issue11188?

With the patch, the results are:

ERROR: test_ctime (test.test_time.TimeTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/dje/src/cpython/Lib/test/test_time.py", line 239, in test_ctime
    self.assertEqual(time.ctime(testval)[20:], str(year))
OverflowError: ctime() timestamp argument out of range

======================================================================
FAIL: test_mktime_error (test.test_time.TimeTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/dje/src/cpython/Lib/test/test_time.py", line 371, in test_mktime_error
    self.assertEqual(time.strftime('%Z', tt), tzname)
AssertionError: 'LMT' != 'PST'
- LMT
+ PST

which is what I would expect based on the patch and the new AIX-specific error message. What did you expect?

> which is what I would expect based on the patch and the new AIX-specific error message. What did you expect?

No, mktime() and ctime() should both raise OverflowError. It looks like mktime() doesn't fail. Can you please try the following examples on AIX?

Examples on Linux:

>>> time.mktime((-100, 1, 10) + (0,)*6)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: year out of range
>>> time.mktime((100, 1, 10) + (0,)*6)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: year out of range
>>> time.mktime((1900, 1, 10) + (0,)*6)
-2208211761.0
>>> time.mktime((1930, 1, 10) + (0,)*6)
-1261526400.0
>>> time.mktime((1969,12,31, 23,59,59, 0,0,0))
-3601.0

>>> time.mktime((-100, 1, 10) + (0,)*6)
-897577382.0
>>> time.mktime((100, 1, 10) + (0,)*6)
1118888922.0
>>> time.mktime((1900, 1, 10) + (0,)*6)
2086784896.0
>>> time.mktime((1930, 1, 10) + (0,)*6)
-1261497600.0
>>> time.mktime((1969,12,31, 23,59,59, 0,0,0))
28799.0

@David Edelsohn: Oh nice, mktime() has an integer overflow on AIX. Could you please try to apply attached mktime_aix.patch, run test_time and try again my examples of msg211616? Thanks.

With the latest patch, test_time passes.

>>> time.mktime((-100, 1, 10) + (0,)*6)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
OverflowError: mktime argument out of range
>>> time.mktime((-100, 1, 10) + (0,)*6)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
OverflowError: mktime argument out of range
>>> time.mktime((1900, 1, 10) + (0,)*6)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
OverflowError: mktime argument out of range
>>> time.mktime((1930, 1, 10) + (0,)*6)
-1261497600.0
>>> time.mktime((1969,12,31, 23,59,59, 0,0,0))
28799.0

New changeset 502c8b7e8ad2 by Victor Stinner in branch 'default':
Issue #19748: On AIX, time.mktime() now raises an OverflowError for year
http://hg.python.org/cpython/rev/502c8b7e8ad2

> With the latest patch, test_time passes.

Ok, thanks for your tests. But time.ctime() is still wrong. Can you please try the following examples with the latest Python version (502c8b7e8ad2) and then retry with pylocaltime_aix.patch?

>>> time.ctime(-2**29)
'Sat Dec 27 06:11:28 1952'
>>> time.ctime(2**29)
'Mon Jan  5 19:48:32 1987'

I would like to know if pylocaltime_aix.patch is needed or not.

WITHOUT the patch to timemodule.c:

>>> time.ctime(-2**29)
'Fri Dec 26 21:11:28 1952'
>>> time.ctime(2**29)
'Mon Jan  5 10:48:32 1987'

ERROR: test_mktime (test.test_time.TimeTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/dje/src/cpython/Lib/test/test_time.py", line 354, in test_mktime
    self.assertEqual(time.mktime(tt), t)
OverflowError: mktime argument out of range

WITH the patch:

>>> time.ctime(-2**29)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
OverflowError: ctime() timestamp argument out of range
>>> time.ctime(2**29)
'Mon Jan  5 10:48:32 1987'

OK (skipped=8)
1 test OK.

"""
WITHOUT the patch to timemodule.c:

>>> time.ctime(-2**29)
'Fri Dec 26 21:11:28 1952'
"""

Oh, in this case, pylocaltime_aix.patch is wrong. There was a bug in mktime(), but ctime() works fine.

"""
  File "/home/dje/src/cpython/Lib/test/test_time.py", line 354, in test_mktime
    self.assertEqual(time.mktime(tt), t)
OverflowError: mktime argument out of range
"""

Oh, I don't understand this one. On Linux I get:

>>> time.localtime(-2)
time.struct_time(tm_year=1970, tm_mon=1, tm_mday=1, tm_hour=0, tm_min=59, tm_sec=58, tm_wday=3, tm_yday=1, tm_isdst=0)
>>> time.localtime(-1)
time.struct_time(tm_year=1970, tm_mon=1, tm_mday=1, tm_hour=0, tm_min=59, tm_sec=59, tm_wday=3, tm_yday=1, tm_isdst=0)
>>> time.localtime(0)
time.struct_time(tm_year=1970, tm_mon=1, tm_mday=1, tm_hour=1, tm_min=0, tm_sec=0, tm_wday=3, tm_yday=1, tm_isdst=0)
>>> time.localtime(1)
time.struct_time(tm_year=1970, tm_mon=1, tm_mday=1, tm_hour=1, tm_min=0, tm_sec=1, tm_wday=3, tm_yday=1, tm_isdst=0)

And:

>>> time.mktime(time.localtime(-2)) 
-2.0
>>> time.mktime(time.localtime(-1)) 
-1.0
>>> time.mktime(time.localtime(0)) 
0.0
>>> time.mktime(time.localtime(1)) 
1.0

>>> time.localtime(-2)
time.struct_time(tm_year=1969, tm_mon=12, tm_mday=31, tm_hour=15, tm_min=59, tm_sec=58, tm_wday=2, tm_yday=365, tm_isdst=0)
>>> time.localtime(-1)
time.struct_time(tm_year=1969, tm_mon=12, tm_mday=31, tm_hour=15, tm_min=59, tm_sec=59, tm_wday=2, tm_yday=365, tm_isdst=0)
>>> time.localtime(0)
time.struct_time(tm_year=1969, tm_mon=12, tm_mday=31, tm_hour=16, tm_min=0, tm_sec=0, tm_wday=2, tm_yday=365, tm_isdst=0)
>>> time.localtime(1)
time.struct_time(tm_year=1969, tm_mon=12, tm_mday=31, tm_hour=16, tm_min=0, tm_sec=1, tm_wday=2, tm_yday=365, tm_isdst=0)
>>> time.mktime(time.localtime(-2))
-2.0
>>> time.mktime(time.localtime(-1))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
OverflowError: mktime argument out of range
>>> time.mktime(time.localtime(0))
0.0
>>> time.mktime(time.localtime(1))
1.0

New changeset 00e94e454813 by Victor Stinner in branch 'default':
Issue #11188, #19748: mktime() returns -1 on error. On Linux, the tm_wday field
http://hg.python.org/cpython/rev/00e94e454813

>>> time.mktime(time.localtime(-1))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
OverflowError: mktime argument out of range

Oh yes, I now remember it. Sorry, it was 3 years ago:
http://bugs.python.org/issue11188#msg128541

The problem is to detect invalid time tuple and support the timestamp value -1 (1 second before the UNIX epoch).

I modified the test to skip mktime(localtime(-1)) test on AIX: changeset 00e94e454813.

It would be nice to support this specific value, but I don't see how to implement it. It's maybe not possible.

I'm no more interested to work on this issue, I prefer to close it.