The patch as written causes buffer overflow for year >= 10,000:

>>> len(time.asctime( (10000, 1, 1, 0, 0, 0, 0, 1, -1)))
26
>>> len(time.asctime( (100000, 1, 1, 0, 0, 0, 0, 1, -1)))
27

while the buffer is only 26 characters:

+       static char result[26];
+
+       sprintf(result, "%.3s %.3s%3d %.2d:%.2d:%.2d %d\n",

This can be fixed in multiple ways: changing the year format to %.4d, using PyString_Format, or restricting the year to 4 decimal digits in check_bounds.

A nit pick: you can save some static storage by making wday_name and mon_name and possibly increase performance of asctime 2d arrays instead of arrays of pointers to null-terminated strings.  See http://www.opengroup.org/onlinepubs/009695399/functions/asctime.html .

Just as Martin, I am split on whether the patch is correct.  The fact that it is almost a copy of POSIX reference implementation gives some confidence, but that confidence is taken away by the reference implementation having a buffer overflow bug.

I am also not sure that all systems produce the same end of line character.  I would like to hear from Windows experts.