Windows doesn't accept negative timestamps (stated in the comment), yet checks is made against os.name instead of sys.platform.

patch fixes that, and also enables windows ce to pass on this test.

I think this is better than having os.name in ("nt", "ce").

This looks fine to me.  Alexander?

Mark,

I have zero experience with Windows and don't even have a win32 machine to test the patch.

On the other hand the patch is so simple that I think it can be reviewed based on theoretical considerations.

This is probably bikesheding, but I have a slight preference for  os.name in ("nt", "ce").  The reason is that sys.platform is fixed when python is built while os.name is (in theory) determined at run-time.

Also, sys.platform == "win32", appears to be false on 64 bit Windows, but I think it is actually true.

Finally, explicit better than implicit.  A change from if os.name == "nt" to  os.name in ("nt", "ce") gives obviously a strictly wider check. On the other hand it is not obvious to me how the current patch will affect Cygwin platform.

sys.platform will be "win32" for both 32 and 64-bit Windows.

As for Cygwin, os.name is "posix" there, and sys.platform is "cygwin", so it should be unaffected.

The patch looks fine to me, and we do typically use sys.platform more often than the os.name check.

OK, I'll commit it then.

Committed in r81555 (trunk) and r81556 (py3k).  Is this a 2.6 backport candidate?  I don't think so.

Leaving this open to consider using newer unittest.skipIf mechanism.  See attached patch, issue7879.diff.

The skipIf patch looks good to me (though I haven't tested it).

From IRC:

Taggnostr: imho tests and doc updates can be backported

SkipIf patch committed in r81559 (trunk) and r81560 (py3k).

- 26backport committed in r81618.
- merged to release31-maint in r81619.

The skipIf patch blocked from release26-maint (skipIf is new in 2.7) and merged into release31-maint in r81620.