summaryrefslogtreecommitdiff
path: root/Lib/test/test_bigmem.py
blob: 0e545950ee704ed8c8a5184b67647a523b839d17 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
"""Bigmem tests - tests for the 32-bit boundary in containers.

These tests try to exercise the 32-bit boundary that is sometimes, if
rarely, exceeded in practice, but almost never tested.  They are really only
meaningful on 64-bit builds on machines with a *lot* of memory, but the
tests are always run, usually with very low memory limits to make sure the
tests themselves don't suffer from bitrot.  To run them for real, pass a
high memory limit to regrtest, with the -M option.
"""

from test import support
from test.support import bigmemtest, _1G, _2G, _4G

import unittest
import operator
import sys
import functools

# These tests all use one of the bigmemtest decorators to indicate how much
# memory they use and how much memory they need to be even meaningful.  The
# decorators take two arguments: a 'memuse' indicator declaring
# (approximate) bytes per size-unit the test will use (at peak usage), and a
# 'minsize' indicator declaring a minimum *useful* size.  A test that
# allocates a bytestring to test various operations near the end will have a
# minsize of at least 2Gb (or it wouldn't reach the 32-bit limit, so the
# test wouldn't be very useful) and a memuse of 1 (one byte per size-unit,
# if it allocates only one big string at a time.)
#
# When run with a memory limit set, both decorators skip tests that need
# more memory than available to be meaningful.  The precisionbigmemtest will
# always pass minsize as size, even if there is much more memory available.
# The bigmemtest decorator will scale size upward to fill available memory.
#
# Bigmem testing houserules:
#
#  - Try not to allocate too many large objects. It's okay to rely on
#    refcounting semantics, and don't forget that 's = create_largestring()'
#    doesn't release the old 's' (if it exists) until well after its new
#    value has been created. Use 'del s' before the create_largestring call.
#
#  - Do *not* compare large objects using assertEqual, assertIn or similar.
#    It's a lengthy operation and the errormessage will be utterly useless
#    due to its size.  To make sure whether a result has the right contents,
#    better to use the strip or count methods, or compare meaningful slices.
#
#  - Don't forget to test for large indices, offsets and results and such,
#    in addition to large sizes. Anything that probes the 32-bit boundary.
#
#  - When repeating an object (say, a substring, or a small list) to create
#    a large object, make the subobject of a length that is not a power of
#    2. That way, int-wrapping problems are more easily detected.
#
#  - Despite the bigmemtest decorator, all tests will actually be called
#    with a much smaller number too, in the normal test run (5Kb currently.)
#    This is so the tests themselves get frequent testing.
#    Consequently, always make all large allocations based on the
#    passed-in 'size', and don't rely on the size being very large. Also,
#    memuse-per-size should remain sane (less than a few thousand); if your
#    test uses more, adjust 'size' upward, instead.

# BEWARE: it seems that one failing test can yield other subsequent tests to
# fail as well. I do not know whether it is due to memory fragmentation
# issues, or other specifics of the platform malloc() routine.

ascii_char_size = 1
ucs2_char_size = 2
ucs4_char_size = 4


class BaseStrTest:

    def _test_capitalize(self, size):
        _ = self.from_latin1
        SUBSTR = self.from_latin1(' abc def ghi')
        s = _('-') * size + SUBSTR
        caps = s.capitalize()
        self.assertEqual(caps[-len(SUBSTR):],
                         SUBSTR.capitalize())
        self.assertEqual(caps.lstrip(_('-')), SUBSTR)

    @bigmemtest(size=_2G + 10, memuse=1)
    def test_center(self, size):
        SUBSTR = self.from_latin1(' abc def ghi')
        s = SUBSTR.center(size)
        self.assertEqual(len(s), size)
        lpadsize = rpadsize = (len(s) - len(SUBSTR)) // 2
        if len(s) % 2:
            lpadsize += 1
        self.assertEqual(s[lpadsize:-rpadsize], SUBSTR)
        self.assertEqual(s.strip(), SUBSTR.strip())

    @bigmemtest(size=_2G, memuse=2)
    def test_count(self, size):
        _ = self.from_latin1
        SUBSTR = _(' abc def ghi')
        s = _('.') * size + SUBSTR
        self.assertEqual(s.count(_('.')), size)
        s += _('.')
        self.assertEqual(s.count(_('.')), size + 1)
        self.assertEqual(s.count(_(' ')), 3)
        self.assertEqual(s.count(_('i')), 1)
        self.assertEqual(s.count(_('j')), 0)

    @bigmemtest(size=_2G, memuse=2)
    def test_endswith(self, size):
        _ = self.from_latin1
        SUBSTR = _(' abc def ghi')
        s = _('-') * size + SUBSTR
        self.assertTrue(s.endswith(SUBSTR))
        self.assertTrue(s.endswith(s))
        s2 = _('...') + s
        self.assertTrue(s2.endswith(s))
        self.assertFalse(s.endswith(_('a') + SUBSTR))
        self.assertFalse(SUBSTR.endswith(s))

