datetime.fromtimestamp fails with negative fractional times

[jamesh]

BPO	1646728
Nosy	@gvanrossum, @birkenfeld
Files	datetime.patch: patch + unittest

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2007-03-07.16:13:08.000>
created_at = <Date 2007-01-29.02:21:31.000>
labels = ['library']
title = 'datetime.fromtimestamp fails with negative fractional times'
updated_at = <Date 2007-03-07.16:13:08.000>
user = 'https://bugs.python.org/jamesh'

bugs.python.org fields:

activity = <Date 2007-03-07.16:13:08.000>
actor = 'georg.brandl'
assignee = 'none'
closed = True
closed_date = None
closer = None
components = ['Library (Lib)']
creation = <Date 2007-01-29.02:21:31.000>
creator = 'jamesh'
dependencies = []
files = ['2287']
hgrepos = []
issue_num = 1646728
keywords = []
message_count = 13.0
messages = ['31113', '31114', '31115', '31116', '31117', '31118', '31119', '31120', '31121', '31122', '31123', '31124', '31125']
nosy_count = 4.0
nosy_names = ['gvanrossum', 'georg.brandl', 'jamesh', 'ocean-city']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = None
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue1646728'
versions = ['Python 2.5']

[jamesh]

The datetime.fromtimestamp() function works fine with integer timestamps and positive fractional timestamps, but fails if I pass a negative fractional timestamp. For example:

>>> import datetime
>>> datetime.datetime.fromtimestamp(-1.05)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: microsecond must be in 0..999999

It should return the same result as datetime.fromtimestamp(-1) - timedelta(seconds=.5).

The same bug can be triggered in datetime.utcfromtimestamp().

I have been able to reproduce this bug in Python 2.4.4 and Python 2.5 on Linux.

[gvanrossum]

Looks like a bug in the conversion from floats to ints. Anyone care to track it down more precisely?

[jamesh]

The problem seems to be in datetime_from_timestamp() from datetimemodule.c. It should probably be checking to see whether the microseconds value it calculates is negative, and adjust "timet" and "us" accordingly if so.

[gvanrossum]

Attached is a fix. If this is to your liking I'll check it in.
File Added: datetime.patch

[jamesh]

I just tried the patch, and can confirm that it fixes the problem with datetime.fromtimestamp() and datetime.utcfromtimestamp(). The logic in the patch looks correct.

[gvanrossum]

Committed revision 54167.

I'm leaving this open until it's been backported to the 2.5 branch.

[gvanrossum]

Georgbot backported this to 2.5.

[birkenfeld]

Though, the new tests seem to fail on Windows (I noticed that only after backporting, since most other buildbot failures were due to the cmp/key problem in setup.py).

[gvanrossum]

That's too bad. More details?

[birkenfeld]

Not from me, no Windows around.

[ocean-city]

Hello, I'm user of Windows (Now building Python2.5 with VC6)
I heard localtime() can only handle positive time_t on windows,
so datetime.fromtimestamp() also fails for negative value.

>>> datetime.datetime.fromtimestamp(-1.05)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: timestamp out of range for platform localtime()/gmtime() function
>>> datetime.datetime.fromtimestamp(-1)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: timestamp out of range for platform localtime()/gmtime() function

I'll attach workaround for unittest. Probably there is better way
skip this test on non-negative platform though :-)

Index: Lib/test/test_datetime.py
===================================================================

--- Lib/test/test_datetime.py	(revision 54194)
+++ Lib/test/test_datetime.py	(working copy)
@@ -1428,9 +1428,17 @@
     def test_negative_float_fromtimestamp(self):
         # The result is tz-dependent; at least test that this doesn't
         # fail (like it did before bug 1646728 was fixed).
+        try:
+            self.theclass.fromtimestamp(-1)
+        except ValueError: # cannot handle negative value
+            return
         self.theclass.fromtimestamp(-1.05)
 
     def test_negative_float_utcfromtimestamp(self):
+        try:
+            self.theclass.utcfromtimestamp(-1)
+        except ValueError: # cannot handle negative value
+            return
         d = self.theclass.utcfromtimestamp(-1.05)
         self.assertEquals(d, self.theclass(1969, 12, 31, 23, 59, 58, 950000))

[gvanrossum]

Thanks! I'm skipping these tests on Windows now.
Committed revision 54209.

Georgbot, would you be so kind... :-)

[birkenfeld]

Certainly. Backported in rev. 54211.