cpython/Objects/listsort.txt

803 lines
38 KiB
Plaintext
Raw Normal View History

Intro
-----
This describes an adaptive, stable, natural mergesort, modestly called
timsort (hey, I earned it <wink>). It has supernatural performance on many
kinds of partially ordered arrays (less than lg(N!) comparisons needed, and
as few as N-1), yet as fast as Python's previous highly tuned samplesort
hybrid on random arrays.
In a nutshell, the main routine marches over the array once, left to right,
alternately identifying the next run, then merging it into the previous
runs "intelligently". Everything else is complication for speed, and some
hard-won measure of memory efficiency.
Comparison with Python's Samplesort Hybrid
------------------------------------------
+ timsort can require a temp array containing as many as N//2 pointers,
which means as many as 2*N extra bytes on 32-bit boxes. It can be
expected to require a temp array this large when sorting random data; on
data with significant structure, it may get away without using any extra
heap memory. This appears to be the strongest argument against it, but
compared to the size of an object, 2 temp bytes worst-case (also expected-
case for random data) doesn't scare me much.
It turns out that Perl is moving to a stable mergesort, and the code for
that appears always to require a temp array with room for at least N
pointers. (Note that I wouldn't want to do that even if space weren't an
issue; I believe its efforts at memory frugality also save timsort
significant pointer-copying costs, and allow it to have a smaller working
set.)
+ Across about four hours of generating random arrays, and sorting them
under both methods, samplesort required about 1.5% more comparisons
(the program is at the end of this file).
+ In real life, this may be faster or slower on random arrays than
samplesort was, depending on platform quirks. Since it does fewer
comparisons on average, it can be expected to do better the more
expensive a comparison function is. OTOH, it does more data movement
(pointer copying) than samplesort, and that may negate its small
comparison advantage (depending on platform quirks) unless comparison
is very expensive.
+ On arrays with many kinds of pre-existing order, this blows samplesort out
of the water. It's significantly faster than samplesort even on some
cases samplesort was special-casing the snot out of. I believe that lists
very often do have exploitable partial order in real life, and this is the
strongest argument in favor of timsort (indeed, samplesort's special cases
for extreme partial order are appreciated by real users, and timsort goes
much deeper than those, in particular naturally covering every case where
someone has suggested "and it would be cool if list.sort() had a special
case for this too ... and for that ...").
+ Here are exact comparison counts across all the tests in sortperf.py,
when run with arguments "15 20 1".
Column Key:
*sort: random data
\sort: descending data
/sort: ascending data
3sort: ascending, then 3 random exchanges
+sort: ascending, then 10 random at the end
%sort: ascending, then randomly replace 1% of elements w/ random values
~sort: many duplicates
=sort: all equal
!sort: worst case scenario
First the trivial cases, trivial for samplesort because it special-cased
them, and trivial for timsort because it naturally works on runs. Within
an "n" block, the first line gives the # of compares done by samplesort,
the second line by timsort, and the third line is the percentage by
which the samplesort count exceeds the timsort count:
n \sort /sort =sort
------- ------ ------ ------
32768 32768 32767 32767 samplesort
32767 32767 32767 timsort
0.00% 0.00% 0.00% (samplesort - timsort) / timsort
65536 65536 65535 65535
65535 65535 65535
0.00% 0.00% 0.00%
131072 131072 131071 131071
131071 131071 131071
0.00% 0.00% 0.00%
262144 262144 262143 262143
262143 262143 262143
0.00% 0.00% 0.00%
524288 524288 524287 524287
524287 524287 524287
0.00% 0.00% 0.00%
1048576 1048576 1048575 1048575
1048575 1048575 1048575
0.00% 0.00% 0.00%
The algorithms are effectively identical in these cases, except that
timsort does one less compare in \sort.
Now for the more interesting cases. Where lg(x) is the logarithm of x to
the base 2 (e.g., lg(8)=3), lg(n!) is the information-theoretic limit for
the best any comparison-based sorting algorithm can do on average (across
all permutations). When a method gets significantly below that, it's
either astronomically lucky, or is finding exploitable structure in the
data.
1. Combined the base and length arrays into a single array of structs. This is friendlier for caches. 2. Cut MIN_GALLOP to 7, but added a per-sort min_gallop vrbl that adapts the "get into galloping mode" threshold higher when galloping isn't paying, and lower when it is. There's no known case where this hurts. It's (of course) neutral for /sort, \sort and =sort. It also happens to be neutral for !sort. It cuts a tiny # of compares in 3sort and +sort. For *sort, it reduces the # of compares to better than what this used to do when MIN_GALLOP was hardcoded to 10 (it did about 0.1% more *sort compares before, but given how close we are to the limit, this is "a lot"!). %sort used to do about 1.5% more compares, and ~sort about 3.6% more. Here are exact counts: i *sort 3sort +sort %sort ~sort !sort 15 449235 33019 33016 51328 188720 65534 before 448885 33016 33007 50426 182083 65534 after 0.08% 0.01% 0.03% 1.79% 3.65% 0.00% %ch from after 16 963714 65824 65809 103409 377634 131070 962991 65821 65808 101667 364341 131070 0.08% 0.00% 0.00% 1.71% 3.65% 0.00% 17 2059092 131413 131362 209130 755476 262142 2057533 131410 131361 206193 728871 262142 0.08% 0.00% 0.00% 1.42% 3.65% 0.00% 18 4380687 262440 262460 421998 1511174 524286 4377402 262437 262459 416347 1457945 524286 0.08% 0.00% 0.00% 1.36% 3.65% 0.00% 19 9285709 524581 524634 848590 3022584 1048574 9278734 524580 524633 837947 2916107 1048574 0.08% 0.00% 0.00% 1.27% 3.65% 0.00% 20 19621118 1048960 1048942 1715806 6045418 2097150 19606028 1048958 1048941 1694896 5832445 2097150 0.08% 0.00% 0.00% 1.23% 3.65% 0.00% 3. Added some key asserts I overlooked before. 4. Updated the doc file.
2002-08-10 02:21:15 -03:00
n lg(n!) *sort 3sort +sort %sort ~sort !sort
------- ------- ------ ------- ------- ------ ------- --------
32768 444255 453096 453614 32908 452871 130491 469141 old
448885 33016 33007 50426 182083 65534 new
0.94% 1273.92% -0.30% 798.09% -28.33% 615.87% %ch from new
65536 954037 972699 981940 65686 973104 260029 1004607
1. Combined the base and length arrays into a single array of structs. This is friendlier for caches. 2. Cut MIN_GALLOP to 7, but added a per-sort min_gallop vrbl that adapts the "get into galloping mode" threshold higher when galloping isn't paying, and lower when it is. There's no known case where this hurts. It's (of course) neutral for /sort, \sort and =sort. It also happens to be neutral for !sort. It cuts a tiny # of compares in 3sort and +sort. For *sort, it reduces the # of compares to better than what this used to do when MIN_GALLOP was hardcoded to 10 (it did about 0.1% more *sort compares before, but given how close we are to the limit, this is "a lot"!). %sort used to do about 1.5% more compares, and ~sort about 3.6% more. Here are exact counts: i *sort 3sort +sort %sort ~sort !sort 15 449235 33019 33016 51328 188720 65534 before 448885 33016 33007 50426 182083 65534 after 0.08% 0.01% 0.03% 1.79% 3.65% 0.00% %ch from after 16 963714 65824 65809 103409 377634 131070 962991 65821 65808 101667 364341 131070 0.08% 0.00% 0.00% 1.71% 3.65% 0.00% 17 2059092 131413 131362 209130 755476 262142 2057533 131410 131361 206193 728871 262142 0.08% 0.00% 0.00% 1.42% 3.65% 0.00% 18 4380687 262440 262460 421998 1511174 524286 4377402 262437 262459 416347 1457945 524286 0.08% 0.00% 0.00% 1.36% 3.65% 0.00% 19 9285709 524581 524634 848590 3022584 1048574 9278734 524580 524633 837947 2916107 1048574 0.08% 0.00% 0.00% 1.27% 3.65% 0.00% 20 19621118 1048960 1048942 1715806 6045418 2097150 19606028 1048958 1048941 1694896 5832445 2097150 0.08% 0.00% 0.00% 1.23% 3.65% 0.00% 3. Added some key asserts I overlooked before. 4. Updated the doc file.
2002-08-10 02:21:15 -03:00
962991 65821 65808 101667 364341 131070
1.01% 1391.83% -0.19% 857.15% -28.63% 666.47%
131072 2039137 2101881 2091491 131232 2092894 554790 2161379
1. Combined the base and length arrays into a single array of structs. This is friendlier for caches. 2. Cut MIN_GALLOP to 7, but added a per-sort min_gallop vrbl that adapts the "get into galloping mode" threshold higher when galloping isn't paying, and lower when it is. There's no known case where this hurts. It's (of course) neutral for /sort, \sort and =sort. It also happens to be neutral for !sort. It cuts a tiny # of compares in 3sort and +sort. For *sort, it reduces the # of compares to better than what this used to do when MIN_GALLOP was hardcoded to 10 (it did about 0.1% more *sort compares before, but given how close we are to the limit, this is "a lot"!). %sort used to do about 1.5% more compares, and ~sort about 3.6% more. Here are exact counts: i *sort 3sort +sort %sort ~sort !sort 15 449235 33019 33016 51328 188720 65534 before 448885 33016 33007 50426 182083 65534 after 0.08% 0.01% 0.03% 1.79% 3.65% 0.00% %ch from after 16 963714 65824 65809 103409 377634 131070 962991 65821 65808 101667 364341 131070 0.08% 0.00% 0.00% 1.71% 3.65% 0.00% 17 2059092 131413 131362 209130 755476 262142 2057533 131410 131361 206193 728871 262142 0.08% 0.00% 0.00% 1.42% 3.65% 0.00% 18 4380687 262440 262460 421998 1511174 524286 4377402 262437 262459 416347 1457945 524286 0.08% 0.00% 0.00% 1.36% 3.65% 0.00% 19 9285709 524581 524634 848590 3022584 1048574 9278734 524580 524633 837947 2916107 1048574 0.08% 0.00% 0.00% 1.27% 3.65% 0.00% 20 19621118 1048960 1048942 1715806 6045418 2097150 19606028 1048958 1048941 1694896 5832445 2097150 0.08% 0.00% 0.00% 1.23% 3.65% 0.00% 3. Added some key asserts I overlooked before. 4. Updated the doc file.
2002-08-10 02:21:15 -03:00
2057533 131410 131361 206193 728871 262142
2.16% 1491.58% -0.10% 915.02% -23.88% 724.51%
262144 4340409 4464460 4403233 262314 4445884 1107842 4584560
1. Combined the base and length arrays into a single array of structs. This is friendlier for caches. 2. Cut MIN_GALLOP to 7, but added a per-sort min_gallop vrbl that adapts the "get into galloping mode" threshold higher when galloping isn't paying, and lower when it is. There's no known case where this hurts. It's (of course) neutral for /sort, \sort and =sort. It also happens to be neutral for !sort. It cuts a tiny # of compares in 3sort and +sort. For *sort, it reduces the # of compares to better than what this used to do when MIN_GALLOP was hardcoded to 10 (it did about 0.1% more *sort compares before, but given how close we are to the limit, this is "a lot"!). %sort used to do about 1.5% more compares, and ~sort about 3.6% more. Here are exact counts: i *sort 3sort +sort %sort ~sort !sort 15 449235 33019 33016 51328 188720 65534 before 448885 33016 33007 50426 182083 65534 after 0.08% 0.01% 0.03% 1.79% 3.65% 0.00% %ch from after 16 963714 65824 65809 103409 377634 131070 962991 65821 65808 101667 364341 131070 0.08% 0.00% 0.00% 1.71% 3.65% 0.00% 17 2059092 131413 131362 209130 755476 262142 2057533 131410 131361 206193 728871 262142 0.08% 0.00% 0.00% 1.42% 3.65% 0.00% 18 4380687 262440 262460 421998 1511174 524286 4377402 262437 262459 416347 1457945 524286 0.08% 0.00% 0.00% 1.36% 3.65% 0.00% 19 9285709 524581 524634 848590 3022584 1048574 9278734 524580 524633 837947 2916107 1048574 0.08% 0.00% 0.00% 1.27% 3.65% 0.00% 20 19621118 1048960 1048942 1715806 6045418 2097150 19606028 1048958 1048941 1694896 5832445 2097150 0.08% 0.00% 0.00% 1.23% 3.65% 0.00% 3. Added some key asserts I overlooked before. 4. Updated the doc file.
2002-08-10 02:21:15 -03:00
4377402 262437 262459 416347 1457945 524286
1.99% 1577.82% -0.06% 967.83% -24.01% 774.44%
524288 9205096 9453356 9408463 524468 9441930 2218577 9692015
9278734 524580 524633 837947 2916107 1048574
1. Combined the base and length arrays into a single array of structs. This is friendlier for caches. 2. Cut MIN_GALLOP to 7, but added a per-sort min_gallop vrbl that adapts the "get into galloping mode" threshold higher when galloping isn't paying, and lower when it is. There's no known case where this hurts. It's (of course) neutral for /sort, \sort and =sort. It also happens to be neutral for !sort. It cuts a tiny # of compares in 3sort and +sort. For *sort, it reduces the # of compares to better than what this used to do when MIN_GALLOP was hardcoded to 10 (it did about 0.1% more *sort compares before, but given how close we are to the limit, this is "a lot"!). %sort used to do about 1.5% more compares, and ~sort about 3.6% more. Here are exact counts: i *sort 3sort +sort %sort ~sort !sort 15 449235 33019 33016 51328 188720 65534 before 448885 33016 33007 50426 182083 65534 after 0.08% 0.01% 0.03% 1.79% 3.65% 0.00% %ch from after 16 963714 65824 65809 103409 377634 131070 962991 65821 65808 101667 364341 131070 0.08% 0.00% 0.00% 1.71% 3.65% 0.00% 17 2059092 131413 131362 209130 755476 262142 2057533 131410 131361 206193 728871 262142 0.08% 0.00% 0.00% 1.42% 3.65% 0.00% 18 4380687 262440 262460 421998 1511174 524286 4377402 262437 262459 416347 1457945 524286 0.08% 0.00% 0.00% 1.36% 3.65% 0.00% 19 9285709 524581 524634 848590 3022584 1048574 9278734 524580 524633 837947 2916107 1048574 0.08% 0.00% 0.00% 1.27% 3.65% 0.00% 20 19621118 1048960 1048942 1715806 6045418 2097150 19606028 1048958 1048941 1694896 5832445 2097150 0.08% 0.00% 0.00% 1.23% 3.65% 0.00% 3. Added some key asserts I overlooked before. 4. Updated the doc file.
2002-08-10 02:21:15 -03:00
1.88% 1693.52% -0.03% 1026.79% -23.92% 824.30%
1048576 19458756 19950272 19838588 1048766 19912134 4430649 20434212
1. Combined the base and length arrays into a single array of structs. This is friendlier for caches. 2. Cut MIN_GALLOP to 7, but added a per-sort min_gallop vrbl that adapts the "get into galloping mode" threshold higher when galloping isn't paying, and lower when it is. There's no known case where this hurts. It's (of course) neutral for /sort, \sort and =sort. It also happens to be neutral for !sort. It cuts a tiny # of compares in 3sort and +sort. For *sort, it reduces the # of compares to better than what this used to do when MIN_GALLOP was hardcoded to 10 (it did about 0.1% more *sort compares before, but given how close we are to the limit, this is "a lot"!). %sort used to do about 1.5% more compares, and ~sort about 3.6% more. Here are exact counts: i *sort 3sort +sort %sort ~sort !sort 15 449235 33019 33016 51328 188720 65534 before 448885 33016 33007 50426 182083 65534 after 0.08% 0.01% 0.03% 1.79% 3.65% 0.00% %ch from after 16 963714 65824 65809 103409 377634 131070 962991 65821 65808 101667 364341 131070 0.08% 0.00% 0.00% 1.71% 3.65% 0.00% 17 2059092 131413 131362 209130 755476 262142 2057533 131410 131361 206193 728871 262142 0.08% 0.00% 0.00% 1.42% 3.65% 0.00% 18 4380687 262440 262460 421998 1511174 524286 4377402 262437 262459 416347 1457945 524286 0.08% 0.00% 0.00% 1.36% 3.65% 0.00% 19 9285709 524581 524634 848590 3022584 1048574 9278734 524580 524633 837947 2916107 1048574 0.08% 0.00% 0.00% 1.27% 3.65% 0.00% 20 19621118 1048960 1048942 1715806 6045418 2097150 19606028 1048958 1048941 1694896 5832445 2097150 0.08% 0.00% 0.00% 1.23% 3.65% 0.00% 3. Added some key asserts I overlooked before. 4. Updated the doc file.
2002-08-10 02:21:15 -03:00
19606028 1048958 1048941 1694896 5832445 2097150
1.76% 1791.27% -0.02% 1074.83% -24.03% 874.38%
Discussion of cases:
*sort: There's no structure in random data to exploit, so the theoretical
limit is lg(n!). Both methods get close to that, and timsort is hugging
it (indeed, in a *marginal* sense, it's a spectacular improvement --
there's only about 1% left before hitting the wall, and timsort knows
darned well it's doing compares that won't pay on random data -- but so
does the samplesort hybrid). For contrast, Hoare's original random-pivot
quicksort does about 39% more compares than the limit, and the median-of-3
variant about 19% more.
3sort, %sort, and !sort: No contest; there's structure in this data, but
not of the specific kinds samplesort special-cases. Note that structure
in !sort wasn't put there on purpose -- it was crafted as a worst case for
a previous quicksort implementation. That timsort nails it came as a
surprise to me (although it's obvious in retrospect).
+sort: samplesort special-cases this data, and does a few less compares
than timsort. However, timsort runs this case significantly faster on all
boxes we have timings for, because timsort is in the business of merging
runs efficiently, while samplesort does much more data movement in this
(for it) special case.
~sort: samplesort's special cases for large masses of equal elements are
extremely effective on ~sort's specific data pattern, and timsort just
isn't going to get close to that, despite that it's clearly getting a
great deal of benefit out of the duplicates (the # of compares is much less
than lg(n!)). ~sort has a perfectly uniform distribution of just 4
distinct values, and as the distribution gets more skewed, samplesort's
equal-element gimmicks become less effective, while timsort's adaptive
strategies find more to exploit; in a database supplied by Kevin Altis, a
sort on its highly skewed "on which stock exchange does this company's
stock trade?" field ran over twice as fast under timsort.
However, despite that timsort does many more comparisons on ~sort, and
that on several platforms ~sort runs highly significantly slower under
timsort, on other platforms ~sort runs highly significantly faster under
timsort. No other kind of data has shown this wild x-platform behavior,
and we don't have an explanation for it. The only thing I can think of
that could transform what "should be" highly significant slowdowns into
highly significant speedups on some boxes are catastrophic cache effects
in samplesort.
But timsort "should be" slower than samplesort on ~sort, so it's hard
to count that it isn't on some boxes as a strike against it <wink>.
+ Here's the highwater mark for the number of heap-based temp slots (4
bytes each on this box) needed by each test, again with arguments
"15 20 1":
2**i *sort \sort /sort 3sort +sort %sort ~sort =sort !sort
32768 16384 0 0 6256 0 10821 12288 0 16383
65536 32766 0 0 21652 0 31276 24576 0 32767
131072 65534 0 0 17258 0 58112 49152 0 65535
262144 131072 0 0 35660 0 123561 98304 0 131071
524288 262142 0 0 31302 0 212057 196608 0 262143
1048576 524286 0 0 312438 0 484942 393216 0 524287
Discussion: The tests that end up doing (close to) perfectly balanced
merges (*sort, !sort) need all N//2 temp slots (or almost all). ~sort
also ends up doing balanced merges, but systematically benefits a lot from
the preliminary pre-merge searches described under "Merge Memory" later.
%sort approaches having a balanced merge at the end because the random
selection of elements to replace is expected to produce an out-of-order
element near the midpoint. \sort, /sort, =sort are the trivial one-run
cases, needing no merging at all. +sort ends up having one very long run
and one very short, and so gets all the temp space it needs from the small
temparray member of the MergeState struct (note that the same would be
true if the new random elements were prefixed to the sorted list instead,
but not if they appeared "in the middle"). 3sort approaches N//3 temp
slots twice, but the run lengths that remain after 3 random exchanges
clearly has very high variance.
A detailed description of timsort follows.
Runs
----
count_run() returns the # of elements in the next run. A run is either
"ascending", which means non-decreasing:
a0 <= a1 <= a2 <= ...
or "descending", which means strictly decreasing:
a0 > a1 > a2 > ...
Note that a run is always at least 2 long, unless we start at the array's
last element.
The definition of descending is strict, because the main routine reverses
a descending run in-place, transforming a descending run into an ascending
run. Reversal is done via the obvious fast "swap elements starting at each
end, and converge at the middle" method, and that can violate stability if
the slice contains any equal elements. Using a strict definition of
descending ensures that a descending run contains distinct elements.
If an array is random, it's very unlikely we'll see long runs. If a natural
run contains less than minrun elements (see next section), the main loop
artificially boosts it to minrun elements, via a stable binary insertion sort
applied to the right number of array elements following the short natural
run. In a random array, *all* runs are likely to be minrun long as a
result. This has two primary good effects:
1. Random data strongly tends then toward perfectly balanced (both runs have
the same length) merges, which is the most efficient way to proceed when
data is random.
2. Because runs are never very short, the rest of the code doesn't make
heroic efforts to shave a few cycles off per-merge overheads. For
example, reasonable use of function calls is made, rather than trying to
inline everything. Since there are no more than N/minrun runs to begin
with, a few "extra" function calls per merge is barely measurable.
Computing minrun
----------------
If N < 64, minrun is N. IOW, binary insertion sort is used for the whole
array then; it's hard to beat that given the overheads of trying something
fancier (see note BINSORT).
When N is a power of 2, testing on random data showed that minrun values of
16, 32, 64 and 128 worked about equally well. At 256 the data-movement cost
in binary insertion sort clearly hurt, and at 8 the increase in the number
of function calls clearly hurt. Picking *some* power of 2 is important
here, so that the merges end up perfectly balanced (see next section). We
pick 32 as a good value in the sweet range; picking a value at the low end
allows the adaptive gimmicks more opportunity to exploit shorter natural
runs.
Because sortperf.py only tries powers of 2, it took a long time to notice
that 32 isn't a good choice for the general case! Consider N=2112:
>>> divmod(2112, 32)
(66, 0)
>>>
If the data is randomly ordered, we're very likely to end up with 66 runs
each of length 32. The first 64 of these trigger a sequence of perfectly
balanced merges (see next section), leaving runs of lengths 2048 and 64 to
merge at the end. The adaptive gimmicks can do that with fewer than 2048+64
compares, but it's still more compares than necessary, and-- mergesort's
bugaboo relative to samplesort --a lot more data movement (O(N) copies just
to get 64 elements into place).
If we take minrun=33 in this case, then we're very likely to end up with 64
runs each of length 33, and then all merges are perfectly balanced. Better!
What we want to avoid is picking minrun such that in
q, r = divmod(N, minrun)
q is a power of 2 and r>0 (then the last merge only gets r elements into
place, and r < minrun is small compared to N), or q a little larger than a
power of 2 regardless of r (then we've got a case similar to "2112", again
leaving too little work for the last merge to do).
Instead we pick a minrun in range(32, 65) such that N/minrun is exactly a
power of 2, or if that isn't possible, is close to, but strictly less than,
a power of 2. This is easier to do than it may sound: take the first 6
bits of N, and add 1 if any of the remaining bits are set. In fact, that
rule covers every case in this section, including small N and exact powers
of 2; merge_compute_minrun() is a deceptively simple function.
The Merge Pattern
-----------------
In order to exploit regularities in the data, we're merging on natural
run lengths, and they can become wildly unbalanced. That's a Good Thing
for this sort! It means we have to find a way to manage an assortment of
potentially very different run lengths, though.
Stability constrains permissible merging patterns. For example, if we have
3 consecutive runs of lengths
A:10000 B:20000 C:10000
we dare not merge A with C first, because if A, B and C happen to contain
a common element, it would get out of order wrt its occurrence(s) in B. The
merging must be done as (A+B)+C or A+(B+C) instead.
So merging is always done on two consecutive runs at a time, and in-place,
although this may require some temp memory (more on that later).
When a run is identified, its length is passed to found_new_run() to
potentially merge runs on a stack of pending runs. We would like to delay
merging as long as possible in order to exploit patterns that may come up
later, but we like even more to do merging as soon as possible to exploit
that the run just found is still high in the memory hierarchy. We also can't
delay merging "too long" because it consumes memory to remember the runs that
are still unmerged, and the stack has a fixed size.
The original version of this code used the first thing I made up that didn't
obviously suck ;-) It was loosely based on invariants involving the Fibonacci
sequence.
It worked OK, but it was hard to reason about, and was subtle enough that the
intended invariants weren't actually preserved. Researchers discovered that
when trying to complete a computer-generated correctness proof. That was
easily-enough repaired, but the discovery spurred quite a bit of academic
interest in truly good ways to manage incremental merging on the fly.
At least a dozen different approaches were developed, some provably having
near-optimal worst case behavior with respect to the entropy of the
distribution of run lengths. Some details can be found in bpo-34561.
The code now uses the "powersort" merge strategy from:
"Nearly-Optimal Mergesorts: Fast, Practical Sorting Methods
That Optimally Adapt to Existing Runs"
J. Ian Munro and Sebastian Wild
The code is pretty simple, but the justification is quite involved, as it's
based on fast approximations to optimal binary search trees, which are
substantial topics on their own.
Here we'll just cover some pragmatic details:
The `powerloop()` function computes a run's "power". Say two adjacent runs
begin at index s1. The first run has length n1, and the second run (starting
at index s1+n1, called "s2" below) has length n2. The list has total length n.
The "power" of the first run is a small integer, the depth of the node
connecting the two runs in an ideal binary merge tree, where power 1 is the
root node, and the power increases by 1 for each level deeper in the tree.