    @bigmemtest(size=_2G + 10, memuse=2)
    def test_expandtabs(self, size):
        _ = self.from_latin1
        s = _('-') * size
        tabsize = 8
        self.assertTrue(s.expandtabs() == s)
        del s
        slen, remainder = divmod(size, tabsize)
        s = _('       \t') * slen
        s = s.expandtabs(tabsize)
        self.assertEqual(len(s), size - remainder)
        self.assertEqual(len(s.strip(_(' '))), 0)

    @bigmemtest(size=_2G, memuse=2)
    def test_find(self, size):
        _ = self.from_latin1
        SUBSTR = _(' abc def ghi')
        sublen = len(SUBSTR)
        s = _('').join([SUBSTR, _('-') * size, SUBSTR])
        self.assertEqual(s.find(_(' ')), 0)
        self.assertEqual(s.find(SUBSTR), 0)
        self.assertEqual(s.find(_(' '), sublen), sublen + size)
        self.assertEqual(s.find(SUBSTR, len(SUBSTR)), sublen + size)
        self.assertEqual(s.find(_('i')), SUBSTR.find(_('i')))
        self.assertEqual(s.find(_('i'), sublen),
                         sublen + size + SUBSTR.find(_('i')))
        self.assertEqual(s.find(_('i'), size),
                         sublen + size + SUBSTR.find(_('i')))
        self.assertEqual(s.find(_('j')), -1)

    @bigmemtest(size=_2G, memuse=2)
    def test_index(self, size):
        _ = self.from_latin1
        SUBSTR = _(' abc def ghi')
        sublen = len(SUBSTR)
        s = _('').join([SUBSTR, _('-') * size, SUBSTR])
        self.assertEqual(s.index(_(' ')), 0)
        self.assertEqual(s.index(SUBSTR), 0)
        self.assertEqual(s.index(_(' '), sublen), sublen + size)
        self.assertEqual(s.index(SUBSTR, sublen), sublen + size)
        self.assertEqual(s.index(_('i')), SUBSTR.index(_('i')))
        self.assertEqual(s.index(_('i'), sublen),
                         sublen + size + SUBSTR.index(_('i')))
        self.assertEqual(s.index(_('i'), size),
                         sublen + size + SUBSTR.index(_('i')))
        self.assertRaises(ValueError, s.index, _('j'))

    @bigmemtest(size=_2G, memuse=2)
    def test_isalnum(self, size):
        _ = self.from_latin1
        SUBSTR = _('123456')
        s = _('a') * size + SUBSTR
        self.assertTrue(s.isalnum())
        s += _('.')
        self.assertFalse(s.isalnum())

    @bigmemtest(size=_2G, memuse=2)
    def test_isalpha(self, size):
        _ = self.from_latin1
        SUBSTR = _('zzzzzzz')
        s = _('a') * size + SUBSTR
        self.assertTrue(s.isalpha())
        s += _('.')
        self.assertFalse(s.isalpha())

    @bigmemtest(size=_2G, memuse=2)
    def test_isdigit(self, size):
        _ = self.from_latin1
        SUBSTR = _('123456')
        s = _('9') * size + SUBSTR
        self.assertTrue(s.isdigit())
        s += _('z')
        self.assertFalse(s.isdigit())

    @bigmemtest(size=_2G, memuse=2)
    def test_islower(self, size):
        _ = self.from_latin1
        chars = _(''.join(
            chr(c) for c in range(255) if not chr(c).isupper()))
        repeats = size // len(chars) + 2
        s = chars * repeats
        self.assertTrue(s.islower())
        s += _('A')
        self.assertFalse(s.islower())

    @bigmemtest(size=_2G, memuse=2)
    def test_isspace(self, size):
        _ = self.from_latin1
        whitespace = _(' \f\n\r\t\v')
        repeats = size // len(whitespace) + 2
        s = whitespace * repeats
        self.assertTrue(s.isspace())
        s += _('j')
        self.assertFalse(s.isspace())

    @bigmemtest(size=_2G, memuse=2)
    def test_istitle(self, size):
        _ = self.from_latin1
        SUBSTR = _('123456')
        s = _('').join([_('A'), _('a') * size, SUBSTR])
        self.assertTrue(s.istitle())
        s += _('A')
        self.assertTrue(s.istitle())
        s += _('aA')
        self.assertFalse(s.istitle())

    @bigmemtest(size=_2G, memuse=2)
    def test_isupper(self, size):
        _ = self.from_latin1
        chars = _(''.join(
            chr(c) for c in range(255) if not chr(c).islower()))
        repeats = size // len(chars) + 2
        s = chars * repeats
        self.assertTrue(s.isupper())
        s += _('a')
        self.assertFalse(s.isupper())

