requires them. Disable executable bits and shebang lines in test and
benchmark files in order to prevent using a random system python, and in
source files of modules which don't provide command line interface.
svn+ssh://pythondev@svn.python.org/python/branches/py3k
........
r87550 | r.david.murray | 2010-12-28 13:54:13 -0500 (Tue, 28 Dec 2010) | 8 lines
#9824: encode , and ; in cookie values so that browsers don't split on them
There is a small chance of backward incompatibility here, but only for
non-SimpleCookie applications reading SimpleCookie generated cookies. Even
then, any such ap is likely to be handling escaped values already, and it would
take a fairly perverse implementation of unescaping to fail to unescape these
newly escaped chars, so the risk seems minimal.
........
svn+ssh://pythondev@svn.python.org/python/branches/py3k
........
r83393 | georg.brandl | 2010-08-01 10:35:29 +0200 (So, 01 Aug 2010) | 1 line
#1690103: fix initial namespace for code run with trace.main().
........
r83396 | georg.brandl | 2010-08-01 10:52:32 +0200 (So, 01 Aug 2010) | 1 line
#4810: document "--" option separator in timeit help.
........
r83398 | georg.brandl | 2010-08-01 11:06:34 +0200 (So, 01 Aug 2010) | 1 line
#8826: the "expires" attribute value is a date string with spaces, but apparently not all user-agents put it in quotes. Handle that as a special case.
........
r83405 | georg.brandl | 2010-08-01 16:38:17 +0200 (So, 01 Aug 2010) | 1 line
#4943: do not try to include drive letters (and colons) when looking for a probably module name.
........
r83408 | georg.brandl | 2010-08-01 17:30:56 +0200 (So, 01 Aug 2010) | 1 line
#5551: symbolic links never can be mount points. Fixes the fix for #1713.
........
The comment following used to say:
/* We use ~hash instead of hash, as degenerate hash functions, such
as for ints <sigh>, can have lots of leading zeros. It's not
really a performance risk, but better safe than sorry.
12-Dec-00 tim: so ~hash produces lots of leading ones instead --
what's the gain? */
That is, there was never a good reason for doing it. And to the contrary,
as explained on Python-Dev last December, it tended to make the *sum*
(i + incr) & mask (which is the first table index examined in case of
collison) the same "too often" across distinct hashes.
Changing to the simpler "i = hash & mask" reduced the number of string-dict
collisions (== # number of times we go around the lookup for-loop) from about
6 million to 5 million during a full run of the test suite (these are
approximate because the test suite does some random stuff from run to run).
The number of collisions in non-string dicts also decreased, but not as
dramatically.
Note that this may, for a given dict, change the order (wrt previous
releases) of entries exposed by .keys(), .values() and .items(). A number
of std tests suffered bogus failures as a result. For dicts keyed by
small ints, or (less so) by characters, the order is much more likely to be
in increasing order of key now; e.g.,
>>> d = {}
>>> for i in range(10):
... d[i] = i
...
>>> d
{0: 0, 1: 1, 2: 2, 3: 3, 4: 4, 5: 5, 6: 6, 7: 7, 8: 8, 9: 9}
>>>
Unfortunately. people may latch on to that in small examples and draw a
bogus conclusion.
test_support.py
Moved test_extcall's sortdict() into test_support, made it stronger,
and imported sortdict into other std tests that needed it.
test_unicode.py
Excluced cp875 from the "roundtrip over range(128)" test, because
cp875 doesn't have a well-defined inverse for unicode("?", "cp875").
See Python-Dev for excruciating details.
Cookie.py
Chaged various output functions to sort dicts before building
strings from them.
test_extcall
Fiddled the expected-result file. This remains sensitive to native
dict ordering, because, e.g., if there are multiple errors in a
keyword-arg dict (and test_extcall sets up many cases like that), the
specific error Python complains about first depends on native dict
ordering.
add a self-test using doctest. Results:
- The docstring needs to be a raw string because it uses \"...\".
- The oreo example was broken: the Set-Cookie output doesn't add
quotes around "doublestuff".
- I had to change the example that prints the class of a Cookie.Cookie
instance to avoid incorporating an arbitrary object address in the
test output.
Pretty good score for both doctest and the doc string, I'd say!
cookies that contain '=' as part of the value. This patch modifies
Cookie.py to allow '=' as a legal character, and to make the key
search nongreedy so it stops at the first '='.
added test script and expected output file as well
this closes patch 103297.
__all__ attributes will be added to other modules without first submitting
a patch, just adding the necessary line to the test script to verify
more-or-less correct implementation.
Serhiy Storchaka
Guido van Rossum
Ezio Melotti
Berker Peksag
Senthil Kumaran
R. David Murray
Georg Brandl
Senthil Kumaran
Benjamin Peterson
Tim Peters
Walter Dörwald
Andrew M. Kuchling
Raymond Hettinger
Fred Drake
Martin v. Löwis
Skip Montanaro