The power is the least integer L such that the "midpoint interval" contains
a rational number of the form J/2**L. The midpoint interval is the semi-
closed interval:
((s1 + n1/2)/n, (s2 + n2/2)/n]
Yes, that's brain-busting at first ;-) Concretely, if (s1 + n1/2)/n and
(s2 + n2/2)/n are computed to infinite precision in binary, the power L is
the first position at which the 2**-L bit differs between the expansions.
Since the left end of the interval is less than the right end, the first
differing bit must be a 0 bit in the left quotient and a 1 bit in the right
quotient.
`powerloop()` emulates these divisions, 1 bit at a time, using comparisons,
subtractions, and shifts in a loop.
You'll notice the paper uses an O(1) method instead, but that relies on two
things we don't have:
- An O(1) "count leading zeroes" primitive. We can find such a thing as a C
extension on most platforms, but not all, and there's no uniform spelling
on the platforms that support it.
- Integer division on an integer type twice as wide as needed to hold the
list length. But the latter is Py_ssize_t for us, and is typically the
widest native signed integer type the platform supports.
But since runs in our algorithm are almost never very short, the once-per-run
overhead of `powerloop()` seems lost in the noise.
Detail: why is Py_ssize_t "wide enough" in `powerloop()`? We do, after all,
shift integers of that width left by 1. How do we know that won't spill into
the sign bit? The trick is that we have some slop. `n` (the total list
length) is the number of list elements, which is at most 4 times (on a 32-box,
with 4-byte pointers) smaller than than the largest size_t. So at least the
leading two bits of the integers we're using are clear.
Since we can't compute a run's power before seeing the run that follows it,
the most-recently identified run is never merged by `found_new_run()`.
Instead a new run is only used to compute the 2nd-most-recent run's power.
Then adjacent runs are merged so long as their saved power (tree depth) is
greater than that newly computed power. When found_new_run() returns, only
then is a new run pushed on to the stack of pending runs.
A key invariant is that powers on the run stack are strictly decreasing
(starting from the run at the top of the stack).
Note that even powersort's strategy isn't always truly optimal. It can't be.
Computing an optimal merge sequence can be done in time quadratic in the
number of runs, which is very much slower, and also requires finding &
remembering _all_ the runs' lengths (of which there may be billions) in
advance. It's remarkable, though, how close to optimal this strategy gets.
Curious factoid: of all the alternatives I've seen in the literature,
powersort's is the only one that's always truly optimal for a collection of 3
run lengths (for three lengths A B C, it's always optimal to first merge the
shorter of A and C with B).
Merge Memory
------------
Merging adjacent runs of lengths A and B in-place, and in linear time, is
difficult. Theoretical constructions are known that can do it, but they're
too difficult and slow for practical use. But if we have temp memory equal
to min(A, B), it's easy.
If A is smaller (function merge_lo), copy A to a temp array, leave B alone,
and then we can do the obvious merge algorithm left to right, from the temp
area and B, starting the stores into where A used to live. There's always a
free area in the original area comprising a number of elements equal to the
number not yet merged from the temp array (trivially true at the start;
proceed by induction). The only tricky bit is that if a comparison raises an
exception, we have to remember to copy the remaining elements back in from
the temp area, lest the array end up with duplicate entries from B. But
that's exactly the same thing we need to do if we reach the end of B first,
so the exit code is pleasantly common to both the normal and error cases.
If B is smaller (function merge_hi, which is merge_lo's "mirror image"),
much the same, except that we need to merge right to left, copying B into a
temp array and starting the stores at the right end of where B used to live.
A refinement: When we're about to merge adjacent runs A and B, we first do
a form of binary search (more on that later) to see where B[0] should end up
in A. Elements in A preceding that point are already in their final
positions, effectively shrinking the size of A. Likewise we also search to
see where A[-1] should end up in B, and elements of B after that point can
also be ignored. This cuts the amount of temp memory needed by the same
amount.
These preliminary searches may not pay off, and can be expected *not* to
repay their cost if the data is random. But they can win huge in all of
time, copying, and memory savings when they do pay, so this is one of the
"per-merge overheads" mentioned above that we're happy to endure because
there is at most one very short run. It's generally true in this algorithm
that we're willing to gamble a little to win a lot, even though the net
expectation is negative for random data.
Merge Algorithms
----------------
merge_lo() and merge_hi() are where the bulk of the time is spent. merge_lo
deals with runs where A <= B, and merge_hi where A > B. They don't know
whether the data is clustered or uniform, but a lovely thing about merging
is that many kinds of clustering "reveal themselves" by how many times in a
row the winning merge element comes from the same run. We'll only discuss
merge_lo here; merge_hi is exactly analogous.
Merging begins in the usual, obvious way, comparing the first element of A
to the first of B, and moving B[0] to the merge area if it's less than A[0],
else moving A[0] to the merge area. Call that the "one pair at a time"
mode. The only twist here is keeping track of how many times in a row "the
winner" comes from the same run.
If that count reaches MIN_GALLOP, we switch to "galloping mode". Here
we *search* B for where A[0] belongs, and move over all the B's before
that point in one chunk to the merge area, then move A[0] to the merge
area. Then we search A for where B[0] belongs, and similarly move a
slice of A in one chunk. Then back to searching B for where A[0] belongs,
etc. We stay in galloping mode until both searches find slices to copy
less than MIN_GALLOP elements long, at which point we go back to one-pair-
at-a-time mode.
1. Combined the base and length arrays into a single array of structs. This is friendlier for caches. 2. Cut MIN_GALLOP to 7, but added a per-sort min_gallop vrbl that adapts the "get into galloping mode" threshold higher when galloping isn't paying, and lower when it is. There's no known case where this hurts. It's (of course) neutral for /sort, \sort and =sort. It also happens to be neutral for !sort. It cuts a tiny # of compares in 3sort and +sort. For *sort, it reduces the # of compares to better than what this used to do when MIN_GALLOP was hardcoded to 10 (it did about 0.1% more *sort compares before, but given how close we are to the limit, this is "a lot"!). %sort used to do about 1.5% more compares, and ~sort about 3.6% more. Here are exact counts: i *sort 3sort +sort %sort ~sort !sort 15 449235 33019 33016 51328 188720 65534 before 448885 33016 33007 50426 182083 65534 after 0.08% 0.01% 0.03% 1.79% 3.65% 0.00% %ch from after 16 963714 65824 65809 103409 377634 131070 962991 65821 65808 101667 364341 131070 0.08% 0.00% 0.00% 1.71% 3.65% 0.00% 17 2059092 131413 131362 209130 755476 262142 2057533 131410 131361 206193 728871 262142 0.08% 0.00% 0.00% 1.42% 3.65% 0.00% 18 4380687 262440 262460 421998 1511174 524286 4377402 262437 262459 416347 1457945 524286 0.08% 0.00% 0.00% 1.36% 3.65% 0.00% 19 9285709 524581 524634 848590 3022584 1048574 9278734 524580 524633 837947 2916107 1048574 0.08% 0.00% 0.00% 1.27% 3.65% 0.00% 20 19621118 1048960 1048942 1715806 6045418 2097150 19606028 1048958 1048941 1694896 5832445 2097150 0.08% 0.00% 0.00% 1.23% 3.65% 0.00% 3. Added some key asserts I overlooked before. 4. Updated the doc file.
2002-08-10 02:21:15 -03:00
A refinement: The MergeState struct contains the value of min_gallop that
controls when we enter galloping mode, initialized to MIN_GALLOP.
merge_lo() and merge_hi() adjust this higher when galloping isn't paying
1. Combined the base and length arrays into a single array of structs. This is friendlier for caches. 2. Cut MIN_GALLOP to 7, but added a per-sort min_gallop vrbl that adapts the "get into galloping mode" threshold higher when galloping isn't paying, and lower when it is. There's no known case where this hurts. It's (of course) neutral for /sort, \sort and =sort. It also happens to be neutral for !sort. It cuts a tiny # of compares in 3sort and +sort. For *sort, it reduces the # of compares to better than what this used to do when MIN_GALLOP was hardcoded to 10 (it did about 0.1% more *sort compares before, but given how close we are to the limit, this is "a lot"!). %sort used to do about 1.5% more compares, and ~sort about 3.6% more. Here are exact counts: i *sort 3sort +sort %sort ~sort !sort 15 449235 33019 33016 51328 188720 65534 before 448885 33016 33007 50426 182083 65534 after 0.08% 0.01% 0.03% 1.79% 3.65% 0.00% %ch from after 16 963714 65824 65809 103409 377634 131070 962991 65821 65808 101667 364341 131070 0.08% 0.00% 0.00% 1.71% 3.65% 0.00% 17 2059092 131413 131362 209130 755476 262142 2057533 131410 131361 206193 728871 262142 0.08% 0.00% 0.00% 1.42% 3.65% 0.00% 18 4380687 262440 262460 421998 1511174 524286 4377402 262437 262459 416347 1457945 524286 0.08% 0.00% 0.00% 1.36% 3.65% 0.00% 19 9285709 524581 524634 848590 3022584 1048574 9278734 524580 524633 837947 2916107 1048574 0.08% 0.00% 0.00% 1.27% 3.65% 0.00% 20 19621118 1048960 1048942 1715806 6045418 2097150 19606028 1048958 1048941 1694896 5832445 2097150 0.08% 0.00% 0.00% 1.23% 3.65% 0.00% 3. Added some key asserts I overlooked before. 4. Updated the doc file.
2002-08-10 02:21:15 -03:00
off, and lower when it is.
Galloping
---------
Still without loss of generality, assume A is the shorter run. In galloping
mode, we first look for A[0] in B. We do this via "galloping", comparing
A[0] in turn to B[0], B[1], B[3], B[7], ..., B[2**j - 1], ..., until finding
the k such that B[2**(k-1) - 1] < A[0] <= B[2**k - 1]. This takes at most
roughly lg(B) comparisons, and, unlike a straight binary search, favors
finding the right spot early in B (more on that later).
After finding such a k, the region of uncertainty is reduced to 2**(k-1) - 1
consecutive elements, and a straight binary search requires exactly k-1
additional comparisons to nail it (see note REGION OF UNCERTAINTY). Then we
copy all the B's up to that point in one chunk, and then copy A[0]. Note
that no matter where A[0] belongs in B, the combination of galloping + binary
search finds it in no more than about 2*lg(B) comparisons.
If we did a straight binary search, we could find it in no more than
ceiling(lg(B+1)) comparisons -- but straight binary search takes that many
comparisons no matter where A[0] belongs. Straight binary search thus loses
to galloping unless the run is quite long, and we simply can't guess
whether it is in advance.
If data is random and runs have the same length, A[0] belongs at B[0] half
the time, at B[1] a quarter of the time, and so on: a consecutive winning
sub-run in B of length k occurs with probability 1/2**(k+1). So long
winning sub-runs are extremely unlikely in random data, and guessing that a
winning sub-run is going to be long is a dangerous game.
OTOH, if data is lopsided or lumpy or contains many duplicates, long
stretches of winning sub-runs are very likely, and cutting the number of
comparisons needed to find one from O(B) to O(log B) is a huge win.
Galloping compromises by getting out fast if there isn't a long winning
sub-run, yet finding such very efficiently when they exist.
I first learned about the galloping strategy in a related context; see:
"Adaptive Set Intersections, Unions, and Differences" (2000)
Erik D. Demaine, Alejandro López-Ortiz, J. Ian Munro
and its followup(s). An earlier paper called the same strategy
"exponential search":
"Optimistic Sorting and Information Theoretic Complexity"
Peter McIlroy
SODA (Fourth Annual ACM-SIAM Symposium on Discrete Algorithms), pp
467-474, Austin, Texas, 25-27 January 1993.
and it probably dates back to an earlier paper by Bentley and Yao. The
Merged revisions 46753-51188 via svnmerge from svn+ssh://pythondev@svn.python.org/python/trunk ........ r46755 | brett.cannon | 2006-06-08 18:23:04 +0200 (Thu, 08 Jun 2006) | 4 lines Make binascii.hexlify() use s# for its arguments instead of t# to actually match its documentation stating it accepts any read-only buffer. ........ r46757 | brett.cannon | 2006-06-08 19:00:45 +0200 (Thu, 08 Jun 2006) | 8 lines Buffer objects would return the read or write buffer for a wrapped object when the char buffer was requested. Now it actually returns the char buffer if available or raises a TypeError if it isn't (as is raised for the other buffer types if they are not present but requested). Not a backport candidate since it does change semantics of the buffer object (although it could be argued this is enough of a bug to bother backporting). ........ r46760 | andrew.kuchling | 2006-06-09 03:10:17 +0200 (Fri, 09 Jun 2006) | 1 line Update functools section ........ r46762 | tim.peters | 2006-06-09 04:11:02 +0200 (Fri, 09 Jun 2006) | 6 lines Whitespace normalization. Since test_file is implicated in mysterious test failures when followed by test_optparse, if I had any brains I'd look at the checkin that last changed test_file ;-) ........ r46763 | tim.peters | 2006-06-09 05:09:42 +0200 (Fri, 09 Jun 2006) | 5 lines To boost morale :-), force test_optparse to run immediately after test_file until we can figure out how to fix it. (See python-dev; at the moment we don't even know which checkin caused the problem.) ........ r46764 | tim.peters | 2006-06-09 05:51:41 +0200 (Fri, 09 Jun 2006) | 6 lines AutoFileTests.tearDown(): Removed mysterious undocumented try/except. Remove TESTFN. Throughout: used open() instead of file(), and wrapped long lines. ........ r46765 | tim.peters | 2006-06-09 06:02:06 +0200 (Fri, 09 Jun 2006) | 8 lines testUnicodeOpen(): I have no idea why, but making this test clean up after itself appears to fix the test failures when test_optparse follows test_file. test_main(): Get rid of TESTFN no matter what. That's also enough to fix the mystery failures. Doesn't hurt to fix them twice :-) ........ r46766 | tim.peters | 2006-06-09 07:12:40 +0200 (Fri, 09 Jun 2006) | 6 lines Remove the temporary hack to force test_optparse to run immediately after test_file. At least 8 buildbot boxes passed since the underlying problem got fixed, and they all failed before the fix, so there's no point to this anymore. ........ r46767 | neal.norwitz | 2006-06-09 07:54:18 +0200 (Fri, 09 Jun 2006) | 1 line Fix grammar and reflow ........ r46769 | andrew.kuchling | 2006-06-09 12:22:35 +0200 (Fri, 09 Jun 2006) | 1 line Markup fix ........ r46773 | andrew.kuchling | 2006-06-09 15:15:57 +0200 (Fri, 09 Jun 2006) | 1 line [Bug #1472827] Make saxutils.XMLGenerator handle \r\n\t in attribute values by escaping them properly. 2.4 bugfix candidate. ........ r46778 | kristjan.jonsson | 2006-06-09 18:28:01 +0200 (Fri, 09 Jun 2006) | 2 lines Turn off warning about deprecated CRT functions on for VisualStudio .NET 2005. Make the definition #ARRAYSIZE conditional. VisualStudio .NET 2005 already has it defined using a better gimmick. ........ r46779 | phillip.eby | 2006-06-09 18:40:18 +0200 (Fri, 09 Jun 2006) | 2 lines Import wsgiref into the stdlib, as of the external version 0.1-r2181. ........ r46783 | andrew.kuchling | 2006-06-09 18:44:40 +0200 (Fri, 09 Jun 2006) | 1 line Add note about XMLGenerator bugfix ........ r46784 | andrew.kuchling | 2006-06-09 18:46:51 +0200 (Fri, 09 Jun 2006) | 1 line Add note about wsgiref ........ r46785 | brett.cannon | 2006-06-09 19:05:48 +0200 (Fri, 09 Jun 2006) | 2 lines Fix inconsistency in naming within an enum. ........ r46787 | tim.peters | 2006-06-09 19:47:00 +0200 (Fri, 09 Jun 2006) | 2 lines Whitespace normalization. ........ r46792 | georg.brandl | 2006-06-09 20:29:52 +0200 (Fri, 09 Jun 2006) | 3 lines Test file.__exit__. ........ r46794 | brett.cannon | 2006-06-09 20:40:46 +0200 (Fri, 09 Jun 2006) | 2 lines svn:ignore .pyc and .pyo files. ........ r46795 | georg.brandl | 2006-06-09 20:45:48 +0200 (Fri, 09 Jun 2006) | 3 lines RFE #1491485: str/unicode.endswith()/startswith() now accept a tuple as first argument. ........ r46798 | andrew.kuchling | 2006-06-09 21:03:16 +0200 (Fri, 09 Jun 2006) | 1 line Describe startswith()/endswiith() change; add reminder about wsgiref ........ r46799 | tim.peters | 2006-06-09 21:24:44 +0200 (Fri, 09 Jun 2006) | 11 lines Implementing a happy idea from Georg Brandl: make runtest() try to clean up files and directories the tests often leave behind by mistake. This is the first time in history I don't have a bogus "db_home" directory after running the tests ;-) Also worked on runtest's docstring, to say something about all the arguments, and to document the non-obvious return values. New functions runtest_inner() and cleanup_test_droppings() in support of the above. ........ r46800 | andrew.kuchling | 2006-06-09 21:43:25 +0200 (Fri, 09 Jun 2006) | 1 line Remove unused variable ........ r46801 | andrew.kuchling | 2006-06-09 21:56:05 +0200 (Fri, 09 Jun 2006) | 1 line Add some wsgiref text ........ r46803 | thomas.heller | 2006-06-09 21:59:11 +0200 (Fri, 09 Jun 2006) | 1 line set eol-style svn property ........ r46804 | thomas.heller | 2006-06-09 22:01:01 +0200 (Fri, 09 Jun 2006) | 1 line set eol-style svn property ........ r46805 | georg.brandl | 2006-06-09 22:43:48 +0200 (Fri, 09 Jun 2006) | 3 lines Make use of new str.startswith/endswith semantics. Occurences in email and compiler were ignored due to backwards compat requirements. ........ r46806 | brett.cannon | 2006-06-10 00:31:23 +0200 (Sat, 10 Jun 2006) | 4 lines An object with __call__ as an attribute, when called, will have that attribute checked for __call__ itself, and will continue to look until it finds an object without the attribute. This can lead to an infinite recursion. Closes bug #532646, again. Will be backported. ........ r46808 | brett.cannon | 2006-06-10 00:45:54 +0200 (Sat, 10 Jun 2006) | 2 lines Fix bug introduced in rev. 46806 by not having variable declaration at the top of a block. ........ r46812 | georg.brandl | 2006-06-10 08:40:50 +0200 (Sat, 10 Jun 2006) | 4 lines Apply perky's fix for #1503157: "/".join([u"", u""]) raising OverflowError. Also improve error message on overflow. ........ r46817 | martin.v.loewis | 2006-06-10 10:14:03 +0200 (Sat, 10 Jun 2006) | 2 lines Port cygwin kill_python changes from 2.4 branch. ........ r46818 | armin.rigo | 2006-06-10 12:57:40 +0200 (Sat, 10 Jun 2006) | 4 lines SF bug #1503294. PyThreadState_GET() complains if the tstate is NULL, but only in debug mode. ........ r46819 | martin.v.loewis | 2006-06-10 14:23:46 +0200 (Sat, 10 Jun 2006) | 4 lines Patch #1495999: Part two of Windows CE changes. - update header checks, using autoconf - provide dummies for getenv, environ, and GetVersion - adjust MSC_VER check in socketmodule.c ........ r46820 | skip.montanaro | 2006-06-10 16:09:11 +0200 (Sat, 10 Jun 2006) | 1 line document the class, not its initializer ........ r46821 | greg.ward | 2006-06-10 18:40:01 +0200 (Sat, 10 Jun 2006) | 4 lines Sync with Optik docs (rev 518): * restore "Extending optparse" section * document ALWAYS_TYPED_ACTIONS (SF #1449311) ........ r46824 | thomas.heller | 2006-06-10 21:51:46 +0200 (Sat, 10 Jun 2006) | 8 lines Upgrade to ctypes version 0.9.9.7. Summary of changes: - support for 'variable sized' data - support for anonymous structure/union fields - fix severe bug with certain arrays or structures containing more than 256 fields ........ r46825 | thomas.heller | 2006-06-10 21:55:36 +0200 (Sat, 10 Jun 2006) | 8 lines Upgrade to ctypes version 0.9.9.7. Summary of changes: - support for 'variable sized' data - support for anonymous structure/union fields - fix severe bug with certain arrays or structures containing more than 256 fields ........ r46826 | fred.drake | 2006-06-10 22:01:34 +0200 (Sat, 10 Jun 2006) | 4 lines SF patch #1303595: improve description of __builtins__, explaining how it varies between __main__ and other modules, and strongly suggest not touching it but using __builtin__ if absolutely necessary ........ r46827 | fred.drake | 2006-06-10 22:02:58 +0200 (Sat, 10 Jun 2006) | 1 line credit for SF patch #1303595 ........ r46831 | thomas.heller | 2006-06-10 22:29:34 +0200 (Sat, 10 Jun 2006) | 2 lines New docs for ctypes. ........ r46834 | thomas.heller | 2006-06-10 23:07:19 +0200 (Sat, 10 Jun 2006) | 1 line Fix a wrong printf format. ........ r46835 | thomas.heller | 2006-06-10 23:17:58 +0200 (Sat, 10 Jun 2006) | 1 line Fix the second occurrence of the problematic printf format. ........ r46837 | thomas.heller | 2006-06-10 23:56:03 +0200 (Sat, 10 Jun 2006) | 1 line Don't use C++ comment. ........ r46838 | thomas.heller | 2006-06-11 00:01:50 +0200 (Sun, 11 Jun 2006) | 1 line Handle failure of PyMem_Realloc. ........ r46839 | skip.montanaro | 2006-06-11 00:38:13 +0200 (Sun, 11 Jun 2006) | 2 lines Suppress warning on MacOSX about possible use before set of proc. ........ r46840 | tim.peters | 2006-06-11 00:51:45 +0200 (Sun, 11 Jun 2006) | 8 lines shuffle() doscstring: Removed warning about sequence length versus generator period. While this was a real weakness of the older WH generator for lists with just a few dozen elements, and so could potentially bite the naive ;-), the Twister should show excellent behavior up to at least 600 elements. Module docstring: reflowed some jarringly short lines. ........ r46844 | greg.ward | 2006-06-11 02:40:49 +0200 (Sun, 11 Jun 2006) | 4 lines Bug #1361643: fix textwrap.dedent() so it handles tabs appropriately, i.e. do *not* expand tabs, but treat them as whitespace that is not equivalent to spaces. Add a couple of test cases. Clarify docs. ........ r46850 | neal.norwitz | 2006-06-11 07:44:18 +0200 (Sun, 11 Jun 2006) | 5 lines Fix Coverity # 146. newDBSequenceObject would deref dbobj, so it can't be NULL. We know it's not NULL from the ParseTuple and DbObject_Check will verify it's not NULL. ........ r46851 | neal.norwitz | 2006-06-11 07:45:25 +0200 (Sun, 11 Jun 2006) | 4 lines Wrap some long lines Top/Bottom factor out some common expressions Add a XXX comment about widing offset. ........ r46852 | neal.norwitz | 2006-06-11 07:45:47 +0200 (Sun, 11 Jun 2006) | 1 line Add versionadded to doc ........ r46853 | neal.norwitz | 2006-06-11 07:47:14 +0200 (Sun, 11 Jun 2006) | 3 lines Update doc to make it agree with code. Bottom factor out some common code. ........ r46854 | neal.norwitz | 2006-06-11 07:48:14 +0200 (Sun, 11 Jun 2006) | 3 lines f_code can't be NULL based on Frame_New and other code that derefs it. So there doesn't seem to be much point to checking here. ........ r46855 | neal.norwitz | 2006-06-11 09:26:27 +0200 (Sun, 11 Jun 2006) | 1 line Fix errors found by pychecker ........ r46856 | neal.norwitz | 2006-06-11 09:26:50 +0200 (Sun, 11 Jun 2006) | 1 line warnings was imported at module scope, no need to import again ........ r46857 | neal.norwitz | 2006-06-11 09:27:56 +0200 (Sun, 11 Jun 2006) | 5 lines Fix errors found by pychecker. I think these changes are correct, but I'm not sure. Could someone who knows how this module works test it? It can at least start on the cmd line. ........ r46858 | neal.norwitz | 2006-06-11 10:35:14 +0200 (Sun, 11 Jun 2006) | 1 line Fix errors found by pychecker ........ r46859 | ronald.oussoren | 2006-06-11 16:33:36 +0200 (Sun, 11 Jun 2006) | 4 lines This patch improves the L&F of IDLE on OSX. The changes are conditionalized on being in an IDLE.app bundle on darwin. This does a slight reorganisation of the menus and adds support for file-open events. ........ r46860 | greg.ward | 2006-06-11 16:42:41 +0200 (Sun, 11 Jun 2006) | 1 line SF #1366250: optparse docs: fix inconsistency in variable name; minor tweaks. ........ r46861 | greg.ward | 2006-06-11 18:24:11 +0200 (Sun, 11 Jun 2006) | 3 lines Bug #1498146: fix optparse to handle Unicode strings in option help, description, and epilog. ........ r46862 | thomas.heller | 2006-06-11 19:04:22 +0200 (Sun, 11 Jun 2006) | 2 lines Release the GIL during COM method calls, to avoid deadlocks in Python coded COM objects. ........ r46863 | tim.peters | 2006-06-11 21:42:51 +0200 (Sun, 11 Jun 2006) | 2 lines Whitespace normalization. ........ r46864 | tim.peters | 2006-06-11 21:43:49 +0200 (Sun, 11 Jun 2006) | 2 lines Add missing svn:eol-style property to text files. ........ r46865 | ronald.oussoren | 2006-06-11 21:45:57 +0200 (Sun, 11 Jun 2006) | 2 lines Remove message about using make frameworkinstall, that's no longer necesssary ........ r46866 | ronald.oussoren | 2006-06-11 22:23:29 +0200 (Sun, 11 Jun 2006) | 2 lines Use configure to substitute the correct prefix instead of hardcoding ........ r46867 | ronald.oussoren | 2006-06-11 22:24:45 +0200 (Sun, 11 Jun 2006) | 4 lines - Change fixapplepython23.py to ensure that it will run with /usr/bin/python on intel macs. - Fix some minor problems in the installer for OSX ........ r46868 | neal.norwitz | 2006-06-11 22:25:56 +0200 (Sun, 11 Jun 2006) | 5 lines Try to fix several networking tests. The problem is that if hosts have a search path setup, some of these hosts resolve to the wrong address. By appending a period to the hostname, the hostname should only resolve to what we want it to resolve to. Hopefully this doesn't break different bots. ........ r46869 | neal.norwitz | 2006-06-11 22:42:02 +0200 (Sun, 11 Jun 2006) | 7 lines Try to fix another networking test. The problem is that if hosts have a search path setup, some of these hosts resolve to the wrong address. By appending a period to the hostname, the hostname should only resolve to what we want it to resolve to. Hopefully this doesn't break different bots. Also add more info to failure message to aid debugging test failure. ........ r46870 | neal.norwitz | 2006-06-11 22:46:46 +0200 (Sun, 11 Jun 2006) | 4 lines Fix test on PPC64 buildbot. It raised an IOError (really an URLError which derives from an IOError). That seems valid. Env Error includes both OSError and IOError, so this seems like a reasonable fix. ........ r46871 | tim.peters | 2006-06-11 22:52:59 +0200 (Sun, 11 Jun 2006) | 10 lines compare_generic_iter(): Fixed the failure of test_wsgiref's testFileWrapper when running with -O. test_simple_validation_error still fails under -O. That appears to be because wsgiref's validate.py uses `assert` statements all over the place to check arguments for sanity. That should all be changed (it's not a logical error in the software if a user passes bogus arguments, so this isn't a reasonable use for `assert` -- checking external preconditions should generally raise ValueError or TypeError instead, as appropriate). ........ r46872 | neal.norwitz | 2006-06-11 23:38:38 +0200 (Sun, 11 Jun 2006) | 1 line Get test to pass on S/390. Shout if you think this change is incorrect. ........ r46873 | neal.norwitz | 2006-06-12 04:05:55 +0200 (Mon, 12 Jun 2006) | 1 line Cleanup Py_ssize_t a little (get rid of second #ifdef) ........ r46874 | neal.norwitz | 2006-06-12 04:06:17 +0200 (Mon, 12 Jun 2006) | 1 line Fix some Py_ssize_t issues ........ r46875 | neal.norwitz | 2006-06-12 04:06:42 +0200 (Mon, 12 Jun 2006) | 1 line Fix some Py_ssize_t issues ........ r46876 | neal.norwitz | 2006-06-12 04:07:24 +0200 (Mon, 12 Jun 2006) | 2 lines Cleanup: Remove import of types to get StringTypes, we can just use basestring. ........ r46877 | neal.norwitz | 2006-06-12 04:07:57 +0200 (Mon, 12 Jun 2006) | 1 line Don't truncate if size_t is bigger than uint ........ r46878 | neal.norwitz | 2006-06-12 04:08:41 +0200 (Mon, 12 Jun 2006) | 1 line Don't leak the list object if there's an error allocating the item storage. Backport candidate ........ r46879 | neal.norwitz | 2006-06-12 04:09:03 +0200 (Mon, 12 Jun 2006) | 1 line Fix typo. Backport if anyone cares. :-) ........ r46880 | neal.norwitz | 2006-06-12 04:09:34 +0200 (Mon, 12 Jun 2006) | 1 line Fix indentation of case and a Py_ssize_t issue. ........ r46881 | neal.norwitz | 2006-06-12 04:11:18 +0200 (Mon, 12 Jun 2006) | 3 lines Get rid of f_restricted too. Doc the other 4 ints that were already removed at the NeedForSpeed sprint. ........ r46882 | neal.norwitz | 2006-06-12 04:13:21 +0200 (Mon, 12 Jun 2006) | 1 line Fix the socket tests so they can be run concurrently. Backport candidate ........ r46883 | neal.norwitz | 2006-06-12 04:16:10 +0200 (Mon, 12 Jun 2006) | 1 line i and j are initialized below when used. No need to do it twice ........ r46884 | neal.norwitz | 2006-06-12 05:05:03 +0200 (Mon, 12 Jun 2006) | 1 line Remove unused import ........ r46885 | neal.norwitz | 2006-06-12 05:05:40 +0200 (Mon, 12 Jun 2006) | 1 line Impl ssize_t ........ r46886 | neal.norwitz | 2006-06-12 05:33:09 +0200 (Mon, 12 Jun 2006) | 6 lines Patch #1503046, Conditional compilation of zlib.(de)compressobj.copy copy is only in newer versions of zlib. This should allow zlibmodule to work with older versions like the Tru64 buildbot. ........ r46887 | phillip.eby | 2006-06-12 06:04:32 +0200 (Mon, 12 Jun 2006) | 2 lines Sync w/external release 0.1.2. Please see PEP 360 before making changes to external packages. ........ r46888 | martin.v.loewis | 2006-06-12 06:26:31 +0200 (Mon, 12 Jun 2006) | 2 lines Get rid of function pointer cast. ........ r46889 | thomas.heller | 2006-06-12 08:05:57 +0200 (Mon, 12 Jun 2006) | 3 lines I don't know how that happend, but the entire file contents was duplicated. Thanks to Simon Percivall for the heads up. ........ r46890 | nick.coghlan | 2006-06-12 10:19:37 +0200 (Mon, 12 Jun 2006) | 1 line Fix site module docstring to match the code ........ r46891 | nick.coghlan | 2006-06-12 10:23:02 +0200 (Mon, 12 Jun 2006) | 1 line Fix site module docstring to match the code for Mac OSX, too ........ r46892 | nick.coghlan | 2006-06-12 10:27:13 +0200 (Mon, 12 Jun 2006) | 1 line The site module documentation also described the Windows behaviour incorrectly. ........ r46893 | nick.coghlan | 2006-06-12 12:17:11 +0200 (Mon, 12 Jun 2006) | 1 line Make the -m switch conform to the documentation of sys.path by behaving like the -c switch ........ r46894 | kristjan.jonsson | 2006-06-12 17:45:12 +0200 (Mon, 12 Jun 2006) | 2 lines Fix the CRT argument error handling for VisualStudio .NET 2005. Install a CRT error handler and disable the assertion for debug builds. This causes CRT to set errno to EINVAL. This update fixes crash cases in the test suite where the default CRT error handler would cause process exit. ........ r46899 | thomas.heller | 2006-06-12 22:56:48 +0200 (Mon, 12 Jun 2006) | 1 line Add pep-291 compatibility markers. ........ r46901 | ka-ping.yee | 2006-06-13 01:47:52 +0200 (Tue, 13 Jun 2006) | 5 lines Add the uuid module. This module has been tested so far on Windows XP (Python 2.4 and 2.5a2), Mac OS X (Python 2.3, 2.4, and 2.5a2), and Linux (Python 2.4 and 2.5a2). ........ r46902 | tim.peters | 2006-06-13 02:30:01 +0200 (Tue, 13 Jun 2006) | 2 lines Whitespace normalization. ........ r46903 | tim.peters | 2006-06-13 02:30:50 +0200 (Tue, 13 Jun 2006) | 2 lines Added missing svn:eol-style property to text files. ........ r46905 | tim.peters | 2006-06-13 05:30:07 +0200 (Tue, 13 Jun 2006) | 5 lines get_matching_blocks(): rewrote code & comments so they match; added more comments about why it's this way at all; and removed what looked like needless expense (sorting (i, j, k) triples directly should give exactly the same order as sorting (i, (i, j, k)) pairs). ........ r46906 | neal.norwitz | 2006-06-13 06:08:53 +0200 (Tue, 13 Jun 2006) | 1 line Don't fail if another process is listening on our port. ........ r46908 | neal.norwitz | 2006-06-13 10:28:19 +0200 (Tue, 13 Jun 2006) | 2 lines Initialize the type object so pychecker can't crash the interpreter. ........ r46909 | neal.norwitz | 2006-06-13 10:41:06 +0200 (Tue, 13 Jun 2006) | 1 line Verify the crash due to EncodingMap not initialized does not return ........ r46910 | thomas.heller | 2006-06-13 10:56:14 +0200 (Tue, 13 Jun 2006) | 3 lines Add some windows datatypes that were missing from this file, and add the aliases defined in windows header files for the structures. ........ r46911 | thomas.heller | 2006-06-13 11:40:14 +0200 (Tue, 13 Jun 2006) | 3 lines Add back WCHAR, UINT, DOUBLE, _LARGE_INTEGER, _ULARGE_INTEGER. VARIANT_BOOL is a special _ctypes data type, not c_short. ........ r46912 | ronald.oussoren | 2006-06-13 13:19:56 +0200 (Tue, 13 Jun 2006) | 4 lines Linecache contains support for PEP302 loaders, but fails to deal with loaders that return None to indicate that the module is valid but no source is available. This patch fixes that. ........ r46913 | andrew.kuchling | 2006-06-13 13:57:04 +0200 (Tue, 13 Jun 2006) | 1 line Mention uuid module ........ r46915 | walter.doerwald | 2006-06-13 14:02:12 +0200 (Tue, 13 Jun 2006) | 2 lines Fix passing errors to the encoder and decoder functions. ........ r46917 | walter.doerwald | 2006-06-13 14:04:43 +0200 (Tue, 13 Jun 2006) | 3 lines errors is an attribute in the incremental decoder not an argument. ........ r46919 | andrew.macintyre | 2006-06-13 17:04:24 +0200 (Tue, 13 Jun 2006) | 11 lines Patch #1454481: Make thread stack size runtime tunable. Heavily revised, comprising revisions: 46640 - original trunk revision (backed out in r46655) 46647 - markup fix (backed out in r46655) 46692:46918 merged from branch aimacintyre-sf1454481 branch tested on buildbots (Windows buildbots had problems not related to these changes). ........ r46920 | brett.cannon | 2006-06-13 18:06:55 +0200 (Tue, 13 Jun 2006) | 2 lines Remove unused variable. ........ r46921 | andrew.kuchling | 2006-06-13 18:41:41 +0200 (Tue, 13 Jun 2006) | 1 line Add ability to set stack size ........ r46923 | marc-andre.lemburg | 2006-06-13 19:04:26 +0200 (Tue, 13 Jun 2006) | 2 lines Update pybench to version 2.0. ........ r46924 | marc-andre.lemburg | 2006-06-13 19:07:14 +0200 (Tue, 13 Jun 2006) | 2 lines Revert wrong svn copy. ........ r46925 | andrew.macintyre | 2006-06-13 19:14:36 +0200 (Tue, 13 Jun 2006) | 2 lines fix exception usage ........ r46927 | tim.peters | 2006-06-13 20:37:07 +0200 (Tue, 13 Jun 2006) | 2 lines Whitespace normalization. ........ r46928 | marc-andre.lemburg | 2006-06-13 20:56:56 +0200 (Tue, 13 Jun 2006) | 9 lines Updated to pybench 2.0. See svn.python.org/external/pybench-2.0 for the original import of that version. Note that platform.py was not copied over from pybench-2.0 since it is already part of Python 2.5. ........ r46929 | andrew.macintyre | 2006-06-13 21:02:35 +0200 (Tue, 13 Jun 2006) | 5 lines Increase the small thread stack size to get the test to pass reliably on the one buildbot that insists on more than 32kB of thread stack. ........ r46930 | marc-andre.lemburg | 2006-06-13 21:20:07 +0200 (Tue, 13 Jun 2006) | 2 lines Whitespace normalization. ........ r46931 | thomas.heller | 2006-06-13 22:18:43 +0200 (Tue, 13 Jun 2006) | 2 lines More docs for ctypes. ........ r46932 | brett.cannon | 2006-06-13 23:34:24 +0200 (Tue, 13 Jun 2006) | 2 lines Ignore .pyc and .pyo files in Pybench. ........ r46933 | brett.cannon | 2006-06-13 23:46:41 +0200 (Tue, 13 Jun 2006) | 7 lines If a classic class defined a __coerce__() method that just returned its two arguments in reverse, the interpreter would infinitely recourse trying to get a coercion that worked. So put in a recursion check after a coercion is made and the next call to attempt to use the coerced values. Fixes bug #992017 and closes crashers/coerce.py . ........ r46936 | gerhard.haering | 2006-06-14 00:24:47 +0200 (Wed, 14 Jun 2006) | 3 lines Merged changes from external pysqlite 2.3.0 release. Documentation updates will follow in a few hours at the latest. Then we should be ready for beta1. ........ r46937 | brett.cannon | 2006-06-14 00:26:13 +0200 (Wed, 14 Jun 2006) | 2 lines Missed test for rev. 46933; infinite recursion from __coerce__() returning its arguments reversed. ........ r46938 | gerhard.haering | 2006-06-14 00:53:48 +0200 (Wed, 14 Jun 2006) | 2 lines Updated documentation for pysqlite 2.3.0 API. ........ r46939 | tim.peters | 2006-06-14 06:09:25 +0200 (Wed, 14 Jun 2006) | 10 lines SequenceMatcher.get_matching_blocks(): This now guarantees that adjacent triples in the result list describe non-adjacent matching blocks. That's _nice_ to have, and Guido said he wanted it. Not a bugfix candidate: Guido or not ;-), this changes visible endcase semantics (note that some tests had to change), and nothing about this was documented before. Since it was working as designed, and behavior was consistent with the docs, it wasn't "a bug". ........ r46940 | tim.peters | 2006-06-14 06:13:00 +0200 (Wed, 14 Jun 2006) | 2 lines Repaired typo in new comment. ........ r46941 | tim.peters | 2006-06-14 06:15:27 +0200 (Wed, 14 Jun 2006) | 2 lines Whitespace normalization. ........ r46942 | fred.drake | 2006-06-14 06:25:02 +0200 (Wed, 14 Jun 2006) | 3 lines - make some disabled tests run what they intend when enabled - remove some over-zealous triple-quoting ........ r46943 | fred.drake | 2006-06-14 07:04:47 +0200 (Wed, 14 Jun 2006) | 3 lines add tests for two cases that are handled correctly in the current code, but that SF patch 1504676 as written mis-handles ........ r46944 | fred.drake | 2006-06-14 07:15:51 +0200 (Wed, 14 Jun 2006) | 1 line explain an XXX in more detail ........ r46945 | martin.v.loewis | 2006-06-14 07:21:04 +0200 (Wed, 14 Jun 2006) | 1 line Patch #1455898: Incremental mode for "mbcs" codec. ........ r46946 | georg.brandl | 2006-06-14 08:08:31 +0200 (Wed, 14 Jun 2006) | 3 lines Bug #1339007: Shelf objects now don't raise an exception in their __del__ method when initialization failed. ........ r46948 | thomas.heller | 2006-06-14 08:18:15 +0200 (Wed, 14 Jun 2006) | 1 line Fix docstring. ........ r46949 | georg.brandl | 2006-06-14 08:29:07 +0200 (Wed, 14 Jun 2006) | 2 lines Bug #1501122: mention __gt__ &co in description of comparison order. ........ r46951 | thomas.heller | 2006-06-14 09:08:38 +0200 (Wed, 14 Jun 2006) | 1 line Write more docs. ........ r46952 | georg.brandl | 2006-06-14 10:31:39 +0200 (Wed, 14 Jun 2006) | 3 lines Bug #1153163: describe __add__ vs __radd__ behavior when adding objects of same type/of subclasses of the other. ........ r46954 | georg.brandl | 2006-06-14 10:42:11 +0200 (Wed, 14 Jun 2006) | 3 lines Bug #1202018: add some common mime.types locations. ........ r46955 | georg.brandl | 2006-06-14 10:50:03 +0200 (Wed, 14 Jun 2006) | 3 lines Bug #1117556: SimpleHTTPServer now tries to find and use the system's mime.types file for determining MIME types. ........ r46957 | thomas.heller | 2006-06-14 11:09:08 +0200 (Wed, 14 Jun 2006) | 1 line Document paramflags. ........ r46958 | thomas.heller | 2006-06-14 11:20:11 +0200 (Wed, 14 Jun 2006) | 1 line Add an __all__ list, since this module does 'from ctypes import *'. ........ r46959 | andrew.kuchling | 2006-06-14 15:59:15 +0200 (Wed, 14 Jun 2006) | 1 line Add item ........ r46961 | georg.brandl | 2006-06-14 18:46:43 +0200 (Wed, 14 Jun 2006) | 3 lines Bug #805015: doc error in PyUnicode_FromEncodedObject. ........ r46962 | gerhard.haering | 2006-06-15 00:28:37 +0200 (Thu, 15 Jun 2006) | 10 lines - Added version checks in C code to make sure we don't trigger bugs in older SQLite versions. - Added version checks in test suite so that we don't execute tests that we know will fail with older (buggy) SQLite versions. Now, all tests should run against all SQLite versions from 3.0.8 until 3.3.6 (latest one now). The sqlite3 module can be built against all these SQLite versions and the sqlite3 module does its best to not trigger bugs in SQLite, but using SQLite 3.3.3 or later is recommended. ........ r46963 | tim.peters | 2006-06-15 00:38:13 +0200 (Thu, 15 Jun 2006) | 2 lines Whitespace normalization. ........ r46964 | neal.norwitz | 2006-06-15 06:54:29 +0200 (Thu, 15 Jun 2006) | 9 lines Speculative checkin (requires approval of Gerhard Haering) This backs out the test changes in 46962 which prevented crashes by not running the tests via a version check. All the version checks added in that rev were removed from the tests. Code was added to the error handler in connection.c that seems to work with older versions of sqlite including 3.1.3. ........ r46965 | neal.norwitz | 2006-06-15 07:55:49 +0200 (Thu, 15 Jun 2006) | 1 line Try to narrow window of failure on slow/busy boxes (ppc64 buildbot) ........ r46966 | martin.v.loewis | 2006-06-15 08:45:05 +0200 (Thu, 15 Jun 2006) | 2 lines Make import/lookup of mbcs fail on non-Windows systems. ........ r46967 | ronald.oussoren | 2006-06-15 10:14:18 +0200 (Thu, 15 Jun 2006) | 2 lines Patch #1446489 (zipfile: support for ZIP64) ........ r46968 | neal.norwitz | 2006-06-15 10:16:44 +0200 (Thu, 15 Jun 2006) | 6 lines Re-revert this change. Install the version check and don't run the test until Gerhard has time to fully debug the issue. This affects versions before 3.2.1 (possibly only versions earlier than 3.1.3). Based on discussion on python-checkins. ........ r46969 | gregory.p.smith | 2006-06-15 10:52:32 +0200 (Thu, 15 Jun 2006) | 6 lines - bsddb: multithreaded DB access using the simple bsddb module interface now works reliably. It has been updated to use automatic BerkeleyDB deadlock detection and the bsddb.dbutils.DeadlockWrap wrapper to retry database calls that would previously deadlock. [SF python bug #775414] ........ r46970 | gregory.p.smith | 2006-06-15 11:23:52 +0200 (Thu, 15 Jun 2006) | 2 lines minor documentation cleanup. mention the bsddb.db interface explicitly by name. ........ r46971 | neal.norwitz | 2006-06-15 11:57:03 +0200 (Thu, 15 Jun 2006) | 5 lines Steal the trick from test_compiler to print out a slow msg. This will hopefully get the buildbots to pass. Not sure this test will be feasible or even work. But everything is red now, so it can't get much worse. ........ r46972 | neal.norwitz | 2006-06-15 12:24:49 +0200 (Thu, 15 Jun 2006) | 1 line Print some more info to get an idea of how much longer the test will last ........ r46981 | tim.peters | 2006-06-15 20:04:40 +0200 (Thu, 15 Jun 2006) | 6 lines Try to reduce the extreme peak memory and disk-space use of this test. It probably still requires more disk space than most buildbots have, and in any case is still so intrusive that if we don't find another way to test this I'm taking my buildbot offline permanently ;-) ........ r46982 | tim.peters | 2006-06-15 20:06:29 +0200 (Thu, 15 Jun 2006) | 2 lines Whitespace normalization. ........ r46983 | tim.peters | 2006-06-15 20:07:28 +0200 (Thu, 15 Jun 2006) | 2 lines Add missing svn:eol-style property to text files. ........ r46984 | tim.peters | 2006-06-15 20:38:19 +0200 (Thu, 15 Jun 2006) | 2 lines Oops -- I introduced an off-by-6436159488 error. ........ r46990 | neal.norwitz | 2006-06-16 06:30:34 +0200 (Fri, 16 Jun 2006) | 1 line Disable this test until we can determine what to do about it ........ r46991 | neal.norwitz | 2006-06-16 06:31:06 +0200 (Fri, 16 Jun 2006) | 1 line Param name is dir, not directory. Update docstring. Backport candidate ........ r46992 | neal.norwitz | 2006-06-16 06:31:28 +0200 (Fri, 16 Jun 2006) | 1 line Add missing period in comment. ........ r46993 | neal.norwitz | 2006-06-16 06:32:43 +0200 (Fri, 16 Jun 2006) | 1 line Fix whitespace, there are memory leaks in this module. ........ r46995 | fred.drake | 2006-06-17 01:45:06 +0200 (Sat, 17 Jun 2006) | 3 lines SF patch 1504676: Make sgmllib char and entity references pluggable (implementation/tests contributed by Sam Ruby) ........ r46996 | fred.drake | 2006-06-17 03:07:54 +0200 (Sat, 17 Jun 2006) | 1 line fix change that broke the htmllib tests ........ r46998 | martin.v.loewis | 2006-06-17 11:15:14 +0200 (Sat, 17 Jun 2006) | 3 lines Patch #763580: Add name and value arguments to Tkinter variable classes. ........ r46999 | martin.v.loewis | 2006-06-17 11:20:41 +0200 (Sat, 17 Jun 2006) | 2 lines Patch #1096231: Add default argument to wm_iconbitmap. ........ r47000 | martin.v.loewis | 2006-06-17 11:25:15 +0200 (Sat, 17 Jun 2006) | 2 lines Patch #1494750: Destroy master after deleting children. ........ r47003 | george.yoshida | 2006-06-17 18:31:52 +0200 (Sat, 17 Jun 2006) | 2 lines markup fix ........ r47005 | george.yoshida | 2006-06-17 18:39:13 +0200 (Sat, 17 Jun 2006) | 4 lines Update url. Old url returned status code:301 Moved permanently. ........ r47007 | martin.v.loewis | 2006-06-17 20:44:27 +0200 (Sat, 17 Jun 2006) | 2 lines Patch #812986: Update the canvas even if not tracing. ........ r47008 | martin.v.loewis | 2006-06-17 21:03:26 +0200 (Sat, 17 Jun 2006) | 2 lines Patch #815924: Restore ability to pass type= and icon= ........ r47009 | neal.norwitz | 2006-06-18 00:37:45 +0200 (Sun, 18 Jun 2006) | 1 line Fix typo in docstring ........ r47010 | neal.norwitz | 2006-06-18 00:38:15 +0200 (Sun, 18 Jun 2006) | 1 line Fix memory leak reported by valgrind while running test_subprocess ........ r47011 | fred.drake | 2006-06-18 04:57:35 +0200 (Sun, 18 Jun 2006) | 1 line remove unnecessary markup ........ r47013 | neal.norwitz | 2006-06-18 21:35:01 +0200 (Sun, 18 Jun 2006) | 7 lines Prevent spurious leaks when running regrtest.py -R. There may be more issues that crop up from time to time, but this change seems to have been pretty stable (no spurious warnings) for about a week. Other modules which use threads may require similar use of threading_setup/threading_cleanup from test_support. ........ r47014 | neal.norwitz | 2006-06-18 21:37:40 +0200 (Sun, 18 Jun 2006) | 9 lines The hppa ubuntu box sometimes hangs forever in these tests. My guess is that the wait is failing for some reason. Use WNOHANG, so we won't wait until the buildbot kills the test suite. I haven't been able to reproduce the failure, so I'm not sure if this will help or not. Hopefully, this change will cause the test to fail, rather than hang. That will be better since we will get the rest of the test results. It may also help us debug the real problem. ........ r47015 | neal.norwitz | 2006-06-18 22:10:24 +0200 (Sun, 18 Jun 2006) | 1 line Revert 47014 until it is more robust ........ r47016 | thomas.heller | 2006-06-18 23:27:04 +0200 (Sun, 18 Jun 2006) | 6 lines Fix typos. Fix doctest example. Mention in the tutorial that 'errcheck' is explained in the ref manual. Use better wording in some places. Remoce code examples that shouldn't be in the tutorial. Remove some XXX notices. ........ r47017 | georg.brandl | 2006-06-19 00:17:29 +0200 (Mon, 19 Jun 2006) | 3 lines Patch #1507676: improve exception messages in abstract.c, object.c and typeobject.c. ........ r47018 | neal.norwitz | 2006-06-19 07:40:44 +0200 (Mon, 19 Jun 2006) | 1 line Use Py_ssize_t ........ r47019 | georg.brandl | 2006-06-19 08:35:54 +0200 (Mon, 19 Jun 2006) | 3 lines Add news entry about error msg improvement. ........ r47020 | thomas.heller | 2006-06-19 09:07:49 +0200 (Mon, 19 Jun 2006) | 2 lines Try to repair the failing test on the OpenBSD buildbot. Trial and error... ........ r47021 | tim.peters | 2006-06-19 09:45:16 +0200 (Mon, 19 Jun 2006) | 2 lines Whitespace normalization. ........ r47022 | walter.doerwald | 2006-06-19 10:07:50 +0200 (Mon, 19 Jun 2006) | 4 lines Patch #1506645: add Python wrappers for the curses functions is_term_resized, resize_term and resizeterm. This uses three separate configure checks (one for each function). ........ r47023 | walter.doerwald | 2006-06-19 10:14:09 +0200 (Mon, 19 Jun 2006) | 2 lines Make check order match in configure and configure.in. ........ r47024 | tim.peters | 2006-06-19 10:14:28 +0200 (Mon, 19 Jun 2006) | 3 lines Repair KeyError when running test_threaded_import under -R, as reported by Neal on python-dev. ........ r47025 | thomas.heller | 2006-06-19 10:32:46 +0200 (Mon, 19 Jun 2006) | 3 lines Next try to fix the OpenBSD buildbot tests: Use ctypes.util.find_library to locate the C runtime library on platforms where is returns useful results. ........ r47026 | tim.peters | 2006-06-19 11:09:44 +0200 (Mon, 19 Jun 2006) | 13 lines TestHelp.make_parser(): This was making a permanent change to os.environ (setting envar COLUMNS), which at least caused test_float_default() to fail if the tests were run more than once. This repairs the test_optparse -R failures Neal reported on python-dev. It also explains some seemingly bizarre test_optparse failures we saw a couple weeks ago on the buildbots, when test_optparse failed due to test_file failing to clean up after itself, and then test_optparse failed in an entirely different way when regrtest's -w option ran test_optparse a second time. It's now obvious that make_parser() permanently changing os.environ was responsible for the second half of that. ........ r47027 | anthony.baxter | 2006-06-19 14:04:15 +0200 (Mon, 19 Jun 2006) | 2 lines Preparing for 2.5b1. ........ r47029 | fred.drake | 2006-06-19 19:31:16 +0200 (Mon, 19 Jun 2006) | 1 line remove non-working document formats from edist ........ r47030 | gerhard.haering | 2006-06-19 23:17:35 +0200 (Mon, 19 Jun 2006) | 5 lines Fixed a memory leak that was introduced with incorrect usage of the Python weak reference API in pysqlite 2.2.1. Bumbed pysqlite version number to upcoming pysqlite 2.3.1 release. ........ r47032 | ka-ping.yee | 2006-06-20 00:49:36 +0200 (Tue, 20 Jun 2006) | 2 lines Remove Python 2.3 compatibility comment. ........ r47033 | trent.mick | 2006-06-20 01:21:25 +0200 (Tue, 20 Jun 2006) | 2 lines Upgrade pyexpat to expat 2.0.0 (http://python.org/sf/1462338). ........ r47034 | trent.mick | 2006-06-20 01:57:41 +0200 (Tue, 20 Jun 2006) | 3 lines [ 1295808 ] expat symbols should be namespaced in pyexpat (http://python.org/sf/1295808) ........ r47039 | andrew.kuchling | 2006-06-20 13:52:16 +0200 (Tue, 20 Jun 2006) | 1 line Uncomment wsgiref section ........ r47040 | andrew.kuchling | 2006-06-20 14:15:09 +0200 (Tue, 20 Jun 2006) | 1 line Add four library items ........ r47041 | andrew.kuchling | 2006-06-20 14:19:54 +0200 (Tue, 20 Jun 2006) | 1 line Terminology and typography fixes ........ r47042 | andrew.kuchling | 2006-06-20 15:05:12 +0200 (Tue, 20 Jun 2006) | 1 line Add introductory paragraphs summarizing the release; minor edits ........ r47043 | andrew.kuchling | 2006-06-20 15:11:29 +0200 (Tue, 20 Jun 2006) | 1 line Minor edits and rearrangements; markup fix ........ r47044 | andrew.kuchling | 2006-06-20 15:20:30 +0200 (Tue, 20 Jun 2006) | 1 line [Bug #1504456] Mention xml -> xmlcore change ........ r47047 | brett.cannon | 2006-06-20 19:30:26 +0200 (Tue, 20 Jun 2006) | 2 lines Raise TestSkipped when the test socket connection is refused. ........ r47049 | brett.cannon | 2006-06-20 21:20:17 +0200 (Tue, 20 Jun 2006) | 2 lines Fix typo of exception name. ........ r47053 | brett.cannon | 2006-06-21 18:57:57 +0200 (Wed, 21 Jun 2006) | 5 lines At the C level, tuple arguments are passed in directly to the exception constructor, meaning it is treated as *args, not as a single argument. This means using the 'message' attribute won't work (until Py3K comes around), and so one must grab from 'arg' to get the error number. ........ r47054 | andrew.kuchling | 2006-06-21 19:10:18 +0200 (Wed, 21 Jun 2006) | 1 line Link to LibRef module documentation ........ r47055 | andrew.kuchling | 2006-06-21 19:17:10 +0200 (Wed, 21 Jun 2006) | 1 line Note some of Barry's work ........ r47056 | andrew.kuchling | 2006-06-21 19:17:28 +0200 (Wed, 21 Jun 2006) | 1 line Bump version ........ r47057 | georg.brandl | 2006-06-21 19:45:17 +0200 (Wed, 21 Jun 2006) | 3 lines fix [ 1509132 ] compiler module builds incorrect AST for TryExceptFinally ........ r47058 | georg.brandl | 2006-06-21 19:52:36 +0200 (Wed, 21 Jun 2006) | 3 lines Make test_fcntl aware of netbsd3. ........ r47059 | georg.brandl | 2006-06-21 19:53:17 +0200 (Wed, 21 Jun 2006) | 3 lines Patch #1509001: expected skips for netbsd3. ........ r47060 | gerhard.haering | 2006-06-21 22:55:04 +0200 (Wed, 21 Jun 2006) | 2 lines Removed call to enable_callback_tracebacks that slipped in by accident. ........ r47061 | armin.rigo | 2006-06-21 23:58:50 +0200 (Wed, 21 Jun 2006) | 13 lines Fix for an obscure bug introduced by revs 46806 and 46808, with a test. The problem of checking too eagerly for recursive calls is the following: if a RuntimeError is caused by recursion, and if code needs to normalize it immediately (as in the 2nd test), then PyErr_NormalizeException() needs a call to the RuntimeError class to instantiate it, and this hits the recursion limit again... causing PyErr_NormalizeException() to never finish. Moved this particular recursion check to slot_tp_call(), which is not involved in instantiating built-in exceptions. Backport candidate. ........ r47064 | neal.norwitz | 2006-06-22 08:30:50 +0200 (Thu, 22 Jun 2006) | 3 lines Copy the wsgiref package during make install. ........ r47065 | neal.norwitz | 2006-06-22 08:35:30 +0200 (Thu, 22 Jun 2006) | 1 line Reset the doc date to today for the automatic doc builds ........ r47067 | andrew.kuchling | 2006-06-22 15:10:23 +0200 (Thu, 22 Jun 2006) | 1 line Mention how to suppress warnings ........ r47069 | georg.brandl | 2006-06-22 16:46:17 +0200 (Thu, 22 Jun 2006) | 3 lines Set lineno correctly on list, tuple and dict literals. ........ r47070 | georg.brandl | 2006-06-22 16:46:46 +0200 (Thu, 22 Jun 2006) | 4 lines Test for correct compilation of try-except-finally stmt. Test for correct lineno on list, tuple, dict literals. ........ r47071 | fred.drake | 2006-06-22 17:50:08 +0200 (Thu, 22 Jun 2006) | 1 line fix markup nit ........ r47072 | brett.cannon | 2006-06-22 18:49:14 +0200 (Thu, 22 Jun 2006) | 6 lines 'warning's was improperly requiring that a command-line Warning category be both a subclass of Warning and a subclass of types.ClassType. The latter is no longer true thanks to new-style exceptions. Closes bug #1510580. Thanks to AMK for the test. ........ r47073 | ronald.oussoren | 2006-06-22 20:33:54 +0200 (Thu, 22 Jun 2006) | 3 lines MacOSX: Add a message to the first screen of the installer that tells users how to avoid updates to their shell profile. ........ r47074 | georg.brandl | 2006-06-22 21:02:18 +0200 (Thu, 22 Jun 2006) | 3 lines Fix my name ;) ........ r47075 | thomas.heller | 2006-06-22 21:07:36 +0200 (Thu, 22 Jun 2006) | 2 lines Small fixes, mostly in the markup. ........ r47076 | peter.astrand | 2006-06-22 22:06:46 +0200 (Thu, 22 Jun 2006) | 1 line Make it possible to run test_subprocess.py on Python 2.2, which lacks test_support.is_resource_enabled. ........ r47077 | peter.astrand | 2006-06-22 22:21:26 +0200 (Thu, 22 Jun 2006) | 1 line Applied patch #1506758: Prevent MemoryErrors with large MAXFD. ........ r47079 | neal.norwitz | 2006-06-23 05:32:44 +0200 (Fri, 23 Jun 2006) | 1 line Fix refleak ........ r47080 | fred.drake | 2006-06-23 08:03:45 +0200 (Fri, 23 Jun 2006) | 9 lines - SF bug #853506: IP6 address parsing in sgmllib ('[' and ']' were not accepted in unquoted attribute values) - cleaned up tests of character and entity reference decoding so the tests cover the documented relationships among handle_charref, handle_entityref, convert_charref, convert_codepoint, and convert_entityref, without bringing up Unicode issues that sgmllib cannot be involved in ........ r47085 | andrew.kuchling | 2006-06-23 21:23:40 +0200 (Fri, 23 Jun 2006) | 11 lines Fit Makefile for the Python doc environment better; this is a step toward including the howtos in the build process. * Put LaTeX output in ../paper-<whatever>/. * Put HTML output in ../html/ * Explain some of the Makefile variables * Remove some cruft dating to my environment (e.g. the 'web' target) This makefile isn't currently invoked by the documentation build process, so these changes won't destabilize anything. ........ r47086 | hyeshik.chang | 2006-06-23 23:16:18 +0200 (Fri, 23 Jun 2006) | 5 lines Bug #1511381: codec_getstreamcodec() in codec.c is corrected to omit a default "error" argument for NULL pointer. This allows the parser to take a codec from cjkcodecs again. (Reported by Taewook Kang and reviewed by Walter Doerwald) ........ r47091 | ronald.oussoren | 2006-06-25 22:44:16 +0200 (Sun, 25 Jun 2006) | 6 lines Workaround for bug #1512124 Without this patch IDLE will get unresponsive when you open the debugger window on OSX. This is both using the system Tcl/Tk on Tiger as the latest universal download from tk-components.sf.net. ........ r47092 | ronald.oussoren | 2006-06-25 23:14:19 +0200 (Sun, 25 Jun 2006) | 3 lines Drop the calldll demo's for macos, calldll isn't present anymore, no need to keep the demo's around. ........ r47093 | ronald.oussoren | 2006-06-25 23:15:58 +0200 (Sun, 25 Jun 2006) | 3 lines Use a path without a double slash to compile the .py files after installation (macosx, binary installer). This fixes bug #1508369 for python 2.5. ........ r47094 | ronald.oussoren | 2006-06-25 23:19:06 +0200 (Sun, 25 Jun 2006) | 3 lines Also install the .egg-info files in Lib. This will cause wsgiref.egg-info to be installed. ........ r47097 | andrew.kuchling | 2006-06-26 14:40:02 +0200 (Mon, 26 Jun 2006) | 1 line [Bug #1511998] Various comments from Nick Coghlan; thanks! ........ r47098 | andrew.kuchling | 2006-06-26 14:43:43 +0200 (Mon, 26 Jun 2006) | 1 line Describe workaround for PyRange_New()'s removal ........ r47099 | andrew.kuchling | 2006-06-26 15:08:24 +0200 (Mon, 26 Jun 2006) | 5 lines [Bug #1512163] Fix typo. This change will probably break tests on FreeBSD buildbots, but I'll check in a fix for that next. ........ r47100 | andrew.kuchling | 2006-06-26 15:12:16 +0200 (Mon, 26 Jun 2006) | 9 lines [Bug #1512163] Use one set of locking methods, lockf(); remove the flock() calls. On FreeBSD, the two methods lockf() and flock() end up using the same mechanism and the second one fails. A Linux man page claims that the two methods are orthogonal (so locks acquired one way don't interact with locks acquired the other way) but that clearly must be false. ........ r47101 | andrew.kuchling | 2006-06-26 15:23:10 +0200 (Mon, 26 Jun 2006) | 5 lines Add a test for a conflicting lock. On slow machines, maybe the time intervals (2 sec, 0.5 sec) will be too tight. I'll see how the buildbots like it. ........ r47103 | andrew.kuchling | 2006-06-26 16:33:24 +0200 (Mon, 26 Jun 2006) | 1 line Windows doesn't have os.fork(). I'll just disable this test for now ........ r47106 | andrew.kuchling | 2006-06-26 19:00:35 +0200 (Mon, 26 Jun 2006) | 9 lines Attempt to fix build failure on OS X and Debian alpha; the symptom is consistent with os.wait() returning immediately because some other subprocess had previously exited; the test suite then immediately tries to lock the mailbox and gets an error saying it's already locked. To fix this, do a waitpid() so the test suite only continues once the intended child process has exited. ........ r47113 | neal.norwitz | 2006-06-27 06:06:46 +0200 (Tue, 27 Jun 2006) | 1 line Ignore some more warnings in the dynamic linker on an older gentoo ........ r47114 | neal.norwitz | 2006-06-27 06:09:13 +0200 (Tue, 27 Jun 2006) | 6 lines Instead of doing a make test, run the regression tests out of the installed copy. This will hopefully catch problems where directories are added under Lib/ but not to Makefile.pre.in. This breaks out the 2 runs of the test suite with and without -O which is also nicer. ........ r47115 | neal.norwitz | 2006-06-27 06:12:58 +0200 (Tue, 27 Jun 2006) | 5 lines Fix SF bug #1513032, 'make install' failure on FreeBSD 5.3. No need to install lib-old, it's empty in 2.5. ........ r47116 | neal.norwitz | 2006-06-27 06:23:06 +0200 (Tue, 27 Jun 2006) | 1 line Test unimportant change to verify buildbot does not try to build ........ r47117 | neal.norwitz | 2006-06-27 06:26:30 +0200 (Tue, 27 Jun 2006) | 1 line Try again: test unimportant change to verify buildbot does not try to build ........ r47118 | neal.norwitz | 2006-06-27 06:28:56 +0200 (Tue, 27 Jun 2006) | 1 line Verify buildbot picks up these changes (really needs testing after last change to Makefile.pre.in) ........ r47121 | vinay.sajip | 2006-06-27 09:34:37 +0200 (Tue, 27 Jun 2006) | 1 line Removed buggy exception handling in doRollover of rotating file handlers. Exceptions now propagate to caller. ........ r47123 | ronald.oussoren | 2006-06-27 12:08:25 +0200 (Tue, 27 Jun 2006) | 3 lines MacOSX: fix rather dumb buglet that made it impossible to create extensions on OSX 10.3 when using a binary distribution build on 10.4. ........ r47125 | tim.peters | 2006-06-27 13:52:49 +0200 (Tue, 27 Jun 2006) | 2 lines Whitespace normalization. ........ r47128 | ronald.oussoren | 2006-06-27 14:53:52 +0200 (Tue, 27 Jun 2006) | 8 lines Use staticly build copies of zlib and bzip2 to build the OSX installer, that way the resulting binaries have a better change of running on 10.3. This patch also updates the search logic for sleepycat db3/4, without this patch you cannot use a sleepycat build with a non-standard prefix; with this you can (at least on OSX) if you add the prefix to CPPFLAGS/LDFLAGS at configure-time. This change is needed to build the binary installer for OSX. ........ r47131 | ronald.oussoren | 2006-06-27 17:45:32 +0200 (Tue, 27 Jun 2006) | 5 lines macosx: Install a libpython2.5.a inside the framework as a symlink to the actual dylib at the root of the framework, that way tools that expect a unix-like install (python-config, but more importantly external products like mod_python) work correctly. ........ r47137 | neal.norwitz | 2006-06-28 07:03:22 +0200 (Wed, 28 Jun 2006) | 4 lines According to the man pages on Gentoo Linux and Tru64, EACCES or EAGAIN can be returned if fcntl (lockf) fails. This fixes the test failure on Tru64 by checking for either error rather than just EAGAIN. ........ r47139 | neal.norwitz | 2006-06-28 08:28:31 +0200 (Wed, 28 Jun 2006) | 5 lines Fix bug #1512695: cPickle.loads could crash if it was interrupted with a KeyboardInterrupt since PyTuple_Pack was passed a NULL. Will backport. ........ r47142 | nick.coghlan | 2006-06-28 12:41:47 +0200 (Wed, 28 Jun 2006) | 1 line Make full module name available as __module_name__ even when __name__ is set to something else (like '__main__') ........ r47143 | armin.rigo | 2006-06-28 12:49:51 +0200 (Wed, 28 Jun 2006) | 2 lines A couple of crashers of the "won't fix" kind. ........ r47147 | andrew.kuchling | 2006-06-28 16:25:20 +0200 (Wed, 28 Jun 2006) | 1 line [Bug #1508766] Add docs for uuid module; docs written by George Yoshida, with minor rearrangements by me. ........ r47148 | andrew.kuchling | 2006-06-28 16:27:21 +0200 (Wed, 28 Jun 2006) | 1 line [Bug #1508766] Add docs for uuid module; this puts the module in the 'Internet Protocols' section. Arguably this module could also have gone in the chapters on strings or encodings, maybe even the crypto chapter. Fred, please move if you see fit. ........ r47151 | georg.brandl | 2006-06-28 22:23:25 +0200 (Wed, 28 Jun 2006) | 3 lines Fix end_fill(). ........ r47153 | trent.mick | 2006-06-28 22:30:41 +0200 (Wed, 28 Jun 2006) | 2 lines Mention the expat upgrade and pyexpat fix I put in 2.5b1. ........ r47154 | fred.drake | 2006-06-29 02:51:53 +0200 (Thu, 29 Jun 2006) | 6 lines SF bug #1504333: sgmlib should allow angle brackets in quoted values (modified patch by Sam Ruby; changed to use separate REs for start and end tags to reduce matching cost for end tags; extended tests; updated to avoid breaking previous changes to support IPv6 addresses in unquoted attribute values) ........ r47156 | fred.drake | 2006-06-29 04:57:48 +0200 (Thu, 29 Jun 2006) | 1 line document recent bugfixes in sgmllib ........ r47158 | neal.norwitz | 2006-06-29 06:10:08 +0200 (Thu, 29 Jun 2006) | 10 lines Add new utility function, reap_children(), to test_support. This should be called at the end of each test that spawns children (perhaps it should be called from regrtest instead?). This will hopefully prevent some of the unexplained failures in the buildbots (hppa and alpha) during tests that spawn children. The problems were not reproducible. There were many zombies that remained at the end of several tests. In the worst case, this shouldn't cause any more problems, though it may not help either. Time will tell. ........ r47159 | neal.norwitz | 2006-06-29 07:48:14 +0200 (Thu, 29 Jun 2006) | 5 lines This should fix the buildbot failure on s/390 which can't connect to gmail.org. It makes the error message consistent and always sends to stderr. It would be much better for all the networking tests to hit only python.org. ........ r47161 | thomas.heller | 2006-06-29 20:34:15 +0200 (Thu, 29 Jun 2006) | 3 lines Protect the thread api calls in the _ctypes extension module within #ifdef WITH_THREADS/#endif blocks. Found by Sam Rushing. ........ r47162 | martin.v.loewis | 2006-06-29 20:58:44 +0200 (Thu, 29 Jun 2006) | 2 lines Patch #1509163: MS Toolkit Compiler no longer available ........ r47163 | skip.montanaro | 2006-06-29 21:20:09 +0200 (Thu, 29 Jun 2006) | 1 line add string methods to index ........ r47164 | vinay.sajip | 2006-06-30 02:13:08 +0200 (Fri, 30 Jun 2006) | 1 line Fixed bug in fileConfig() which failed to clear logging._handlerList ........ r47166 | tim.peters | 2006-06-30 08:18:39 +0200 (Fri, 30 Jun 2006) | 2 lines Whitespace normalization. ........ r47170 | neal.norwitz | 2006-06-30 09:32:16 +0200 (Fri, 30 Jun 2006) | 1 line Silence compiler warning ........ r47171 | neal.norwitz | 2006-06-30 09:32:46 +0200 (Fri, 30 Jun 2006) | 1 line Another problem reported by Coverity. Backport candidate. ........ r47175 | thomas.heller | 2006-06-30 19:44:54 +0200 (Fri, 30 Jun 2006) | 2 lines Revert the use of PY_FORMAT_SIZE_T in PyErr_Format. ........ r47176 | tim.peters | 2006-06-30 20:34:51 +0200 (Fri, 30 Jun 2006) | 2 lines Remove now-unused fidding with PY_FORMAT_SIZE_T. ........ r47177 | georg.brandl | 2006-06-30 20:47:56 +0200 (Fri, 30 Jun 2006) | 3 lines Document decorator usage of property. ........ r47181 | fred.drake | 2006-06-30 21:29:25 +0200 (Fri, 30 Jun 2006) | 4 lines - consistency nit: always include "()" in \function and \method (*should* be done by the presentation, but that requires changes all over) - avoid spreading the __name meme ........ r47188 | vinay.sajip | 2006-07-01 12:45:20 +0200 (Sat, 01 Jul 2006) | 1 line Added entry for fileConfig() bugfix. ........ r47189 | vinay.sajip | 2006-07-01 12:47:20 +0200 (Sat, 01 Jul 2006) | 1 line Added duplicate call to fileConfig() to ensure that it cleans up after itself correctly. ........ r47190 | martin.v.loewis | 2006-07-01 17:33:37 +0200 (Sat, 01 Jul 2006) | 2 lines Release all forwarded functions in .close. Fixes #1513223. ........ r47191 | fred.drake | 2006-07-01 18:28:20 +0200 (Sat, 01 Jul 2006) | 7 lines SF bug #1296433 (Expat bug #1515266): Unchecked calls to character data handler would cause a segfault. This merges in Expat's lib/xmlparse.c revisions 1.154 and 1.155, which fix this and a closely related problem (the later does not affect Python). Moved the crasher test to the tests for xml.parsers.expat. ........ r47197 | gerhard.haering | 2006-07-02 19:48:30 +0200 (Sun, 02 Jul 2006) | 4 lines The sqlite3 module did cut off data from the SQLite database at the first null character before sending it to a custom converter. This has been fixed now. ........ r47198 | martin.v.loewis | 2006-07-02 20:44:00 +0200 (Sun, 02 Jul 2006) | 1 line Correct arithmetic in access on Win32. Fixes #1513646. ........ r47203 | thomas.heller | 2006-07-03 09:58:09 +0200 (Mon, 03 Jul 2006) | 1 line Cleanup: Remove commented out code. ........ r47204 | thomas.heller | 2006-07-03 09:59:50 +0200 (Mon, 03 Jul 2006) | 1 line Don't run the doctests with Python 2.3 because it doesn't have the ELLIPSIS flag. ........ r47205 | thomas.heller | 2006-07-03 10:04:05 +0200 (Mon, 03 Jul 2006) | 7 lines Fixes so that _ctypes can be compiled with the MingW compiler. It seems that the definition of '__attribute__(x)' was responsible for the compiler ignoring the '__fastcall' attribute on the ffi_closure_SYSV function in libffi_msvc/ffi.c, took me quite some time to figure this out. ........ r47206 | thomas.heller | 2006-07-03 10:08:14 +0200 (Mon, 03 Jul 2006) | 11 lines Add a new function uses_seh() to the _ctypes extension module. This will return True if Windows Structured Exception handling (SEH) is used when calling functions, False otherwise. Currently, only MSVC supports SEH. Fix the test so that it doesn't crash when run with MingW compiled _ctypes. Note that two tests are still failing when mingw is used, I suspect structure layout differences and function calling conventions between MSVC and MingW. ........ r47207 | tim.peters | 2006-07-03 10:23:19 +0200 (Mon, 03 Jul 2006) | 2 lines Whitespace normalization. ........ r47208 | martin.v.loewis | 2006-07-03 11:44:00 +0200 (Mon, 03 Jul 2006) | 3 lines Only setup canvas when it is first created. Fixes #1514703 ........ r47209 | martin.v.loewis | 2006-07-03 12:05:30 +0200 (Mon, 03 Jul 2006) | 3 lines Reimplement turtle.circle using a polyline, to allow correct filling of arcs. Also fixes #1514693. ........ r47210 | martin.v.loewis | 2006-07-03 12:19:49 +0200 (Mon, 03 Jul 2006) | 3 lines Bug #1514693: Update turtle's heading when switching between degrees and radians. ........ r47211 | martin.v.loewis | 2006-07-03 13:12:06 +0200 (Mon, 03 Jul 2006) | 2 lines Document functions added in 2.3 and 2.5. ........ r47212 | martin.v.loewis | 2006-07-03 14:19:50 +0200 (Mon, 03 Jul 2006) | 3 lines Bug #1417699: Reject locale-specific decimal point in float() and atof(). ........ r47213 | martin.v.loewis | 2006-07-03 14:28:58 +0200 (Mon, 03 Jul 2006) | 3 lines Bug #1267547: Put proper recursive setup.py call into the spec file generated by bdist_rpm. ........ r47215 | martin.v.loewis | 2006-07-03 15:01:35 +0200 (Mon, 03 Jul 2006) | 3 lines Patch #825417: Fix timeout processing in expect, read_until. Will backport to 2.4. ........ r47218 | martin.v.loewis | 2006-07-03 15:47:40 +0200 (Mon, 03 Jul 2006) | 2 lines Put method-wrappers into trashcan. Fixes #927248. ........ r47219 | andrew.kuchling | 2006-07-03 16:07:30 +0200 (Mon, 03 Jul 2006) | 1 line [Bug #1515932] Clarify description of slice assignment ........ r47220 | andrew.kuchling | 2006-07-03 16:16:09 +0200 (Mon, 03 Jul 2006) | 4 lines [Bug #1511911] Clarify description of optional arguments to sorted() by improving the xref to the section on lists, and by copying the explanations of the arguments (with a slight modification). ........ r47223 | kristjan.jonsson | 2006-07-03 16:59:05 +0200 (Mon, 03 Jul 2006) | 1 line Fix build problems with the platform SDK on windows. It is not sufficient to test for the C compiler version when determining if we have the secure CRT from microsoft. Must test with an undocumented macro, __STDC_SECURE_LIB__ too. ........ r47224 | ronald.oussoren | 2006-07-04 14:30:22 +0200 (Tue, 04 Jul 2006) | 7 lines Sync the darwin/x86 port libffi with the copy in PyObjC. This fixes a number of bugs in that port. The most annoying ones were due to some subtle differences between the document ABI and the actual implementation :-( (there are no python unittests that fail without this patch, but without it some of libffi's unittests fail). ........ r47234 | georg.brandl | 2006-07-05 10:21:00 +0200 (Wed, 05 Jul 2006) | 3 lines Remove remaining references to OverflowWarning. ........ r47236 | thomas.heller | 2006-07-05 11:13:56 +0200 (Wed, 05 Jul 2006) | 3 lines Fix the bitfield test when _ctypes is compiled with MingW. Structures containing bitfields may have different layout on MSVC and MingW . ........ r47237 | thomas.wouters | 2006-07-05 13:03:49 +0200 (Wed, 05 Jul 2006) | 15 lines Fix bug in passing tuples to string.Template. All other values (with working str() or repr()) would work, just not multi-value tuples. Probably not a backport candidate, since it changes the behaviour of passing a single-element tuple: >>> string.Template("$foo").substitute(dict(foo=(1,))) '(1,)' versus '1' ........ r47241 | georg.brandl | 2006-07-05 16:18:45 +0200 (Wed, 05 Jul 2006) | 2 lines Patch #1517490: fix glitches in filter() docs. ........ r47244 | georg.brandl | 2006-07-05 17:50:05 +0200 (Wed, 05 Jul 2006) | 2 lines no need to elaborate "string". ........ r47251 | neal.norwitz | 2006-07-06 06:28:59 +0200 (Thu, 06 Jul 2006) | 3 lines Fix refleaks reported by Shane Hathaway in SF patch #1515361. This change contains only the changes related to leaking the copy variable. ........ r47253 | fred.drake | 2006-07-06 07:13:22 +0200 (Thu, 06 Jul 2006) | 4 lines - back out Expat change; the final fix to Expat will be different - change the pyexpat wrapper to not be so sensitive to this detail of the Expat implementation (the ex-crasher test still passes) ........ r47257 | neal.norwitz | 2006-07-06 08:45:08 +0200 (Thu, 06 Jul 2006) | 1 line Add a NEWS entry for a recent pyexpat fix ........ r47258 | martin.v.loewis | 2006-07-06 08:55:58 +0200 (Thu, 06 Jul 2006) | 2 lines Add sqlite3.dll to the DLLs component, not to the TkDLLs component. Fixes #1517388. ........ r47259 | martin.v.loewis | 2006-07-06 09:05:21 +0200 (Thu, 06 Jul 2006) | 1 line Properly quote compileall and Lib paths in case TARGETDIR has a space. ........ r47260 | thomas.heller | 2006-07-06 09:50:18 +0200 (Thu, 06 Jul 2006) | 5 lines Revert the change done in svn revision 47206: Add a new function uses_seh() to the _ctypes extension module. This will return True if Windows Structured Exception handling (SEH) is used when calling functions, False otherwise. ........ r47261 | armin.rigo | 2006-07-06 09:58:18 +0200 (Thu, 06 Jul 2006) | 3 lines A couple of examples about how to attack the fact that _PyType_Lookup() returns a borrowed ref. Many of the calls are open to attack. ........ r47262 | thomas.heller | 2006-07-06 10:28:14 +0200 (Thu, 06 Jul 2006) | 2 lines The test that calls a function with invalid arguments and catches the resulting Windows access violation will not be run by default. ........ r47263 | thomas.heller | 2006-07-06 10:48:35 +0200 (Thu, 06 Jul 2006) | 5 lines Patch #1517790: It is now possible to use custom objects in the ctypes foreign function argtypes sequence as long as they provide a from_param method, no longer is it required that the object is a ctypes type. ........ r47264 | thomas.heller | 2006-07-06 10:58:40 +0200 (Thu, 06 Jul 2006) | 2 lines Document the Struture and Union constructors. ........ r47265 | thomas.heller | 2006-07-06 11:11:22 +0200 (Thu, 06 Jul 2006) | 2 lines Document the changes in svn revision 47263, from patch #1517790. ........ r47267 | ronald.oussoren | 2006-07-06 12:13:35 +0200 (Thu, 06 Jul 2006) | 7 lines This patch solves the problem Skip was seeing with zlib, this patch ensures that configure uses similar compiler flags as setup.py when doing the zlib test. Without this patch configure would use the first shared library on the linker path, with this patch it uses the first shared or static library on that path just like setup.py. ........ r47268 | thomas.wouters | 2006-07-06 12:48:28 +0200 (Thu, 06 Jul 2006) | 4 lines NEWS entry for r47267: fixing configure's zlib probing. ........ r47269 | fredrik.lundh | 2006-07-06 14:29:24 +0200 (Thu, 06 Jul 2006) | 3 lines added XMLParser alias for cElementTree compatibility ........ r47271 | nick.coghlan | 2006-07-06 14:53:04 +0200 (Thu, 06 Jul 2006) | 1 line Revert the __module_name__ changes made in rev 47142. We'll revisit this in Python 2.6 ........ r47272 | nick.coghlan | 2006-07-06 15:04:56 +0200 (Thu, 06 Jul 2006) | 1 line Update the tutorial section on relative imports ........ r47273 | nick.coghlan | 2006-07-06 15:35:27 +0200 (Thu, 06 Jul 2006) | 1 line Ignore ImportWarning by default ........ r47274 | nick.coghlan | 2006-07-06 15:41:34 +0200 (Thu, 06 Jul 2006) | 1 line Cover ImportWarning, PendingDeprecationWarning and simplefilter() in the warnings module docs ........ r47275 | nick.coghlan | 2006-07-06 15:47:18 +0200 (Thu, 06 Jul 2006) | 1 line Add NEWS entries for the ImportWarning change and documentation update ........ r47276 | andrew.kuchling | 2006-07-06 15:57:28 +0200 (Thu, 06 Jul 2006) | 1 line ImportWarning is now silent by default ........ r47277 | thomas.heller | 2006-07-06 17:06:05 +0200 (Thu, 06 Jul 2006) | 2 lines Document the correct return type of PyLong_AsUnsignedLongLongMask. ........ r47278 | hyeshik.chang | 2006-07-06 17:21:52 +0200 (Thu, 06 Jul 2006) | 2 lines Add a testcase for r47086 which fixed a bug in codec_getstreamcodec(). ........ r47279 | hyeshik.chang | 2006-07-06 17:39:24 +0200 (Thu, 06 Jul 2006) | 3 lines Test using all CJK encodings for the testcases which don't require specific encodings. ........ r47280 | martin.v.loewis | 2006-07-06 21:28:03 +0200 (Thu, 06 Jul 2006) | 2 lines Properly generate logical file ids. Fixes #1515998. Also correct typo in Control.mapping. ........ r47287 | neal.norwitz | 2006-07-07 08:03:15 +0200 (Fri, 07 Jul 2006) | 17 lines Restore rev 47014: The hppa ubuntu box sometimes hangs forever in these tests. My guess is that the wait is failing for some reason. Use WNOHANG, so we won't wait until the buildbot kills the test suite. I haven't been able to reproduce the failure, so I'm not sure if this will help or not. Hopefully, this change will cause the test to fail, rather than hang. That will be better since we will get the rest of the test results. It may also help us debug the real problem. *** The reason this originally failed was because there were many zombie children outstanding before rev 47158 cleaned them up. There are still hangs in test_subprocess that need to be addressed, but that will take more work. This should close some holes. ........ r47289 | georg.brandl | 2006-07-07 10:15:12 +0200 (Fri, 07 Jul 2006) | 3 lines Fix RFC number. ........ r50489 | neal.norwitz | 2006-07-08 07:31:37 +0200 (Sat, 08 Jul 2006) | 1 line Fix SF bug #1519018: 'as' is now validated properly in import statements ........ r50490 | georg.brandl | 2006-07-08 14:15:27 +0200 (Sat, 08 Jul 2006) | 3 lines Add an additional test for bug #1519018. ........ r50491 | tim.peters | 2006-07-08 21:55:05 +0200 (Sat, 08 Jul 2006) | 2 lines Whitespace normalization. ........ r50493 | neil.schemenauer | 2006-07-09 18:16:34 +0200 (Sun, 09 Jul 2006) | 2 lines Fix AST compiler bug #1501934: incorrect LOAD/STORE_GLOBAL generation. ........ r50495 | neil.schemenauer | 2006-07-09 23:19:29 +0200 (Sun, 09 Jul 2006) | 2 lines Fix SF bug 1441486: bad unary minus folding in compiler. ........ r50497 | neal.norwitz | 2006-07-10 00:14:42 +0200 (Mon, 10 Jul 2006) | 4 lines On 64 bit systems, int literals that use less than 64 bits are now ints rather than longs. This also fixes the test for eval(-sys.maxint - 1). ........ r50500 | neal.norwitz | 2006-07-10 02:04:44 +0200 (Mon, 10 Jul 2006) | 4 lines Bug #1512814, Fix incorrect lineno's when code at module scope started after line 256. ........ r50501 | neal.norwitz | 2006-07-10 02:05:34 +0200 (Mon, 10 Jul 2006) | 1 line Fix doco. Backport candidate. ........ r50503 | neal.norwitz | 2006-07-10 02:23:17 +0200 (Mon, 10 Jul 2006) | 5 lines Part of SF patch #1484695. This removes dead code. The chksum was already verified in .frombuf() on the lines above. If there was a problem an exception is raised, so there was no way this condition could have been true. ........ r50504 | neal.norwitz | 2006-07-10 03:18:57 +0200 (Mon, 10 Jul 2006) | 3 lines Patch #1516912: improve Modules support for OpenVMS. ........ r50506 | neal.norwitz | 2006-07-10 04:36:41 +0200 (Mon, 10 Jul 2006) | 7 lines Patch #1504046: Add documentation for xml.etree. /F wrote the text docs, Englebert Gruber massaged it to latex and I did some more massaging to try and improve the consistency and fix some name mismatches between the declaration and text. ........ r50509 | martin.v.loewis | 2006-07-10 09:23:48 +0200 (Mon, 10 Jul 2006) | 2 lines Introduce DISTUTILS_USE_SDK as a flag to determine whether the SDK environment should be used. Fixes #1508010. ........ r50510 | martin.v.loewis | 2006-07-10 09:26:41 +0200 (Mon, 10 Jul 2006) | 1 line Change error message to indicate that VS2003 is necessary to build extension modules, not the .NET SDK. ........ r50511 | martin.v.loewis | 2006-07-10 09:29:41 +0200 (Mon, 10 Jul 2006) | 1 line Add svn:ignore. ........ r50512 | anthony.baxter | 2006-07-10 09:41:04 +0200 (Mon, 10 Jul 2006) | 1 line preparing for 2.5b2 ........ r50513 | thomas.heller | 2006-07-10 11:10:28 +0200 (Mon, 10 Jul 2006) | 2 lines Fix bug #1518190: accept any integer or long value in the ctypes.c_void_p constructor. ........ r50514 | thomas.heller | 2006-07-10 11:31:06 +0200 (Mon, 10 Jul 2006) | 3 lines Fixed a segfault when ctypes.wintypes were imported on non-Windows machines. ........ r50516 | thomas.heller | 2006-07-10 13:11:10 +0200 (Mon, 10 Jul 2006) | 3 lines Assigning None to pointer type structure fields possible overwrote wrong fields. ........ r50517 | thomas.heller | 2006-07-10 13:17:37 +0200 (Mon, 10 Jul 2006) | 5 lines Moved the ctypes news entries from the 'Library' section into the 'Extension Modules' section where they belong, probably. This destroyes the original order of the news entries, don't know if that is important or not. ........ r50526 | phillip.eby | 2006-07-10 21:03:29 +0200 (Mon, 10 Jul 2006) | 2 lines Fix SF#1516184 and add a test to prevent regression. ........ r50528 | phillip.eby | 2006-07-10 21:18:35 +0200 (Mon, 10 Jul 2006) | 2 lines Fix SF#1457312: bad socket error handling in distutils "upload" command. ........ r50537 | peter.astrand | 2006-07-10 22:39:49 +0200 (Mon, 10 Jul 2006) | 1 line Make it possible to run test_subprocess.py with Python 2.2, which lacks test_support.reap_children(). ........ r50541 | tim.peters | 2006-07-10 23:08:24 +0200 (Mon, 10 Jul 2006) | 5 lines After approval from Anthony, merge the tim-current_frames branch into the trunk. This adds a new sys._current_frames() function, which returns a dict mapping thread id to topmost thread stack frame. ........ r50542 | tim.peters | 2006-07-10 23:11:49 +0200 (Mon, 10 Jul 2006) | 2 lines Whitespace normalization. ........ r50553 | martin.v.loewis | 2006-07-11 00:11:28 +0200 (Tue, 11 Jul 2006) | 4 lines Patch #1519566: Remove unused _tofill member. Make begin_fill idempotent. Update demo2 to demonstrate filling of concave shapes. ........ r50567 | anthony.baxter | 2006-07-11 04:04:09 +0200 (Tue, 11 Jul 2006) | 4 lines #1494314: Fix a regression with high-numbered sockets in 2.4.3. This means that select() on sockets > FD_SETSIZE (typically 1024) work again. The patch makes sockets use poll() internally where available. ........ r50568 | tim.peters | 2006-07-11 04:17:48 +0200 (Tue, 11 Jul 2006) | 2 lines Whitespace normalization. ........ r50575 | thomas.heller | 2006-07-11 18:42:05 +0200 (Tue, 11 Jul 2006) | 1 line Add missing Py_DECREF. ........ r50576 | thomas.heller | 2006-07-11 18:44:25 +0200 (Tue, 11 Jul 2006) | 1 line Add missing Py_DECREFs. ........ r50579 | andrew.kuchling | 2006-07-11 19:20:16 +0200 (Tue, 11 Jul 2006) | 1 line Bump version number; add sys._current_frames ........ r50582 | thomas.heller | 2006-07-11 20:28:35 +0200 (Tue, 11 Jul 2006) | 3 lines When a foreign function is retrived by calling __getitem__ on a ctypes library instance, do not set it as attribute. ........ r50583 | thomas.heller | 2006-07-11 20:40:50 +0200 (Tue, 11 Jul 2006) | 2 lines Change the ctypes version number to 1.0.0. ........ r50597 | neal.norwitz | 2006-07-12 07:26:17 +0200 (Wed, 12 Jul 2006) | 3 lines Bug #1520864: unpacking singleton tuples in for loop (for x, in) work again. ........ r50598 | neal.norwitz | 2006-07-12 07:26:35 +0200 (Wed, 12 Jul 2006) | 1 line Fix function name in error msg ........ r50599 | neal.norwitz | 2006-07-12 07:27:46 +0200 (Wed, 12 Jul 2006) | 4 lines Fix uninitialized memory read reported by Valgrind when running doctest. This could happen if size == 0. ........ r50600 | neal.norwitz | 2006-07-12 09:28:29 +0200 (Wed, 12 Jul 2006) | 1 line Actually change the MAGIC #. Create a new section for 2.5c1 and mention the impact of changing the MAGIC #. ........ r50601 | thomas.heller | 2006-07-12 10:43:47 +0200 (Wed, 12 Jul 2006) | 3 lines Fix #1467450: ctypes now uses RTLD_GLOBAL by default on OSX 10.3 to load shared libraries. ........ r50604 | thomas.heller | 2006-07-12 16:25:18 +0200 (Wed, 12 Jul 2006) | 3 lines Fix the wrong description of LibraryLoader.LoadLibrary, and document the DEFAULT_MODE constant. ........ r50607 | georg.brandl | 2006-07-12 17:31:17 +0200 (Wed, 12 Jul 2006) | 3 lines Accept long options "--help" and "--version". ........ r50617 | thomas.heller | 2006-07-13 11:53:47 +0200 (Thu, 13 Jul 2006) | 3 lines A misspelled preprocessor symbol caused ctypes to be always compiled without thread support. Replaced WITH_THREADS with WITH_THREAD. ........ r50619 | thomas.heller | 2006-07-13 19:01:14 +0200 (Thu, 13 Jul 2006) | 3 lines Fix #1521375. When running with root priviledges, 'gcc -o /dev/null' did overwrite /dev/null. Use a temporary file instead of /dev/null. ........ r50620 | thomas.heller | 2006-07-13 19:05:13 +0200 (Thu, 13 Jul 2006) | 2 lines Fix misleading words. ........ r50622 | andrew.kuchling | 2006-07-13 19:37:26 +0200 (Thu, 13 Jul 2006) | 1 line Typo fix ........ r50629 | georg.brandl | 2006-07-14 09:12:54 +0200 (Fri, 14 Jul 2006) | 3 lines Patch #1521874: grammar errors in doanddont.tex. ........ r50630 | neal.norwitz | 2006-07-14 09:20:04 +0200 (Fri, 14 Jul 2006) | 1 line Try to improve grammar further. ........ r50631 | martin.v.loewis | 2006-07-14 11:58:55 +0200 (Fri, 14 Jul 2006) | 1 line Extend build_ssl to Win64, using VSExtComp. ........ r50632 | martin.v.loewis | 2006-07-14 14:10:09 +0200 (Fri, 14 Jul 2006) | 1 line Add debug output to analyse buildbot failure. ........ r50633 | martin.v.loewis | 2006-07-14 14:31:05 +0200 (Fri, 14 Jul 2006) | 1 line Fix Debug build of _ssl. ........ r50636 | andrew.kuchling | 2006-07-14 15:32:38 +0200 (Fri, 14 Jul 2006) | 1 line Mention new options ........ r50638 | peter.astrand | 2006-07-14 16:04:45 +0200 (Fri, 14 Jul 2006) | 1 line Bug #1223937: CalledProcessError.errno -> CalledProcessError.returncode. ........ r50640 | thomas.heller | 2006-07-14 17:01:05 +0200 (Fri, 14 Jul 2006) | 4 lines Make the prototypes of our private PyUnicode_FromWideChar and PyUnicode_AsWideChar replacement functions compatible to the official functions by using Py_ssize_t instead of int. ........ r50643 | thomas.heller | 2006-07-14 19:51:14 +0200 (Fri, 14 Jul 2006) | 3 lines Patch #1521817: The index range checking on ctypes arrays containing exactly one element is enabled again. ........ r50647 | thomas.heller | 2006-07-14 20:22:50 +0200 (Fri, 14 Jul 2006) | 2 lines Updates for the ctypes documentation. ........ r50655 | fredrik.lundh | 2006-07-14 23:45:48 +0200 (Fri, 14 Jul 2006) | 3 lines typo ........ r50664 | george.yoshida | 2006-07-15 18:03:49 +0200 (Sat, 15 Jul 2006) | 2 lines Bug #15187702 : ext/win-cookbook.html has a broken link to distutils ........ r50667 | bob.ippolito | 2006-07-15 18:53:15 +0200 (Sat, 15 Jul 2006) | 1 line Patch #1220874: Update the binhex module for Mach-O. ........ r50671 | fred.drake | 2006-07-16 03:21:20 +0200 (Sun, 16 Jul 2006) | 1 line clean up some link markup ........ r50673 | neal.norwitz | 2006-07-16 03:50:38 +0200 (Sun, 16 Jul 2006) | 4 lines Bug #1512814, Fix incorrect lineno's when code within a function had more than 255 blank lines. Byte codes need to go first, line #s second. ........ r50674 | neal.norwitz | 2006-07-16 04:00:32 +0200 (Sun, 16 Jul 2006) | 5 lines a & b were dereffed above, so they are known to be valid pointers. z is known to be NULL, nothing to DECREF. Reported by Klockwork, #107. ........ r50675 | neal.norwitz | 2006-07-16 04:02:57 +0200 (Sun, 16 Jul 2006) | 5 lines self is dereffed (and passed as first arg), so it's known to be good. func is returned from PyArg_ParseTuple and also dereffed. Reported by Klocwork, #30 (self one at least). ........ r50676 | neal.norwitz | 2006-07-16 04:05:35 +0200 (Sun, 16 Jul 2006) | 4 lines proto was dereffed above and is known to be good. No need for X. Reported by Klocwork, #39. ........ r50677 | neal.norwitz | 2006-07-16 04:15:27 +0200 (Sun, 16 Jul 2006) | 5 lines Fix memory leaks in some conditions. Reported by Klocwork #152. ........ r50678 | neal.norwitz | 2006-07-16 04:17:36 +0200 (Sun, 16 Jul 2006) | 4 lines Fix memory leak under some conditions. Reported by Klocwork, #98. ........ r50679 | neal.norwitz | 2006-07-16 04:22:30 +0200 (Sun, 16 Jul 2006) | 8 lines Use sizeof(buffer) instead of duplicating the constants to ensure they won't be wrong. The real change is to pass (bufsz - 1) to PyOS_ascii_formatd and 1 to strncat. strncat copies n+1 bytes from src (not dest). Reported by Klocwork #58. ........ r50680 | neal.norwitz | 2006-07-16 04:32:03 +0200 (Sun, 16 Jul 2006) | 5 lines Handle a NULL name properly. Reported by Klocwork #67 ........ r50681 | neal.norwitz | 2006-07-16 04:35:47 +0200 (Sun, 16 Jul 2006) | 6 lines PyFunction_SetDefaults() is documented as taking None or a tuple. A NULL would crash the PyTuple_Check(). Now make NULL return a SystemError. Reported by Klocwork #73. ........ r50683 | neal.norwitz | 2006-07-17 02:55:45 +0200 (Mon, 17 Jul 2006) | 5 lines Stop INCREFing name, then checking if it's NULL. name (f_name) should never be NULL so assert it. Fix one place where we could have passed NULL. Reported by Klocwork #66. ........ r50684 | neal.norwitz | 2006-07-17 02:57:15 +0200 (Mon, 17 Jul 2006) | 5 lines otherset is known to be non-NULL based on checks before and DECREF after. DECREF otherset rather than XDECREF in error conditions too. Reported by Klockwork #154. ........ r50685 | neal.norwitz | 2006-07-17 02:59:04 +0200 (Mon, 17 Jul 2006) | 7 lines Reported by Klocwork #151. v2 can be NULL if exception2 is NULL. I don't think that condition can happen, but I'm not sure it can't either. Now the code will protect against either being NULL. ........ r50686 | neal.norwitz | 2006-07-17 03:00:16 +0200 (Mon, 17 Jul 2006) | 1 line Add NEWS entry for a bunch of fixes due to warnings produced by Klocworks static analysis tool. ........ r50687 | fred.drake | 2006-07-17 07:47:52 +0200 (Mon, 17 Jul 2006) | 3 lines document xmlcore (still minimal; needs mention in each of the xml.* modules) SF bug #1504456 (partial) ........ r50688 | georg.brandl | 2006-07-17 15:23:46 +0200 (Mon, 17 Jul 2006) | 3 lines Remove usage of sets module (patch #1500609). ........ r50689 | georg.brandl | 2006-07-17 15:26:33 +0200 (Mon, 17 Jul 2006) | 3 lines Add missing NEWS item (#1522771) ........ r50690 | andrew.kuchling | 2006-07-17 18:47:54 +0200 (Mon, 17 Jul 2006) | 1 line Attribute more features ........ r50692 | kurt.kaiser | 2006-07-17 23:59:27 +0200 (Mon, 17 Jul 2006) | 8 lines Patch 1479219 - Tal Einat 1. 'as' highlighted as builtin in comment string on import line 2. Comments such as "#False identity" which start with a keyword immediately after the '#' character aren't colored as comments. 3. u or U beginning unicode string not correctly highlighted Closes bug 1325071 ........ r50693 | barry.warsaw | 2006-07-18 01:07:51 +0200 (Tue, 18 Jul 2006) | 16 lines decode_rfc2231(): Be more robust against buggy RFC 2231 encodings. Specifically, instead of raising a ValueError when there is a single tick in the parameter, simply return that the entire string unquoted, with None for both the charset and the language. Also, if there are more than 2 ticks in the parameter, interpret the first three parts as the standard RFC 2231 parts, then the rest of the parts as the encoded string. Test cases added. Original fewer-than-3-parts fix by Tokio Kikuchi. Resolves SF bug # 1218081. I will back port the fix and tests to Python 2.4 (email 3.0) and Python 2.3 (email 2.5). Also, bump the version number to email 4.0.1, removing the 'alpha' moniker. ........ r50695 | kurt.kaiser | 2006-07-18 06:03:16 +0200 (Tue, 18 Jul 2006) | 2 lines Rebinding Tab key was inserting 'tab' instead of 'Tab'. Bug 1179168. ........ r50696 | brett.cannon | 2006-07-18 06:41:36 +0200 (Tue, 18 Jul 2006) | 6 lines Fix bug #1520914. Starting in 2.4, time.strftime() began to check the bounds of values in the time tuple passed in. Unfortunately people came to rely on undocumented behaviour of setting unneeded values to 0, regardless of if it was within the valid range. Now those values force the value internally to the minimum value when 0 is passed in. ........ r50697 | facundo.batista | 2006-07-18 14:16:13 +0200 (Tue, 18 Jul 2006) | 1 line Comments and docs cleanups, and some little fixes, provided by Santiágo Peresón ........ r50704 | martin.v.loewis | 2006-07-18 19:46:31 +0200 (Tue, 18 Jul 2006) | 2 lines Patch #1524429: Use repr instead of backticks again. ........ r50706 | tim.peters | 2006-07-18 23:55:15 +0200 (Tue, 18 Jul 2006) | 2 lines Whitespace normalization. ........ r50708 | tim.peters | 2006-07-19 02:03:19 +0200 (Wed, 19 Jul 2006) | 18 lines SF bug 1524317: configure --without-threads fails to build Moved the code for _PyThread_CurrentFrames() up, so it's no longer in a huge "#ifdef WITH_THREAD" block (I didn't realize it /was/ in one). Changed test_sys's test_current_frames() so it passes with or without thread supported compiled in. Note that test_sys fails when Python is compiled without threads, but for an unrelated reason (the old test_exit() fails with an indirect ImportError on the `thread` module). There are also other unrelated compilation failures without threads, in extension modules (like ctypes); at least the core compiles again. Do we really support --without-threads? If so, there are several problems remaining. ........ r50713 | thomas.heller | 2006-07-19 11:09:32 +0200 (Wed, 19 Jul 2006) | 4 lines Make sure the _ctypes extension can be compiled when WITH_THREAD is not defined on Windows, even if that configuration is probably not supported at all. ........ r50715 | martin.v.loewis | 2006-07-19 19:18:32 +0200 (Wed, 19 Jul 2006) | 4 lines Revert r50706 (Whitespace normalization) and r50697: Comments and docs cleanups, and some little fixes per recommendation from Raymond Hettinger. ........ r50719 | phillip.eby | 2006-07-20 17:54:16 +0200 (Thu, 20 Jul 2006) | 4 lines Fix SF#1516184 (again) and add a test to prevent regression. (There was a problem with empty filenames still causing recursion) ........ r50720 | georg.brandl | 2006-07-20 18:28:39 +0200 (Thu, 20 Jul 2006) | 3 lines Guard for _active being None in __del__ method. ........ r50721 | vinay.sajip | 2006-07-20 18:28:39 +0200 (Thu, 20 Jul 2006) | 1 line Updated documentation for TimedRotatingFileHandler relating to how rollover files are named. The previous documentation was wrongly the same as for RotatingFileHandler. ........ r50731 | fred.drake | 2006-07-20 22:11:57 +0200 (Thu, 20 Jul 2006) | 1 line markup fix ........ r50739 | kurt.kaiser | 2006-07-21 00:22:52 +0200 (Fri, 21 Jul 2006) | 7 lines Avoid occasional failure to detect closing paren properly. Patch 1407280 Tal Einat M ParenMatch.py M NEWS.txt M CREDITS.txt ........ r50740 | vinay.sajip | 2006-07-21 01:20:12 +0200 (Fri, 21 Jul 2006) | 1 line Addressed SF#1524081 by using a dictionary to map level names to syslog priority names, rather than a string.lower(). ........ r50741 | neal.norwitz | 2006-07-21 07:29:58 +0200 (Fri, 21 Jul 2006) | 1 line Add some asserts that we got good params passed ........ r50742 | neal.norwitz | 2006-07-21 07:31:02 +0200 (Fri, 21 Jul 2006) | 5 lines Move the initialization of some pointers earlier. The problem is that if we call Py_DECREF(frame) like we do if allocating locals fails, frame_dealloc() will try to use these bogus values and crash. ........ r50743 | neal.norwitz | 2006-07-21 07:32:28 +0200 (Fri, 21 Jul 2006) | 4 lines Handle allocation failures gracefully. Found with failmalloc. Many (all?) of these could be backported. ........ r50745 | neal.norwitz | 2006-07-21 09:59:02 +0200 (Fri, 21 Jul 2006) | 1 line Speel initialise write. Tanks Anthony. ........ r50746 | neal.norwitz | 2006-07-21 09:59:47 +0200 (Fri, 21 Jul 2006) | 2 lines Handle more memory allocation failures without crashing. ........ r50754 | barry.warsaw | 2006-07-21 16:51:07 +0200 (Fri, 21 Jul 2006) | 23 lines More RFC 2231 improvements for the email 4.0 package. As Mark Sapiro rightly points out there are really two types of continued headers defined in this RFC (i.e. "encoded" parameters with the form "name*0*=" and unencoded parameters with the form "name*0="), but we were were handling them both the same way and that isn't correct. This patch should be much more RFC compliant in that only encoded params are %-decoded and the charset/language information is only extract if there are any encoded params in the segments. If there are no encoded params then the RFC says that there will be no charset/language parts. Note however that this will change the return value for Message.get_param() in some cases. For example, whereas before if you had all unencoded param continuations you would have still gotten a 3-tuple back from this method (with charset and language == None), you will now get just a string. I don't believe this is a backward incompatible change though because the documentation for this method already indicates that either return value is possible and that you must do an isinstance(val, tuple) check to discriminate between the two. (Yeah that API kind of sucks but we can't change /that/ without breaking code.) Test cases, some documentation updates, and a NEWS item accompany this patch. ........ r50759 | georg.brandl | 2006-07-21 19:36:31 +0200 (Fri, 21 Jul 2006) | 3 lines Fix check for empty list (vs. None). ........ r50771 | brett.cannon | 2006-07-22 00:44:07 +0200 (Sat, 22 Jul 2006) | 2 lines Remove an XXX marker in a comment. ........ r50773 | neal.norwitz | 2006-07-22 18:20:49 +0200 (Sat, 22 Jul 2006) | 1 line Fix more memory allocation issues found with failmalloc. ........ r50774 | neal.norwitz | 2006-07-22 19:00:57 +0200 (Sat, 22 Jul 2006) | 1 line Don't fail if the directory already exists ........ r50775 | greg.ward | 2006-07-23 04:25:53 +0200 (Sun, 23 Jul 2006) | 6 lines Be a lot smarter about whether this test passes: instead of assuming that a 2.93 sec audio file will always take 3.1 sec (as it did on the hardware I had when I first wrote the test), expect that it will take 2.93 sec +/- 10%, and only fail if it's outside of that range. Compute the expected ........ r50776 | kurt.kaiser | 2006-07-23 06:19:49 +0200 (Sun, 23 Jul 2006) | 2 lines Tooltips failed on new-syle class __init__ args. Bug 1027566 Loren Guthrie ........ r50777 | neal.norwitz | 2006-07-23 09:50:36 +0200 (Sun, 23 Jul 2006) | 1 line Handle more mem alloc issues found with failmalloc ........ r50778 | neal.norwitz | 2006-07-23 09:51:58 +0200 (Sun, 23 Jul 2006) | 5 lines If the for loop isn't entered, entryblock will be NULL. If passed to stackdepth_walk it will be dereffed. Not sure if I found with failmalloc or Klockwork #55. ........ r50779 | neal.norwitz | 2006-07-23 09:53:14 +0200 (Sun, 23 Jul 2006) | 4 lines Move the initialization of size_a down below the check for a being NULL. Reported by Klocwork #106 ........ r50780 | neal.norwitz | 2006-07-23 09:55:55 +0200 (Sun, 23 Jul 2006) | 9 lines Check the allocation of b_objects and return if there was a failure. Also fix a few memory leaks in other failure scenarios. It seems that if b_objects == Py_None, we will have an extra ref to b_objects. Add XXX comment so hopefully someone documents why the else isn't necessary or adds it in. Reported by Klocwork #20 ........ r50781 | neal.norwitz | 2006-07-23 09:57:11 +0200 (Sun, 23 Jul 2006) | 2 lines Fix memory leaks spotted by Klocwork #37. ........ r50782 | neal.norwitz | 2006-07-23 09:59:00 +0200 (Sun, 23 Jul 2006) | 5 lines nextlink can be NULL if teedataobject_new fails, so use XINCREF. Ensure that dataobj is never NULL. Reported by Klocwork #102 ........ r50783 | neal.norwitz | 2006-07-23 10:01:43 +0200 (Sun, 23 Jul 2006) | 8 lines Ensure we don't write beyond errText. I think I got this right, but it definitely could use some review to ensure I'm not off by one and there's no possible overflow/wrap-around of bytes_left. Reported by Klocwork #1. Fix a problem if there is a failure allocating self->db. Found with failmalloc. ........ r50784 | ronald.oussoren | 2006-07-23 11:41:09 +0200 (Sun, 23 Jul 2006) | 3 lines Without this patch CMD-W won't close EditorWindows on MacOS X. This solves part of bug #1517990. ........ r50785 | ronald.oussoren | 2006-07-23 11:46:11 +0200 (Sun, 23 Jul 2006) | 5 lines Fix for bug #1517996: Class and Path browsers show Tk menu This patch replaces the menubar that is used by AquaTk for windows without a menubar of their own by one that is more appropriate for IDLE. ........ r50786 | andrew.macintyre | 2006-07-23 14:57:02 +0200 (Sun, 23 Jul 2006) | 2 lines Build updates for OS/2 EMX port ........ r50787 | andrew.macintyre | 2006-07-23 15:00:04 +0200 (Sun, 23 Jul 2006) | 3 lines bugfix: PyThread_start_new_thread() returns the thread ID, not a flag; will backport. ........ r50789 | andrew.macintyre | 2006-07-23 15:04:00 +0200 (Sun, 23 Jul 2006) | 2 lines Get mailbox module working on OS/2 EMX port. ........ r50791 | greg.ward | 2006-07-23 18:05:51 +0200 (Sun, 23 Jul 2006) | 1 line Resync optparse with Optik 1.5.3: minor tweaks for/to tests. ........ r50794 | martin.v.loewis | 2006-07-24 07:05:22 +0200 (Mon, 24 Jul 2006) | 2 lines Update list of unsupported systems. Fixes #1510853. ........ r50795 | martin.v.loewis | 2006-07-24 12:26:33 +0200 (Mon, 24 Jul 2006) | 1 line Patch #1448199: Release GIL around ConnectRegistry. ........ r50796 | martin.v.loewis | 2006-07-24 13:54:53 +0200 (Mon, 24 Jul 2006) | 3 lines Patch #1232023: Don't include empty path component from registry, so that the current directory does not get added to sys.path. Also fixes #1526785. ........ r50797 | martin.v.loewis | 2006-07-24 14:54:17 +0200 (Mon, 24 Jul 2006) | 3 lines Bug #1524310: Properly report errors from FindNextFile in os.listdir. Will backport to 2.4. ........ r50800 | georg.brandl | 2006-07-24 15:28:57 +0200 (Mon, 24 Jul 2006) | 7 lines Patch #1523356: fix determining include dirs in python-config. Also don't install "python-config" when doing altinstall, but always install "python-config2.x" and make a link to it like with the main executable. ........ r50802 | georg.brandl | 2006-07-24 15:46:47 +0200 (Mon, 24 Jul 2006) | 3 lines Patch #1527744: right order of includes in order to have HAVE_CONIO_H defined properly. ........ r50803 | georg.brandl | 2006-07-24 16:09:56 +0200 (Mon, 24 Jul 2006) | 3 lines Patch #1515343: Fix printing of deprecated string exceptions with a value in the traceback module. ........ r50804 | kurt.kaiser | 2006-07-24 19:13:23 +0200 (Mon, 24 Jul 2006) | 7 lines EditorWindow failed when used stand-alone if sys.ps1 not set. Bug 1010370 Dave Florek M EditorWindow.py M PyShell.py M NEWS.txt ........ r50805 | kurt.kaiser | 2006-07-24 20:05:51 +0200 (Mon, 24 Jul 2006) | 6 lines - EditorWindow.test() was failing. Bug 1417598 M EditorWindow.py M ScriptBinding.py M NEWS.txt ........ r50808 | georg.brandl | 2006-07-24 22:11:35 +0200 (Mon, 24 Jul 2006) | 3 lines Repair accidental NameError. ........ r50809 | tim.peters | 2006-07-24 23:02:15 +0200 (Mon, 24 Jul 2006) | 2 lines Whitespace normalization. ........ r50810 | greg.ward | 2006-07-25 04:11:12 +0200 (Tue, 25 Jul 2006) | 3 lines Don't use standard assert: want tests to fail even when run with -O. Delete cruft. ........ r50811 | tim.peters | 2006-07-25 06:07:22 +0200 (Tue, 25 Jul 2006) | 10 lines current_frames_with_threads(): There's actually no way to guess /which/ line the spawned thread is in at the time sys._current_frames() is called: we know it finished enter_g.set(), but can't know whether the instruction counter has advanced to the following leave_g.wait(). The latter is overwhelming most likely, but not guaranteed, and I see that the "x86 Ubuntu dapper (icc) trunk" buildbot found it on the other line once. Changed the test so it passes in either case. ........ r50815 | martin.v.loewis | 2006-07-25 11:53:12 +0200 (Tue, 25 Jul 2006) | 2 lines Bug #1525817: Don't truncate short lines in IDLE's tool tips. ........ r50816 | martin.v.loewis | 2006-07-25 12:05:47 +0200 (Tue, 25 Jul 2006) | 3 lines Bug #978833: Really close underlying socket in _socketobject.close. Will backport to 2.4. ........ r50817 | martin.v.loewis | 2006-07-25 12:11:14 +0200 (Tue, 25 Jul 2006) | 1 line Revert incomplete checkin. ........ r50819 | georg.brandl | 2006-07-25 12:22:34 +0200 (Tue, 25 Jul 2006) | 4 lines Patch #1525766: correctly pass onerror arg to recursive calls of pkg.walk_packages. Also improve the docstrings. ........ r50825 | brett.cannon | 2006-07-25 19:32:20 +0200 (Tue, 25 Jul 2006) | 2 lines Add comment for changes to test_ossaudiodev. ........ r50826 | brett.cannon | 2006-07-25 19:34:36 +0200 (Tue, 25 Jul 2006) | 3 lines Fix a bug in the messages for an assert failure where not enough arguments to a string were being converted in the format. ........ r50828 | armin.rigo | 2006-07-25 20:09:57 +0200 (Tue, 25 Jul 2006) | 2 lines Document why is and is not a good way to fix the gc_inspection crasher. ........ r50829 | armin.rigo | 2006-07-25 20:11:07 +0200 (Tue, 25 Jul 2006) | 5 lines Added another crasher, which hit me today (I was not intentionally writing such code, of course, but it took some gdb time to figure out what my bug was). ........ r50830 | armin.rigo | 2006-07-25 20:38:39 +0200 (Tue, 25 Jul 2006) | 3 lines Document the crashers that will not go away soon as "won't fix", and explain why. ........ r50831 | ronald.oussoren | 2006-07-25 21:13:35 +0200 (Tue, 25 Jul 2006) | 3 lines Install the compatibility symlink to libpython.a on OSX using 'ln -sf' instead of 'ln -s', this avoid problems when reinstalling python. ........ r50832 | ronald.oussoren | 2006-07-25 21:20:54 +0200 (Tue, 25 Jul 2006) | 7 lines Fix for bug #1525447 (renaming to MacOSmodule.c would also work, but not without causing problems for anyone that is on a case-insensitive filesystem). Setup.py tries to compile the MacOS extension from MacOSmodule.c, while the actual file is named macosmodule.c. This is no problem on the (default) case-insensitive filesystem, but doesn't work on case-sensitive filesystems. ........ r50833 | ronald.oussoren | 2006-07-25 22:28:55 +0200 (Tue, 25 Jul 2006) | 7 lines Fix bug #1517990: IDLE keybindings on OSX This adds a new key definition for OSX, which is slightly different from the classic mac definition. Also add NEWS item for a couple of bugfixes I added recently. ........ r50834 | tim.peters | 2006-07-26 00:30:24 +0200 (Wed, 26 Jul 2006) | 2 lines Whitespace normalization. ........ r50839 | neal.norwitz | 2006-07-26 06:00:18 +0200 (Wed, 26 Jul 2006) | 1 line Hmm, only python2.x is installed, not plain python. Did that change recently? ........ r50840 | barry.warsaw | 2006-07-26 07:54:46 +0200 (Wed, 26 Jul 2006) | 6 lines Forward port some fixes that were in email 2.5 but for some reason didn't make it into email 4.0. Specifically, in Message.get_content_charset(), handle RFC 2231 headers that contain an encoding not known to Python, or a character in the data that isn't in the charset encoding. Also forward port the appropriate unit tests. ........ r50841 | georg.brandl | 2006-07-26 09:23:32 +0200 (Wed, 26 Jul 2006) | 3 lines NEWS entry for #1525766. ........ r50842 | georg.brandl | 2006-07-26 09:40:17 +0200 (Wed, 26 Jul 2006) | 3 lines Bug #1459963: properly capitalize HTTP header names. ........ r50843 | georg.brandl | 2006-07-26 10:03:10 +0200 (Wed, 26 Jul 2006) | 6 lines Part of bug #1523610: fix miscalculation of buffer length. Also add a guard against NULL in converttuple and add a test case (that previously would have crashed). ........ r50844 | martin.v.loewis | 2006-07-26 14:12:56 +0200 (Wed, 26 Jul 2006) | 3 lines Bug #978833: Really close underlying socket in _socketobject.close. Fix httplib.HTTPConnection.getresponse to not close the socket if it is still needed for the response. ........ r50845 | andrew.kuchling | 2006-07-26 19:16:52 +0200 (Wed, 26 Jul 2006) | 1 line [Bug #1471938] Fix build problem on Solaris 8 by conditionalizing the use of mvwgetnstr(); it was conditionalized a few lines below. Fix from Paul Eggert. I also tried out the STRICT_SYSV_CURSES case and am therefore removing the 'untested' comment. ........ r50846 | andrew.kuchling | 2006-07-26 19:18:01 +0200 (Wed, 26 Jul 2006) | 1 line Correct error message ........ r50847 | andrew.kuchling | 2006-07-26 19:19:39 +0200 (Wed, 26 Jul 2006) | 1 line Minor grammar fix ........ r50848 | andrew.kuchling | 2006-07-26 19:22:21 +0200 (Wed, 26 Jul 2006) | 1 line Put news item in right section ........ r50850 | andrew.kuchling | 2006-07-26 20:03:12 +0200 (Wed, 26 Jul 2006) | 1 line Use sys.exc_info() ........ r50851 | andrew.kuchling | 2006-07-26 20:15:45 +0200 (Wed, 26 Jul 2006) | 1 line Use sys.exc_info() ........ r50852 | phillip.eby | 2006-07-26 21:48:27 +0200 (Wed, 26 Jul 2006) | 4 lines Allow the 'onerror' argument to walk_packages() to catch any Exception, not just ImportError. This allows documentation tools to better skip unimportable packages. ........ r50854 | tim.peters | 2006-07-27 01:23:15 +0200 (Thu, 27 Jul 2006) | 2 lines Whitespace normalization. ........ r50855 | tim.peters | 2006-07-27 03:14:53 +0200 (Thu, 27 Jul 2006) | 21 lines Bug #1521947: possible bug in mystrtol.c with recent gcc. In general, C doesn't define anything about what happens when an operation on a signed integral type overflows, and PyOS_strtol() did several formally undefined things of that nature on signed longs. Some version of gcc apparently tries to exploit that now, and PyOS_strtol() could fail to detect overflow then. Tried to repair all that, although it seems at least as likely to me that we'll get screwed by bad platform definitions for LONG_MIN and/or LONG_MAX now. For that reason, I don't recommend backporting this. Note that I have no box on which this makes a lick of difference -- can't really test it, except to note that it didn't break anything on my boxes. Silent change: PyOS_strtol() used to return the hard-coded 0x7fffffff in case of overflow. Now it returns LONG_MAX. They're the same only on 32-bit boxes (although C doesn't guarantee that either ...). ........ r50856 | neal.norwitz | 2006-07-27 05:51:58 +0200 (Thu, 27 Jul 2006) | 6 lines Don't kill a normal instance of python running on windows when checking to kill a cygwin instance. build\\python.exe was matching a normal windows instance. Prefix that with a \\ to ensure build is a directory and not PCbuild. As discussed on python-dev. ........ r50857 | neal.norwitz | 2006-07-27 05:55:39 +0200 (Thu, 27 Jul 2006) | 5 lines Closure can't be NULL at this point since we know it's a tuple. Reported by Klocwork # 74. ........ r50858 | neal.norwitz | 2006-07-27 06:04:50 +0200 (Thu, 27 Jul 2006) | 1 line No functional change. Add comment and assert to describe why there cannot be overflow which was reported by Klocwork. Discussed on python-dev ........ r50859 | martin.v.loewis | 2006-07-27 08:38:16 +0200 (Thu, 27 Jul 2006) | 3 lines Bump distutils version to 2.5, as several new features have been introduced since 2.4. ........ r50860 | andrew.kuchling | 2006-07-27 14:18:20 +0200 (Thu, 27 Jul 2006) | 1 line Reformat docstring; fix typo ........ r50861 | georg.brandl | 2006-07-27 17:05:36 +0200 (Thu, 27 Jul 2006) | 6 lines Add test_main() methods. These three tests were never run by regrtest.py. We really need a simpler testing framework. ........ r50862 | tim.peters | 2006-07-27 17:09:20 +0200 (Thu, 27 Jul 2006) | 2 lines News for patch #1529686. ........ r50863 | tim.peters | 2006-07-27 17:11:00 +0200 (Thu, 27 Jul 2006) | 2 lines Whitespace normalization. ........ r50864 | georg.brandl | 2006-07-27 17:38:33 +0200 (Thu, 27 Jul 2006) | 3 lines Amend news entry. ........ r50865 | georg.brandl | 2006-07-27 18:08:15 +0200 (Thu, 27 Jul 2006) | 3 lines Make uuid test suite pass on this box by requesting output with LC_ALL=C. ........ r50866 | andrew.kuchling | 2006-07-27 20:37:33 +0200 (Thu, 27 Jul 2006) | 1 line Add example ........ r50867 | thomas.heller | 2006-07-27 20:39:55 +0200 (Thu, 27 Jul 2006) | 9 lines Remove code that is no longer used (ctypes.com). Fix the DllGetClassObject and DllCanUnloadNow so that they forward the call to the comtypes.server.inprocserver module. The latter was never documented, never used by published code, and didn't work anyway, so I think it does not deserve a NEWS entry (but I might be wrong). ........ r50868 | andrew.kuchling | 2006-07-27 20:41:21 +0200 (Thu, 27 Jul 2006) | 1 line Typo fix ('publically' is rare, poss. non-standard) ........ r50869 | andrew.kuchling | 2006-07-27 20:42:41 +0200 (Thu, 27 Jul 2006) | 1 line Add missing word ........ r50870 | andrew.kuchling | 2006-07-27 20:44:10 +0200 (Thu, 27 Jul 2006) | 1 line Repair typos ........ r50872 | andrew.kuchling | 2006-07-27 20:53:33 +0200 (Thu, 27 Jul 2006) | 1 line Update URL; add example ........ r50873 | andrew.kuchling | 2006-07-27 21:07:29 +0200 (Thu, 27 Jul 2006) | 1 line Add punctuation mark; add some examples ........ r50874 | andrew.kuchling | 2006-07-27 21:11:07 +0200 (Thu, 27 Jul 2006) | 1 line Mention base64 module; rewrite last sentence to be more positive ........ r50875 | andrew.kuchling | 2006-07-27 21:12:49 +0200 (Thu, 27 Jul 2006) | 1 line If binhex is higher-level than binascii, it should come first in the chapter ........ r50876 | tim.peters | 2006-07-27 22:47:24 +0200 (Thu, 27 Jul 2006) | 28 lines check_node(): stop spraying mystery output to stderr. When a node number disagrees, keep track of all sources & the node numbers they reported, and stick all that in the error message. Changed all callers to supply a non-empty "source" argument; made the "source" argument non-optional. On my box, test_uuid still fails, but with the less confusing output: AssertionError: different sources disagree on node: from source 'getnode1', node was 00038a000015 from source 'getnode2', node was 00038a000015 from source 'ipconfig', node was 001111b2b7bf Only the last one appears to be correct; e.g., C:\Code\python\PCbuild>getmac Physical Address Transport Name =================== ========================================================== 00-11-11-B2-B7-BF \Device\Tcpip_{190FB163-5AFD-4483-86A1-2FE16AC61FF1} 62-A1-AC-6C-FD-BE \Device\Tcpip_{8F77DF5A-EA3D-4F1D-975E-D472CEE6438A} E2-1F-01-C6-5D-88 \Device\Tcpip_{CD18F76B-2EF3-409F-9B8A-6481EE70A1E4} I can't find anything on my box with MAC 00-03-8a-00-00-15, and am not clear on where that comes from. ........ r50878 | andrew.kuchling | 2006-07-28 00:40:05 +0200 (Fri, 28 Jul 2006) | 1 line Reword paragraph ........ r50879 | andrew.kuchling | 2006-07-28 00:49:38 +0200 (Fri, 28 Jul 2006) | 1 line Add example ........ r50880 | andrew.kuchling | 2006-07-28 00:49:54 +0200 (Fri, 28 Jul 2006) | 1 line Add example ........ r50881 | barry.warsaw | 2006-07-28 01:43:15 +0200 (Fri, 28 Jul 2006) | 27 lines Patch #1520294: Support for getset and member descriptors in types.py, inspect.py, and pydoc.py. Specifically, this allows for querying the type of an object against these built-in C types and more importantly, for getting their docstrings printed in the interactive interpreter's help() function. This patch includes a new built-in module called _types which provides definitions of getset and member descriptors for use by the types.py module. These types are exposed as types.GetSetDescriptorType and types.MemberDescriptorType. Query functions are provided as inspect.isgetsetdescriptor() and inspect.ismemberdescriptor(). The implementations of these are robust enough to work with Python implementations other than CPython, which may not have these fundamental types. The patch also includes documentation and test suite updates. I commit these changes now under these guiding principles: 1. Silence is assent. The release manager has not said "no", and of the few people that cared enough to respond to the thread, the worst vote was "0". 2. It's easier to ask for forgiveness than permission. 3. It's so dang easy to revert stuff in svn, that you could view this as a forcing function. :) Windows build patches will follow. ........ r50882 | tim.peters | 2006-07-28 01:44:37 +0200 (Fri, 28 Jul 2006) | 4 lines Bug #1529297: The rewrite of doctest for Python 2.4 unintentionally lost that tests are sorted by name before being run. ``DocTestFinder`` has been changed to sort the list of tests it returns. ........ r50883 | tim.peters | 2006-07-28 01:45:48 +0200 (Fri, 28 Jul 2006) | 2 lines Whitespace normalization. ........ r50884 | tim.peters | 2006-07-28 01:46:36 +0200 (Fri, 28 Jul 2006) | 2 lines Add missing svn:eol-style property to text files. ........ r50885 | barry.warsaw | 2006-07-28 01:50:40 +0200 (Fri, 28 Jul 2006) | 4 lines Enable the building of the _types module on Windows. Note that this has only been tested for VS 2003 since that's all I have. ........ r50887 | tim.peters | 2006-07-28 02:23:15 +0200 (Fri, 28 Jul 2006) | 7 lines defdict_reduce(): Plug leaks. We didn't notice these before because test_defaultdict didn't actually do anything before Georg fixed that earlier today. Neal's next refleak run then showed test_defaultdict leaking 9 references on each run. That's repaired by this checkin. ........ r50888 | tim.peters | 2006-07-28 02:30:00 +0200 (Fri, 28 Jul 2006) | 2 lines News about the repaired memory leak in defaultdict. ........ r50889 | gregory.p.smith | 2006-07-28 03:35:25 +0200 (Fri, 28 Jul 2006) | 7 lines - pybsddb Bug #1527939: bsddb module DBEnv dbremove and dbrename methods now allow their database parameter to be None as the sleepycat API allows. Also adds an appropriate test case for DBEnv.dbrename and dbremove. ........ r50895 | neal.norwitz | 2006-07-28 06:22:34 +0200 (Fri, 28 Jul 2006) | 1 line Ensure the actual number matches the expected count ........ r50896 | tim.peters | 2006-07-28 06:51:59 +0200 (Fri, 28 Jul 2006) | 6 lines Live with that "the hardware address" is an ill-defined concept, and that different ways of trying to find "the hardware address" may return different results. Certainly true on both of my Windows boxes, and in different ways (see whining on python-dev). ........ r50897 | neal.norwitz | 2006-07-28 09:21:27 +0200 (Fri, 28 Jul 2006) | 3 lines Try to find the MAC addr on various flavours of Unix. This seems hopeless. The reduces the test_uuid failures, but there's still another method failing. ........ r50898 | martin.v.loewis | 2006-07-28 09:45:49 +0200 (Fri, 28 Jul 2006) | 2 lines Add UUID for upcoming 2.5b3. ........ r50899 | matt.fleming | 2006-07-28 13:27:27 +0200 (Fri, 28 Jul 2006) | 3 lines Allow socketmodule to compile on NetBSD -current, whose bluetooth API differs from both Linux and FreeBSD. Accepted by Neal Norwitz. ........ r50900 | andrew.kuchling | 2006-07-28 14:07:12 +0200 (Fri, 28 Jul 2006) | 1 line [Patch #1529811] Correction to description of r|* mode ........ r50901 | andrew.kuchling | 2006-07-28 14:18:22 +0200 (Fri, 28 Jul 2006) | 1 line Typo fix ........ r50902 | andrew.kuchling | 2006-07-28 14:32:43 +0200 (Fri, 28 Jul 2006) | 1 line Add example ........ r50903 | andrew.kuchling | 2006-07-28 14:33:19 +0200 (Fri, 28 Jul 2006) | 1 line Add example ........ r50904 | andrew.kuchling | 2006-07-28 14:45:55 +0200 (Fri, 28 Jul 2006) | 1 line Don't overwrite built-in name; add some blank lines for readability ........ r50905 | andrew.kuchling | 2006-07-28 14:48:07 +0200 (Fri, 28 Jul 2006) | 1 line Add example. Should I propagate this example to all the other DBM-ish modules, too? ........ r50912 | georg.brandl | 2006-07-28 20:31:39 +0200 (Fri, 28 Jul 2006) | 3 lines Patch #1529686: also run test_email_codecs with regrtest.py. ........ r50913 | georg.brandl | 2006-07-28 20:36:01 +0200 (Fri, 28 Jul 2006) | 3 lines Fix spelling. ........ r50915 | thomas.heller | 2006-07-28 21:42:40 +0200 (Fri, 28 Jul 2006) | 3 lines Remove a useless XXX comment. Cosmetic changes to the code so that the #ifdef _UNICODE block doesn't mess emacs code formatting. ........ r50916 | phillip.eby | 2006-07-28 23:12:07 +0200 (Fri, 28 Jul 2006) | 5 lines Bug #1529871: The speed enhancement patch #921466 broke Python's compliance with PEP 302. This was fixed by adding an ``imp.NullImporter`` type that is used in ``sys.path_importer_cache`` to cache non-directory paths and avoid excessive filesystem operations during imports. ........ r50917 | phillip.eby | 2006-07-28 23:31:54 +0200 (Fri, 28 Jul 2006) | 2 lines Fix svn merge spew. ........ r50918 | thomas.heller | 2006-07-28 23:43:20 +0200 (Fri, 28 Jul 2006) | 4 lines Patch #1529514: More openbsd platforms for ctypes. Regenerated Modules/_ctypes/libffi/configure with autoconf 2.59. Approved by Neal. ........ r50922 | georg.brandl | 2006-07-29 10:51:21 +0200 (Sat, 29 Jul 2006) | 2 lines Bug #835255: The "closure" argument to new.function() is now documented. ........ r50924 | georg.brandl | 2006-07-29 11:33:26 +0200 (Sat, 29 Jul 2006) | 3 lines Bug #1441397: The compiler module now recognizes module and function docstrings correctly as it did in Python 2.4. ........ r50925 | georg.brandl | 2006-07-29 12:25:46 +0200 (Sat, 29 Jul 2006) | 4 lines Revert rev 42617, it was introduced to work around bug #1441397. test_compiler now passes again. ........ r50926 | fred.drake | 2006-07-29 15:22:49 +0200 (Sat, 29 Jul 2006) | 1 line update target version number ........ r50927 | andrew.kuchling | 2006-07-29 15:56:48 +0200 (Sat, 29 Jul 2006) | 1 line Add example ........ r50928 | andrew.kuchling | 2006-07-29 16:04:47 +0200 (Sat, 29 Jul 2006) | 1 line Update URL ........ r50930 | andrew.kuchling | 2006-07-29 16:08:15 +0200 (Sat, 29 Jul 2006) | 1 line Reword paragraph to match the order of the subsequent sections ........ r50931 | andrew.kuchling | 2006-07-29 16:21:15 +0200 (Sat, 29 Jul 2006) | 1 line [Bug #1529157] Mention raw_input() and input(); while I'm at it, reword the description a bit ........ r50932 | andrew.kuchling | 2006-07-29 16:42:48 +0200 (Sat, 29 Jul 2006) | 1 line [Bug #1519571] Document some missing functions: setup(), title(), done() ........ r50933 | andrew.kuchling | 2006-07-29 16:43:55 +0200 (Sat, 29 Jul 2006) | 1 line Fix docstring punctuation ........ r50934 | andrew.kuchling | 2006-07-29 17:10:32 +0200 (Sat, 29 Jul 2006) | 1 line [Bug #1414697] Change docstring of set/frozenset types to specify that the contents are unique. Raymond, please feel free to edit or revert. ........ r50935 | andrew.kuchling | 2006-07-29 17:35:21 +0200 (Sat, 29 Jul 2006) | 1 line [Bug #1530382] Document SSL.server(), .issuer() methods ........ r50936 | andrew.kuchling | 2006-07-29 17:42:46 +0200 (Sat, 29 Jul 2006) | 1 line Typo fix ........ r50937 | andrew.kuchling | 2006-07-29 17:43:13 +0200 (Sat, 29 Jul 2006) | 1 line Tweak wording ........ r50938 | matt.fleming | 2006-07-29 17:55:30 +0200 (Sat, 29 Jul 2006) | 2 lines Fix typo ........ r50939 | andrew.kuchling | 2006-07-29 17:57:08 +0200 (Sat, 29 Jul 2006) | 6 lines [Bug #1528258] Mention that the 'data' argument can be None. The constructor docs referred the reader to the add_data() method's docs, but they weren't very helpful. I've simply copied an earlier explanation of 'data' that's more useful. ........ r50940 | andrew.kuchling | 2006-07-29 18:08:40 +0200 (Sat, 29 Jul 2006) | 1 line Set bug/patch count. Take a bow, everyone! ........ r50941 | fred.drake | 2006-07-29 18:56:15 +0200 (Sat, 29 Jul 2006) | 18 lines expunge the xmlcore changes: 41667, 41668 - initial switch to xmlcore 47044 - mention of xmlcore in What's New 50687 - mention of xmlcore in the library reference re-apply xmlcore changes to xml: 41674 - line ending changes (re-applied manually), directory props 41677 - add cElementTree wrapper 41678 - PSF licensing for etree 41812 - whitespace normalization 42724 - fix svn:eol-style settings 43681, 43682 - remove Python version-compatibility cruft from minidom 46773 - fix encoding of \r\n\t in attr values in saxutils 47269 - added XMLParser alias for cElementTree compatibility additional tests were added in Lib/test/test_sax.py that failed with the xmlcore changes; these relate to SF bugs #1511497, #1513611 ........ r50942 | andrew.kuchling | 2006-07-29 20:14:07 +0200 (Sat, 29 Jul 2006) | 17 lines Reorganize the docs for 'file' and 'open()' after some discussion with Fred. We want to encourage users to write open() when opening a file, but open() was described with a single paragraph and 'file' had lots of explanation of the mode and bufsize arguments. I've shrunk the description of 'file' to cross-reference to the 'File objects' section, and to open() for an explanation of the arguments. open() now has all the paragraphs about the mode string. The bufsize argument was moved up so that it isn't buried at the end; now there's 1 paragraph on mode, 1 on bufsize, and then 3 more on mode. Various other edits and rearrangements were made in the process. It's probably best to read the final text and not to try to make sense of the diffs. ........ r50943 | fred.drake | 2006-07-29 20:19:19 +0200 (Sat, 29 Jul 2006) | 1 line restore test un-intentionally removed in the xmlcore purge (revision 50941) ........ r50944 | fred.drake | 2006-07-29 20:33:29 +0200 (Sat, 29 Jul 2006) | 3 lines make the reference to older versions of the documentation a link to the right page on python.org ........ r50945 | fred.drake | 2006-07-29 21:09:01 +0200 (Sat, 29 Jul 2006) | 1 line document the footnote usage pattern ........ r50947 | fred.drake | 2006-07-29 21:14:10 +0200 (Sat, 29 Jul 2006) | 1 line emphasize and oddball nuance of LaTeX comment syntax ........ r50948 | andrew.kuchling | 2006-07-29 21:24:04 +0200 (Sat, 29 Jul 2006) | 1 line [Patch #1490989 from Skip Montanaro] Mention debugging builds in the API documentation. I've changed Skip's patch to point to Misc/SpecialBuilds and fiddled with the markup a bit. ........ r50949 | neal.norwitz | 2006-07-29 21:29:35 +0200 (Sat, 29 Jul 2006) | 6 lines Disable these tests until they are reliable across platforms. These problems may mask more important, real problems. One or both methods are known to fail on: Solaris, OpenBSD, Debian, Ubuntu. They pass on Windows and some Linux boxes. ........ r50950 | andrew.kuchling | 2006-07-29 21:50:37 +0200 (Sat, 29 Jul 2006) | 1 line [Patch #1068277] Clarify that os.path.exists() can return False depending on permissions. Fred approved committing this patch in December 2004! ........ r50952 | fred.drake | 2006-07-29 22:04:42 +0200 (Sat, 29 Jul 2006) | 6 lines SF bug #1193966: Weakref types documentation misplaced The information about supporting weakrefs with types defined in C extensions is moved to the Extending & Embedding manual. Py_TPFLAGS_HAVE_WEAKREFS is no longer mentioned since it is part of Py_TPFLAGS_DEFAULT. ........ r50953 | skip.montanaro | 2006-07-29 22:06:05 +0200 (Sat, 29 Jul 2006) | 4 lines Add a comment to the csv reader documentation that explains why the treatment of newlines changed in 2.5. Pulled almost verbatim from a comment by Andrew McNamara in <http://python.org/sf/1465014>. ........ r50954 | neal.norwitz | 2006-07-29 22:20:52 +0200 (Sat, 29 Jul 2006) | 3 lines If the executable doesn't exist, there's no reason to try to start it. This prevents garbage about command not found being printed on Solaris. ........ r50955 | fred.drake | 2006-07-29 22:21:25 +0200 (Sat, 29 Jul 2006) | 1 line fix minor markup error that introduced extra punctuation ........ r50957 | neal.norwitz | 2006-07-29 22:37:08 +0200 (Sat, 29 Jul 2006) | 3 lines Disable test_getnode too, since this is also unreliable. ........ r50958 | andrew.kuchling | 2006-07-29 23:27:12 +0200 (Sat, 29 Jul 2006) | 1 line Follow TeX's conventions for hyphens ........ r50959 | andrew.kuchling | 2006-07-29 23:30:21 +0200 (Sat, 29 Jul 2006) | 1 line Fix case for 'Unix' ........ r50960 | fred.drake | 2006-07-30 01:34:57 +0200 (Sun, 30 Jul 2006) | 1 line markup cleanups ........ r50961 | andrew.kuchling | 2006-07-30 02:27:34 +0200 (Sun, 30 Jul 2006) | 1 line Minor typo fixes ........ r50962 | andrew.kuchling | 2006-07-30 02:37:56 +0200 (Sun, 30 Jul 2006) | 1 line [Bug #793553] Correct description of keyword arguments for SSL authentication ........ r50963 | tim.peters | 2006-07-30 02:58:15 +0200 (Sun, 30 Jul 2006) | 2 lines Whitespace normalization. ........ r50964 | fred.drake | 2006-07-30 05:03:43 +0200 (Sun, 30 Jul 2006) | 1 line lots of markup nits, most commonly Unix/unix --> \UNIX ........ r50965 | fred.drake | 2006-07-30 07:41:28 +0200 (Sun, 30 Jul 2006) | 1 line update information on wxPython, from Robin Dunn ........ r50966 | fred.drake | 2006-07-30 07:49:49 +0200 (Sun, 30 Jul 2006) | 4 lines remove possibly-outdated comment on what GUI toolkit is most commonly used; it is hard to know whether this is right, and it does not add valuable reference information at any rate ........ r50967 | fred.drake | 2006-07-30 07:55:39 +0200 (Sun, 30 Jul 2006) | 3 lines - remove yet another reference to how commonly Tkinter is (thought to be) used - fix an internal section reference ........ r50968 | neal.norwitz | 2006-07-30 08:53:31 +0200 (Sun, 30 Jul 2006) | 4 lines Patch #1531113: Fix augmented assignment with yield expressions. Also fix a SystemError when trying to assign to yield expressions. ........ r50969 | neal.norwitz | 2006-07-30 08:55:48 +0200 (Sun, 30 Jul 2006) | 5 lines Add PyErr_WarnEx() so C code can pass the stacklevel to warnings.warn(). This provides the proper warning for struct.pack(). PyErr_Warn() is now deprecated in favor of PyErr_WarnEx(). As mentioned by Tim Peters on python-dev. ........ r50970 | neal.norwitz | 2006-07-30 08:57:04 +0200 (Sun, 30 Jul 2006) | 3 lines Bug #1515471: string.replace() accepts character buffers again. Pass the char* and size around rather than PyObject's. ........ r50971 | neal.norwitz | 2006-07-30 08:59:13 +0200 (Sun, 30 Jul 2006) | 1 line Whitespace normalization ........ r50973 | georg.brandl | 2006-07-30 12:53:32 +0200 (Sun, 30 Jul 2006) | 3 lines Clarify that __op__ methods must return NotImplemented if they don't support the operation. ........ r50974 | georg.brandl | 2006-07-30 13:07:23 +0200 (Sun, 30 Jul 2006) | 3 lines Bug #1002398: The documentation for os.path.sameopenfile now correctly refers to file descriptors, not file objects. ........ r50977 | martin.v.loewis | 2006-07-30 15:00:31 +0200 (Sun, 30 Jul 2006) | 3 lines Don't copy directory stat times in shutil.copytree on Windows Fixes #1525866. ........ r50978 | martin.v.loewis | 2006-07-30 15:14:05 +0200 (Sun, 30 Jul 2006) | 3 lines Base __version__ on sys.version_info, as distutils is no longer maintained separatedly. ........ r50979 | martin.v.loewis | 2006-07-30 15:27:31 +0200 (Sun, 30 Jul 2006) | 3 lines Mention Cygwin in distutils error message about a missing VS 2003. Fixes #1257728. ........ r50982 | martin.v.loewis | 2006-07-30 16:09:47 +0200 (Sun, 30 Jul 2006) | 5 lines Drop usage of test -e in configure as it is not portable. Fixes #1439538 Will backport to 2.4 Also regenerate pyconfig.h.in. ........ r50984 | georg.brandl | 2006-07-30 18:20:10 +0200 (Sun, 30 Jul 2006) | 3 lines Fix makefile changes for python-config. ........ r50985 | george.yoshida | 2006-07-30 18:37:37 +0200 (Sun, 30 Jul 2006) | 2 lines Rename struct.pack_to to struct.pack_into as changed in revision 46642. ........ r50986 | george.yoshida | 2006-07-30 18:41:30 +0200 (Sun, 30 Jul 2006) | 2 lines Typo fix ........ r50987 | neal.norwitz | 2006-07-30 21:18:13 +0200 (Sun, 30 Jul 2006) | 1 line Add some asserts and update comments ........ r50988 | neal.norwitz | 2006-07-30 21:18:38 +0200 (Sun, 30 Jul 2006) | 1 line Verify that the signal handlers were really called ........ r50989 | neal.norwitz | 2006-07-30 21:20:42 +0200 (Sun, 30 Jul 2006) | 3 lines Try to prevent hangs on Tru64/Alpha buildbot. I'm not certain this will help and may need to be reverted if it causes problems. ........ r50990 | georg.brandl | 2006-07-30 22:18:51 +0200 (Sun, 30 Jul 2006) | 2 lines Bug #1531349: right <-> left glitch in __rop__ description. ........ r50992 | tim.peters | 2006-07-31 03:46:03 +0200 (Mon, 31 Jul 2006) | 2 lines Whitespace normalization. ........ r50993 | andrew.mcnamara | 2006-07-31 04:27:48 +0200 (Mon, 31 Jul 2006) | 2 lines Redo the comment about the 2.5 change in quoted-newline handling. ........ r50994 | tim.peters | 2006-07-31 04:40:23 +0200 (Mon, 31 Jul 2006) | 10 lines ZipFile.close(): Killed one of the struct.pack deprecation warnings on Win32. Also added an XXX about the line: pos3 = self.fp.tell() `pos3` is never referenced, and I have no idea what the code intended to do instead. ........ r50996 | tim.peters | 2006-07-31 04:53:03 +0200 (Mon, 31 Jul 2006) | 8 lines ZipFile.close(): Kill the other struct.pack deprecation warning on Windows. Afraid I can't detect a pattern to when the pack formats decide to use a signed or unsigned format code -- appears nearly arbitrary to my eyes. So I left all the pack formats alone and changed the special-case data values instead. ........ r50997 | skip.montanaro | 2006-07-31 05:09:45 +0200 (Mon, 31 Jul 2006) | 1 line minor tweaks ........ r50998 | skip.montanaro | 2006-07-31 05:11:11 +0200 (Mon, 31 Jul 2006) | 1 line minor tweaks ........ r50999 | andrew.kuchling | 2006-07-31 14:20:24 +0200 (Mon, 31 Jul 2006) | 1 line Add refcounts for PyErr_WarnEx ........ r51000 | andrew.kuchling | 2006-07-31 14:39:05 +0200 (Mon, 31 Jul 2006) | 9 lines Document PyErr_WarnEx. (Bad Neal! No biscuit!) Is the explanation of the 'stacklevel' parameter clear? Please feel free to edit it. I don't have LaTeX installed on this machine, so haven't verified that the markup is correct. Will check tonight, or maybe the automatic doc build will tell me. ........ r51001 | andrew.kuchling | 2006-07-31 14:52:26 +0200 (Mon, 31 Jul 2006) | 1 line Add PyErr_WarnEx() ........ r51002 | andrew.kuchling | 2006-07-31 15:18:27 +0200 (Mon, 31 Jul 2006) | 1 line Mention csv newline changes ........ r51003 | andrew.kuchling | 2006-07-31 17:22:58 +0200 (Mon, 31 Jul 2006) | 1 line Typo fix ........ r51004 | andrew.kuchling | 2006-07-31 17:23:43 +0200 (Mon, 31 Jul 2006) | 1 line Remove reference to notation ........ r51005 | georg.brandl | 2006-07-31 18:00:34 +0200 (Mon, 31 Jul 2006) | 3 lines Fix function name. ........ r51006 | andrew.kuchling | 2006-07-31 18:10:24 +0200 (Mon, 31 Jul 2006) | 1 line [Bug #1514540] Instead of putting the standard types in a section, put them in a chapter of their own. This means string methods will now show up in the ToC. (Should the types come before or after the functions+exceptions+constants chapter? I've put them after, for now.) ........ r51007 | andrew.kuchling | 2006-07-31 18:22:05 +0200 (Mon, 31 Jul 2006) | 1 line [Bug #848556] Remove \d* from second alternative to avoid exponential case when repeating match ........ r51008 | andrew.kuchling | 2006-07-31 18:27:57 +0200 (Mon, 31 Jul 2006) | 1 line Update list of files; fix a typo ........ r51013 | andrew.kuchling | 2006-08-01 18:24:30 +0200 (Tue, 01 Aug 2006) | 1 line typo fix ........ r51018 | thomas.heller | 2006-08-01 18:54:43 +0200 (Tue, 01 Aug 2006) | 2 lines Fix a potential segfault and various potentail refcount leaks in the cast() function. ........ r51020 | thomas.heller | 2006-08-01 19:46:10 +0200 (Tue, 01 Aug 2006) | 1 line Minimal useful docstring for CopyComPointer. ........ r51021 | andrew.kuchling | 2006-08-01 20:16:15 +0200 (Tue, 01 Aug 2006) | 8 lines [Patch #1520905] Attempt to suppress core file created by test_subprocess.py. Patch by Douglas Greiman. The test_run_abort() testcase produces a core file on Unix systems, even though the test is successful. This can be confusing or alarming to someone who runs 'make test' and then finds that the Python interpreter apparently crashed. ........ r51023 | georg.brandl | 2006-08-01 20:49:24 +0200 (Tue, 01 Aug 2006) | 3 lines os.urandom no longer masks unrelated exceptions like SystemExit or KeyboardInterrupt. ........ r51025 | thomas.heller | 2006-08-01 21:14:15 +0200 (Tue, 01 Aug 2006) | 2 lines Speed up PyType_stgdict and PyObject_stgdict. ........ r51027 | ronald.oussoren | 2006-08-01 22:30:31 +0200 (Tue, 01 Aug 2006) | 3 lines Make sure the postinstall action that optionally updates the user's profile on MacOS X actually works correctly in all cases. ........ r51028 | ronald.oussoren | 2006-08-01 23:00:57 +0200 (Tue, 01 Aug 2006) | 4 lines This fixes bug #1527397: PythonLauncher runs scripts with the wrong working directory. It also fixes a bug where PythonLauncher failed to launch scripts when the scriptname (or the path to the script) contains quotes. ........ r51031 | tim.peters | 2006-08-02 05:27:46 +0200 (Wed, 02 Aug 2006) | 2 lines Whitespace normalization. ........ r51032 | tim.peters | 2006-08-02 06:12:36 +0200 (Wed, 02 Aug 2006) | 19 lines Try to squash struct.pack warnings on the "amd64 gentoo trunk" buildbot (& possibly other 64-bit boxes) during test_gzip. The native zlib crc32 function returns an unsigned 32-bit integer, which the Python wrapper implicitly casts to C long. Therefore the same crc can "look negative" on a 32-bit box but "look positive" on a 64-bit box. This patch papers over that platform difference when writing the crc to file. It may be better to change the Python wrapper, either to make the result "look positive" on all platforms (which means it may have to return a Python long at times on a 32-bit box), or to keep the sign the same across boxes. But that would be a visible change in what users see, while the current hack changes no visible behavior (well, apart from stopping the struct deprecation warning). Note that the module-level write32() function is no longer used. ........ r51033 | neal.norwitz | 2006-08-02 06:27:11 +0200 (Wed, 02 Aug 2006) | 4 lines Prevent memory leak on error. Reported by Klocwork #36 ........ r51034 | tim.peters | 2006-08-02 07:20:08 +0200 (Wed, 02 Aug 2006) | 9 lines _Stream.close(): Try to kill struct.pack() warnings when writing the crc to file on the "PPC64 Debian trunk" buildbot when running test_tarfile. This is again a case where the native zlib crc is an unsigned 32-bit int, but the Python wrapper implicitly casts it to signed C long, so that "the sign bit looks different" on different platforms. ........ r51035 | ronald.oussoren | 2006-08-02 08:10:10 +0200 (Wed, 02 Aug 2006) | 2 lines Updated documentation for the script that builds the OSX installer. ........ r51036 | neal.norwitz | 2006-08-02 08:14:22 +0200 (Wed, 02 Aug 2006) | 2 lines _PyWeakref_GetWeakrefCount() now returns a Py_ssize_t instead of long. ........ r51037 | neal.norwitz | 2006-08-02 08:15:10 +0200 (Wed, 02 Aug 2006) | 1 line v is already checked for NULL, so just DECREF it ........ r51038 | neal.norwitz | 2006-08-02 08:19:19 +0200 (Wed, 02 Aug 2006) | 1 line Let us know when there was a problem and the child had to kill the parent ........ r51039 | neal.norwitz | 2006-08-02 08:46:21 +0200 (Wed, 02 Aug 2006) | 5 lines Patch #1519025 and bug #926423: If a KeyboardInterrupt occurs during a socket operation on a socket with a timeout, the exception will be caught correctly. Previously, the exception was not caught. ........ r51040 | neal.norwitz | 2006-08-02 09:09:32 +0200 (Wed, 02 Aug 2006) | 1 line Add some explanation about Klocwork and Coverity static analysis ........ r51041 | anthony.baxter | 2006-08-02 09:43:09 +0200 (Wed, 02 Aug 2006) | 1 line pre-release machinations ........ r51043 | thomas.heller | 2006-08-02 13:35:31 +0200 (Wed, 02 Aug 2006) | 4 lines A few nore words about what ctypes does. Document that using the wrong calling convention can also raise 'ValueError: Procedure called with the wrong number of arguments'. ........ r51045 | thomas.heller | 2006-08-02 14:00:13 +0200 (Wed, 02 Aug 2006) | 1 line Fix a mistake. ........ r51046 | martin.v.loewis | 2006-08-02 15:53:55 +0200 (Wed, 02 Aug 2006) | 3 lines Correction of patch #1455898: In the mbcs decoder, set final=False for stream decoder, but final=True for the decode function. ........ r51049 | tim.peters | 2006-08-02 20:19:35 +0200 (Wed, 02 Aug 2006) | 2 lines Add missing svn:eol-style property to text files. ........ r51079 | neal.norwitz | 2006-08-04 06:50:21 +0200 (Fri, 04 Aug 2006) | 3 lines Bug #1531405, format_exception no longer raises an exception if str(exception) raised an exception. ........ r51080 | neal.norwitz | 2006-08-04 06:58:47 +0200 (Fri, 04 Aug 2006) | 11 lines Bug #1191458: tracing over for loops now produces a line event on each iteration. I'm not positive this is the best way to handle this. I'm also not sure that there aren't other cases where the lnotab is generated incorrectly. It would be great if people that use pdb or tracing could test heavily. Also: * Remove dead/duplicated code that wasn't used/necessary because we already handled the docstring prior to entering the loop. * add some debugging code into the compiler (#if 0'd out). ........ r51081 | neal.norwitz | 2006-08-04 07:09:28 +0200 (Fri, 04 Aug 2006) | 4 lines Bug #1333982: string/number constants were inappropriately stored in the byte code and co_consts even if they were not used, ie immediately popped off the stack. ........ r51082 | neal.norwitz | 2006-08-04 07:12:19 +0200 (Fri, 04 Aug 2006) | 1 line There were really two issues ........ r51084 | fred.drake | 2006-08-04 07:17:21 +0200 (Fri, 04 Aug 2006) | 1 line SF patch #1534048 (bug #1531003): fix typo in error message ........ r51085 | gregory.p.smith | 2006-08-04 07:17:47 +0200 (Fri, 04 Aug 2006) | 3 lines fix typos ........ r51087 | georg.brandl | 2006-08-04 08:03:53 +0200 (Fri, 04 Aug 2006) | 3 lines Fix bug caused by first decrefing, then increfing. ........ r51109 | neil.schemenauer | 2006-08-04 18:20:30 +0200 (Fri, 04 Aug 2006) | 5 lines Fix the 'compiler' package to generate correct code for MAKE_CLOSURE. In the 2.5 development cycle, MAKE_CLOSURE as changed to take free variables as a tuple rather than as individual items on the stack. Closes patch #1534084. ........ r51110 | georg.brandl | 2006-08-04 20:03:37 +0200 (Fri, 04 Aug 2006) | 3 lines Change fix for segfaulting property(), add a NEWS entry and a test. ........ r51111 | georg.brandl | 2006-08-04 20:07:34 +0200 (Fri, 04 Aug 2006) | 3 lines Better fix for bug #1531405, not executing str(value) twice. ........ r51112 | thomas.heller | 2006-08-04 20:17:40 +0200 (Fri, 04 Aug 2006) | 1 line On Windows, make PyErr_Warn an exported function again. ........ r51113 | thomas.heller | 2006-08-04 20:57:34 +0200 (Fri, 04 Aug 2006) | 4 lines Fix #1530448 - fix ctypes build failure on solaris 10. The '-mimpure-text' linker flag is required when linking _ctypes.so. ........ r51114 | thomas.heller | 2006-08-04 21:49:31 +0200 (Fri, 04 Aug 2006) | 3 lines Fix #1534738: win32 debug version of _msi must be _msi_d.pyd, not _msi.pyd. Fix the name of the pdb file as well. ........ r51115 | andrew.kuchling | 2006-08-04 22:37:43 +0200 (Fri, 04 Aug 2006) | 1 line Typo fixes ........ r51116 | andrew.kuchling | 2006-08-04 23:10:03 +0200 (Fri, 04 Aug 2006) | 1 line Fix mangled sentence ........ r51118 | tim.peters | 2006-08-05 00:00:35 +0200 (Sat, 05 Aug 2006) | 2 lines Whitespace normalization. ........ r51119 | bob.ippolito | 2006-08-05 01:59:21 +0200 (Sat, 05 Aug 2006) | 5 lines Fix #1530559, struct.pack raises TypeError where it used to convert. Passing float arguments to struct.pack when integers are expected now triggers a DeprecationWarning. ........ r51123 | georg.brandl | 2006-08-05 08:10:54 +0200 (Sat, 05 Aug 2006) | 3 lines Patch #1534922: correct and enhance unittest docs. ........ r51126 | georg.brandl | 2006-08-06 09:06:33 +0200 (Sun, 06 Aug 2006) | 2 lines Bug #1535182: really test the xreadlines() method of bz2 objects. ........ r51128 | georg.brandl | 2006-08-06 09:26:21 +0200 (Sun, 06 Aug 2006) | 4 lines Bug #1535081: A leading underscore has been added to the names of the md5 and sha modules, so add it in Modules/Setup.dist too. ........ r51129 | georg.brandl | 2006-08-06 10:23:54 +0200 (Sun, 06 Aug 2006) | 3 lines Bug #1535165: fixed a segfault in input() and raw_input() when sys.stdin is closed. ........ r51131 | georg.brandl | 2006-08-06 11:17:16 +0200 (Sun, 06 Aug 2006) | 2 lines Don't produce output in test_builtin. ........ r51133 | andrew.macintyre | 2006-08-06 14:37:03 +0200 (Sun, 06 Aug 2006) | 4 lines test_threading now skips testing alternate thread stack sizes on platforms that don't support changing thread stack size. ........ r51134 | andrew.kuchling | 2006-08-07 00:07:04 +0200 (Mon, 07 Aug 2006) | 2 lines [Patch #1464056] Ensure that we use the panelw library when linking with ncursesw. Once I see how the buildbots react, I'll backport this to 2.4. ........ r51137 | georg.brandl | 2006-08-08 13:52:34 +0200 (Tue, 08 Aug 2006) | 3 lines webbrowser: Silence stderr output if no gconftool or gnome browser found ........ r51138 | georg.brandl | 2006-08-08 13:56:21 +0200 (Tue, 08 Aug 2006) | 7 lines Remove "non-mapping" and "non-sequence" from TypeErrors raised by PyMapping_Size and PySequence_Size. Because len() tries first sequence, then mapping size, it will always raise a "non-mapping object has no len" error which is confusing. ........ r51139 | thomas.heller | 2006-08-08 19:37:00 +0200 (Tue, 08 Aug 2006) | 3 lines memcmp() can return values other than -1, 0, and +1 but tp_compare must not. ........ r51140 | thomas.heller | 2006-08-08 19:39:20 +0200 (Tue, 08 Aug 2006) | 1 line Remove accidently committed, duplicated test. ........ r51147 | andrew.kuchling | 2006-08-08 20:50:14 +0200 (Tue, 08 Aug 2006) | 1 line Reword paragraph to clarify ........ r51148 | andrew.kuchling | 2006-08-08 20:56:08 +0200 (Tue, 08 Aug 2006) | 1 line Move obmalloc item into C API section ........ r51149 | andrew.kuchling | 2006-08-08 21:00:14 +0200 (Tue, 08 Aug 2006) | 1 line 'Other changes' section now has only one item; move the item elsewhere and remove the section ........ r51150 | andrew.kuchling | 2006-08-08 21:00:34 +0200 (Tue, 08 Aug 2006) | 1 line Bump version number ........ r51151 | georg.brandl | 2006-08-08 22:11:22 +0200 (Tue, 08 Aug 2006) | 2 lines Bug #1536828: typo: TypeType should have been StringType. ........ r51153 | georg.brandl | 2006-08-08 22:13:13 +0200 (Tue, 08 Aug 2006) | 2 lines Bug #1536660: separate two words. ........ r51155 | georg.brandl | 2006-08-08 22:48:10 +0200 (Tue, 08 Aug 2006) | 3 lines ``str`` is now the same object as ``types.StringType``. ........ r51156 | tim.peters | 2006-08-09 02:52:26 +0200 (Wed, 09 Aug 2006) | 2 lines Whitespace normalization. ........ r51158 | georg.brandl | 2006-08-09 09:03:22 +0200 (Wed, 09 Aug 2006) | 4 lines Introduce an upper bound on tuple nesting depth in C argument format strings; fixes rest of #1523610. ........ r51160 | martin.v.loewis | 2006-08-09 09:57:39 +0200 (Wed, 09 Aug 2006) | 4 lines __hash__ may now return long int; the final hash value is obtained by invoking hash on the long int. Fixes #1536021. ........ r51168 | andrew.kuchling | 2006-08-09 15:03:41 +0200 (Wed, 09 Aug 2006) | 1 line [Bug #1536021] Mention __hash__ change ........ r51169 | andrew.kuchling | 2006-08-09 15:57:05 +0200 (Wed, 09 Aug 2006) | 1 line [Patch #1534027] Add notes on locale module changes ........ r51170 | andrew.kuchling | 2006-08-09 16:05:35 +0200 (Wed, 09 Aug 2006) | 1 line Add missing 'self' parameters ........ r51171 | andrew.kuchling | 2006-08-09 16:06:19 +0200 (Wed, 09 Aug 2006) | 1 line Reindent code ........ r51172 | armin.rigo | 2006-08-09 16:55:26 +0200 (Wed, 09 Aug 2006) | 2 lines Fix and test for an infinite C recursion. ........ r51173 | ronald.oussoren | 2006-08-09 16:56:33 +0200 (Wed, 09 Aug 2006) | 2 lines It's unlikely that future versions will require _POSIX_C_SOURCE ........ r51178 | armin.rigo | 2006-08-09 17:37:26 +0200 (Wed, 09 Aug 2006) | 2 lines Concatenation on a long string breaks (SF #1526585). ........ r51180 | kurt.kaiser | 2006-08-09 18:46:15 +0200 (Wed, 09 Aug 2006) | 8 lines 1. When used w/o subprocess, all exceptions were preceeded by an error message claiming they were IDLE internal errors (since 1.2a1). 2. Add Ronald Oussoren to CREDITS M NEWS.txt M PyShell.py M CREDITS.txt ........ r51181 | kurt.kaiser | 2006-08-09 19:47:15 +0200 (Wed, 09 Aug 2006) | 4 lines As a slight enhancement to the previous checkin, improve the internal error reporting by moving message to IDLE console. ........ r51182 | andrew.kuchling | 2006-08-09 20:23:14 +0200 (Wed, 09 Aug 2006) | 1 line Typo fix ........ r51183 | kurt.kaiser | 2006-08-09 22:34:46 +0200 (Wed, 09 Aug 2006) | 2 lines ToggleTab dialog was setting indent to 8 even if cancelled (since 1.2a1). ........ r51184 | martin.v.loewis | 2006-08-10 01:42:18 +0200 (Thu, 10 Aug 2006) | 2 lines Add some commentary on -mimpure-text. ........ r51185 | tim.peters | 2006-08-10 02:58:49 +0200 (Thu, 10 Aug 2006) | 2 lines Add missing svn:eol-style property to text files. ........ r51186 | kurt.kaiser | 2006-08-10 03:41:17 +0200 (Thu, 10 Aug 2006) | 2 lines Changing tokenize (39046) to detect dedent broke tabnanny check (since 1.2a1) ........ r51187 | tim.peters | 2006-08-10 05:01:26 +0200 (Thu, 10 Aug 2006) | 13 lines test_copytree_simple(): This was leaving behind two new temp directories each time it ran, at least on Windows. Several changes: explicitly closed all files; wrapped long lines; stopped suppressing errors when removing a file or directory fails (removing /shouldn't/ fail!); and changed what appeared to be incorrect usage of os.removedirs() (that doesn't remove empty directories at and /under/ the given path, instead it must be given an empty leaf directory and then deletes empty directories moving /up/ the path -- could be that the conceptually simpler shutil.rmtree() was really actually intended here). ........
2006-08-11 11:57:12 -03:00
McIlroy paper in particular has good analysis of a mergesort that's
probably strongly related to this one in its galloping strategy.
Galloping with a Broken Leg
---------------------------
So why don't we always gallop? Because it can lose, on two counts:
1. While we're willing to endure small per-merge overheads, per-comparison
overheads are a different story. Calling Yet Another Function per
comparison is expensive, and gallop_left() and gallop_right() are
too long-winded for sane inlining.
2. Galloping can-- alas --require more comparisons than linear one-at-time
search, depending on the data.
#2 requires details. If A[0] belongs before B[0], galloping requires 1
compare to determine that, same as linear search, except it costs more
to call the gallop function. If A[0] belongs right before B[1], galloping
requires 2 compares, again same as linear search. On the third compare,
galloping checks A[0] against B[3], and if it's <=, requires one more
compare to determine whether A[0] belongs at B[2] or B[3]. That's a total
of 4 compares, but if A[0] does belong at B[2], linear search would have
discovered that in only 3 compares, and that's a huge loss! Really. It's
an increase of 33% in the number of compares needed, and comparisons are
expensive in Python.
index in B where # compares linear # gallop # binary gallop
A[0] belongs search needs compares compares total
---------------- ----------------- -------- -------- ------
0 1 1 0 1
1 2 2 0 2
2 3 3 1 4
3 4 3 1 4
4 5 4 2 6
5 6 4 2 6
6 7 4 2 6
7 8 4 2 6
8 9 5 3 8
9 10 5 3 8
10 11 5 3 8
11 12 5 3 8
...
In general, if A[0] belongs at B[i], linear search requires i+1 comparisons
to determine that, and galloping a total of 2*floor(lg(i))+2 comparisons.
The advantage of galloping is unbounded as i grows, but it doesn't win at
all until i=6. Before then, it loses twice (at i=2 and i=4), and ties
at the other values. At and after i=6, galloping always wins.
We can't guess in advance when it's going to win, though, so we do one pair
at a time until the evidence seems strong that galloping may pay. MIN_GALLOP
1. Combined the base and length arrays into a single array of structs. This is friendlier for caches. 2. Cut MIN_GALLOP to 7, but added a per-sort min_gallop vrbl that adapts the "get into galloping mode" threshold higher when galloping isn't paying, and lower when it is. There's no known case where this hurts. It's (of course) neutral for /sort, \sort and =sort. It also happens to be neutral for !sort. It cuts a tiny # of compares in 3sort and +sort. For *sort, it reduces the # of compares to better than what this used to do when MIN_GALLOP was hardcoded to 10 (it did about 0.1% more *sort compares before, but given how close we are to the limit, this is "a lot"!). %sort used to do about 1.5% more compares, and ~sort about 3.6% more. Here are exact counts: i *sort 3sort +sort %sort ~sort !sort 15 449235 33019 33016 51328 188720 65534 before 448885 33016 33007 50426 182083 65534 after 0.08% 0.01% 0.03% 1.79% 3.65% 0.00% %ch from after 16 963714 65824 65809 103409 377634 131070 962991 65821 65808 101667 364341 131070 0.08% 0.00% 0.00% 1.71% 3.65% 0.00% 17 2059092 131413 131362 209130 755476 262142 2057533 131410 131361 206193 728871 262142 0.08% 0.00% 0.00% 1.42% 3.65% 0.00% 18 4380687 262440 262460 421998 1511174 524286 4377402 262437 262459 416347 1457945 524286 0.08% 0.00% 0.00% 1.36% 3.65% 0.00% 19 9285709 524581 524634 848590 3022584 1048574 9278734 524580 524633 837947 2916107 1048574 0.08% 0.00% 0.00% 1.27% 3.65% 0.00% 20 19621118 1048960 1048942 1715806 6045418 2097150 19606028 1048958 1048941 1694896 5832445 2097150 0.08% 0.00% 0.00% 1.23% 3.65% 0.00% 3. Added some key asserts I overlooked before. 4. Updated the doc file.
2002-08-10 02:21:15 -03:00
is 7, and that's pretty strong evidence. However, if the data is random, it
simply will trigger galloping mode purely by luck every now and again, and
it's quite likely to hit one of the losing cases next. On the other hand,
in cases like ~sort, galloping always pays, and MIN_GALLOP is larger than it
"should be" then. So the MergeState struct keeps a min_gallop variable
that merge_lo and merge_hi adjust: the longer we stay in galloping mode,
the smaller min_gallop gets, making it easier to transition back to
galloping mode (if we ever leave it in the current merge, and at the
start of the next merge). But whenever the gallop loop doesn't pay,
min_gallop is increased by one, making it harder to transition back
1. Combined the base and length arrays into a single array of structs. This is friendlier for caches. 2. Cut MIN_GALLOP to 7, but added a per-sort min_gallop vrbl that adapts the "get into galloping mode" threshold higher when galloping isn't paying, and lower when it is. There's no known case where this hurts. It's (of course) neutral for /sort, \sort and =sort. It also happens to be neutral for !sort. It cuts a tiny # of compares in 3sort and +sort. For *sort, it reduces the # of compares to better than what this used to do when MIN_GALLOP was hardcoded to 10 (it did about 0.1% more *sort compares before, but given how close we are to the limit, this is "a lot"!). %sort used to do about 1.5% more compares, and ~sort about 3.6% more. Here are exact counts: i *sort 3sort +sort %sort ~sort !sort 15 449235 33019 33016 51328 188720 65534 before 448885 33016 33007 50426 182083 65534 after 0.08% 0.01% 0.03% 1.79% 3.65% 0.00% %ch from after 16 963714 65824 65809 103409 377634 131070 962991 65821 65808 101667 364341 131070 0.08% 0.00% 0.00% 1.71% 3.65% 0.00% 17 2059092 131413 131362 209130 755476 262142 2057533 131410 131361 206193 728871 262142 0.08% 0.00% 0.00% 1.42% 3.65% 0.00% 18 4380687 262440 262460 421998 1511174 524286 4377402 262437 262459 416347 1457945 524286 0.08% 0.00% 0.00% 1.36% 3.65% 0.00% 19 9285709 524581 524634 848590 3022584 1048574 9278734 524580 524633 837947 2916107 1048574 0.08% 0.00% 0.00% 1.27% 3.65% 0.00% 20 19621118 1048960 1048942 1715806 6045418 2097150 19606028 1048958 1048941 1694896 5832445 2097150 0.08% 0.00% 0.00% 1.23% 3.65% 0.00% 3. Added some key asserts I overlooked before. 4. Updated the doc file.
2002-08-10 02:21:15 -03:00
to galloping mode (and again both within a merge and across merges). For
random data, this all but eliminates the gallop penalty: min_gallop grows
large enough that we almost never get into galloping mode. And for cases
like ~sort, min_gallop can fall to as low as 1. This seems to work well,
but in all it's a minor improvement over using a fixed MIN_GALLOP value.
Galloping Complication
----------------------
The description above was for merge_lo. merge_hi has to merge "from the
other end", and really needs to gallop starting at the last element in a run
instead of the first. Galloping from the first still works, but does more
comparisons than it should (this is significant -- I timed it both ways). For
this reason, the gallop_left() and gallop_right() (see note LEFT OR RIGHT)
functions have a "hint" argument, which is the index at which galloping
should begin. So galloping can actually start at any index, and proceed at
offsets of 1, 3, 7, 15, ... or -1, -3, -7, -15, ... from the starting index.
In the code as I type it's always called with either 0 or n-1 (where n is
the # of elements in a run). It's tempting to try to do something fancier,
melding galloping with some form of interpolation search; for example, if
we're merging a run of length 1 with a run of length 10000, index 5000 is
probably a better guess at the final result than either 0 or 9999. But
it's unclear how to generalize that intuition usefully, and merging of
wildly unbalanced runs already enjoys excellent performance.
~sort is a good example of when balanced runs could benefit from a better
hint value: to the extent possible, this would like to use a starting
offset equal to the previous value of acount/bcount. Doing so saves about
10% of the compares in ~sort. However, doing so is also a mixed bag,
hurting other cases.
Comparing Average # of Compares on Random Arrays
------------------------------------------------
1. Combined the base and length arrays into a single array of structs. This is friendlier for caches. 2. Cut MIN_GALLOP to 7, but added a per-sort min_gallop vrbl that adapts the "get into galloping mode" threshold higher when galloping isn't paying, and lower when it is. There's no known case where this hurts. It's (of course) neutral for /sort, \sort and =sort. It also happens to be neutral for !sort. It cuts a tiny # of compares in 3sort and +sort. For *sort, it reduces the # of compares to better than what this used to do when MIN_GALLOP was hardcoded to 10 (it did about 0.1% more *sort compares before, but given how close we are to the limit, this is "a lot"!). %sort used to do about 1.5% more compares, and ~sort about 3.6% more. Here are exact counts: i *sort 3sort +sort %sort ~sort !sort 15 449235 33019 33016 51328 188720 65534 before 448885 33016 33007 50426 182083 65534 after 0.08% 0.01% 0.03% 1.79% 3.65% 0.00% %ch from after 16 963714 65824 65809 103409 377634 131070 962991 65821 65808 101667 364341 131070 0.08% 0.00% 0.00% 1.71% 3.65% 0.00% 17 2059092 131413 131362 209130 755476 262142 2057533 131410 131361 206193 728871 262142 0.08% 0.00% 0.00% 1.42% 3.65% 0.00% 18 4380687 262440 262460 421998 1511174 524286 4377402 262437 262459 416347 1457945 524286 0.08% 0.00% 0.00% 1.36% 3.65% 0.00% 19 9285709 524581 524634 848590 3022584 1048574 9278734 524580 524633 837947 2916107 1048574 0.08% 0.00% 0.00% 1.27% 3.65% 0.00% 20 19621118 1048960 1048942 1715806 6045418 2097150 19606028 1048958 1048941 1694896 5832445 2097150 0.08% 0.00% 0.00% 1.23% 3.65% 0.00% 3. Added some key asserts I overlooked before. 4. Updated the doc file.
2002-08-10 02:21:15 -03:00
[NOTE: This was done when the new algorithm used about 0.1% more compares
on random data than does its current incarnation.]
Here list.sort() is samplesort, and list.msort() this sort:
"""
import random
from time import clock as now
def fill(n):
from random import random
2009-09-01 04:34:27 -03:00
return [random() for i in range(n)]
def mycmp(x, y):
global ncmp
ncmp += 1
return cmp(x, y)
def timeit(values, method):
global ncmp
X = values[:]
bound = getattr(X, method)
ncmp = 0
t1 = now()
bound(mycmp)
t2 = now()
return t2-t1, ncmp
format = "%5s %9.2f %11d"
f2 = "%5s %9.2f %11.2f"
def drive():
count = sst = sscmp = mst = mscmp = nelts = 0
while True:
n = random.randrange(100000)
nelts += n
x = fill(n)
t, c = timeit(x, 'sort')
sst += t
sscmp += c
t, c = timeit(x, 'msort')
mst += t
mscmp += c
count += 1
if count % 10:
continue
print "count", count, "nelts", nelts
print format % ("sort", sst, sscmp)
print format % ("msort", mst, mscmp)
print f2 % ("", (sst-mst)*1e2/mst, (sscmp-mscmp)*1e2/mscmp)
drive()
"""
I ran this on Windows and kept using the computer lightly while it was
running. time.clock() is wall-clock time on Windows, with better than
microsecond resolution. samplesort started with a 1.52% #-of-comparisons
disadvantage, fell quickly to 1.48%, and then fluctuated within that small
range. Here's the last chunk of output before I killed the job:
count 2630 nelts 130906543
sort 6110.80 1937887573
msort 6002.78 1909389381
1.80 1.49
We've done nearly 2 billion comparisons apiece at Python speed there, and
that's enough <wink>.
For random arrays of size 2 (yes, there are only 2 interesting ones),
samplesort has a 50%(!) comparison disadvantage. This is a consequence of
samplesort special-casing at most one ascending run at the start, then
falling back to the general case if it doesn't find an ascending run
immediately. The consequence is that it ends up using two compares to sort
[2, 1]. Gratifyingly, timsort doesn't do any special-casing, so had to be
taught how to deal with mixtures of ascending and descending runs
efficiently in all cases.
NOTES
-----
BINSORT
A "binary insertion sort" is just like a textbook insertion sort, but instead
of locating the correct position of the next item via linear (one at a time)
search, an equivalent to Python's bisect.bisect_right is used to find the
correct position in logarithmic time. Most texts don't mention this
variation, and those that do usually say it's not worth the bother: insertion
sort remains quadratic (expected and worst cases) either way. Speeding the
search doesn't reduce the quadratic data movement costs.
But in CPython's case, comparisons are extraordinarily expensive compared to
moving data, and the details matter. Moving objects is just copying
pointers. Comparisons can be arbitrarily expensive (can invoke arbitrary
user-supplied Python code), but even in simple cases (like 3 < 4) _all_
decisions are made at runtime: what's the type of the left comparand? the
type of the right? do they need to be coerced to a common type? where's the
code to compare these types? And so on. Even the simplest Python comparison
triggers a large pile of C-level pointer dereferences, conditionals, and
function calls.
So cutting the number of compares is almost always measurably helpful in
CPython, and the savings swamp the quadratic-time data movement costs for
reasonable minrun values.
LEFT OR RIGHT
gallop_left() and gallop_right() are akin to the Python bisect module's
bisect_left() and bisect_right(): they're the same unless the slice they're
searching contains a (at least one) value equal to the value being searched
for. In that case, gallop_left() returns the position immediately before the
leftmost equal value, and gallop_right() the position immediately after the
rightmost equal value. The distinction is needed to preserve stability. In
general, when merging adjacent runs A and B, gallop_left is used to search
thru B for where an element from A belongs, and gallop_right to search thru A
for where an element from B belongs.
REGION OF UNCERTAINTY
Two kinds of confusion seem to be common about the claim that after finding
a k such that
B[2**(k-1) - 1] < A[0] <= B[2**k - 1]
then a binary search requires exactly k-1 tries to find A[0]'s proper
location. For concreteness, say k=3, so B[3] < A[0] <= B[7].
The first confusion takes the form "OK, then the region of uncertainty is at
indices 3, 4, 5, 6 and 7: that's 5 elements, not the claimed 2**(k-1) - 1 =
3"; or the region is viewed as a Python slice and the objection is "but that's
the slice B[3:7], so has 7-3 = 4 elements". Resolution: we've already
compared A[0] against B[3] and against B[7], so A[0]'s correct location is
already known wrt _both_ endpoints. What remains is to find A[0]'s correct
location wrt B[4], B[5] and B[6], which spans 3 elements. Or in general, the
slice (leaving off both endpoints) (2**(k-1)-1)+1 through (2**k-1)-1
inclusive = 2**(k-1) through (2**k-1)-1 inclusive, which has
(2**k-1)-1 - 2**(k-1) + 1 =
2**k-1 - 2**(k-1) =
2*2**(k-1)-1 - 2**(k-1) =
(2-1)*2**(k-1) - 1 =
2**(k-1) - 1
elements.
The second confusion: "k-1 = 2 binary searches can find the correct location
among 2**(k-1) = 4 elements, but you're only applying it to 3 elements: we
could make this more efficient by arranging for the region of uncertainty to
span 2**(k-1) elements." Resolution: that confuses "elements" with
"locations". In a slice with N elements, there are N+1 _locations_. In the
example, with the region of uncertainty B[4], B[5], B[6], there are 4
locations: before B[4], between B[4] and B[5], between B[5] and B[6], and
after B[6]. In general, across 2**(k-1)-1 elements, there are 2**(k-1)
locations. That's why k-1 binary searches are necessary and sufficient.
OPTIMIZATION OF INDIVIDUAL COMPARISONS
As noted above, even the simplest Python comparison triggers a large pile of
C-level pointer dereferences, conditionals, and function calls. This can be
partially mitigated by pre-scanning the data to determine whether the data is
homogeneous with respect to type. If so, it is sometimes possible to
substitute faster type-specific comparisons for the slower, generic
PyObject_RichCompareBool.