    @bigmemtest(size=_2G, memuse=2)
    def test_join(self, size):
        _ = self.from_latin1
        s = _('A') * size
        x = s.join([_('aaaaa'), _('bbbbb')])
        self.assertEqual(x.count(_('a')), 5)
        self.assertEqual(x.count(_('b')), 5)
        self.assertTrue(x.startswith(_('aaaaaA')))
        self.assertTrue(x.endswith(_('Abbbbb')))

    @bigmemtest(size=_2G + 10, memuse=1)
    def test_ljust(self, size):
        _ = self.from_latin1
        SUBSTR = _(' abc def ghi')
        s = SUBSTR.ljust(size)
        self.assertTrue(s.startswith(SUBSTR + _('  ')))
        self.assertEqual(len(s), size)
        self.assertEqual(s.strip(), SUBSTR.strip())

    @bigmemtest(size=_2G + 10, memuse=2)
    def test_lower(self, size):
        _ = self.from_latin1
        s = _('A') * size
        s = s.lower()
        self.assertEqual(len(s), size)
        self.assertEqual(s.count(_('a')), size)

    @bigmemtest(size=_2G + 10, memuse=1)
    def test_lstrip(self, size):
        _ = self.from_latin1
        SUBSTR = _('abc def ghi')
        s = SUBSTR.rjust(size)
        self.assertEqual(len(s), size)
        self.assertEqual(s.lstrip(), SUBSTR.lstrip())
        del s
        s = SUBSTR.ljust(size)
        self.assertEqual(len(s), size)
        # Type-specific optimization
        if isinstance(s, (str, bytes)):
            stripped = s.lstrip()
            self.assertTrue(stripped is s)

    @bigmemtest(size=_2G + 10, memuse=2)
    def test_replace(self, size):
        _ = self.from_latin1
        replacement = _('a')
        s = _(' ') * size
        s = s.replace(_(' '), replacement)
        self.assertEqual(len(s), size)
        self.assertEqual(s.count(replacement), size)
        s = s.replace(replacement, _(' '), size - 4)
        self.assertEqual(len(s), size)
        self.assertEqual(s.count(replacement), 4)
        self.assertEqual(s[-10:], _('      aaaa'))

    @bigmemtest(size=_2G, memuse=2)
    def test_rfind(self, size):
        _ = self.from_latin1
        SUBSTR = _(' abc def ghi')
        sublen = len(SUBSTR)
        s = _('').join([SUBSTR, _('-') * size, SUBSTR])
        self.assertEqual(s.rfind(_(' ')), sublen + size + SUBSTR.rfind(_(' ')))
        self.assertEqual(s.rfind(SUBSTR), sublen + size)
        self.assertEqual(s.rfind(_(' '), 0, size), SUBSTR.rfind(_(' ')))
        self.assertEqual(s.rfind(SUBSTR, 0, sublen + size), 0)
        self.assertEqual(s.rfind(_('i')), sublen + size + SUBSTR.rfind(_('i')))
        self.assertEqual(s.rfind(_('i'), 0, sublen), SUBSTR.rfind(_('i')))
        self.assertEqual(s.rfind(_('i'), 0, sublen + size),
                         SUBSTR.rfind(_('i')))
        self.assertEqual(s.rfind(_('j')), -1)

    @bigmemtest(size=_2G, memuse=2)
    def test_rindex(self, size):
        _ = self.from_latin1
        SUBSTR = _(' abc def ghi')
        sublen = len(SUBSTR)
        s = _('').join([SUBSTR, _('-') * size, SUBSTR])
        self.assertEqual(s.rindex(_(' ')),
                         sublen + size + SUBSTR.rindex(_(' ')))
        self.assertEqual(s.rindex(SUBSTR), sublen + size)
        self.assertEqual(s.rindex(_(' '), 0, sublen + size - 1),
                         SUBSTR.rindex(_(' ')))
        self.assertEqual(s.rindex(SUBSTR, 0, sublen + size), 0)
        self.assertEqual(s.rindex(_('i')),
                         sublen + size + SUBSTR.rindex(_('i')))
        self.assertEqual(s.rindex(_('i'), 0, sublen), SUBSTR.rindex(_('i')))
        self.assertEqual(s.rindex(_('i'), 0, sublen + size),
                         SUBSTR.rindex(_('i')))
        self.assertRaises(ValueError, s.rindex, _('j'))

    @bigmemtest(size=_2G + 10, memuse=1)
    def test_rjust(self, size):
        _ = self.from_latin1
        SUBSTR = _(' abc def ghi')
        s = SUBSTR.ljust(size)
        self.assertTrue(s.startswith(SUBSTR + _('  ')))
        self.assertEqual(len(s), size)
        self.assertEqual(s.strip(), SUBSTR.strip())

    @bigmemtest(size=_2G + 10, memuse=1)
    def test_rstrip(self, size):
        _ = self.from_latin1
        SUBSTR = _(' abc def ghi')
        s = SUBSTR.ljust(size)
        self.assertEqual(len(s), size)
        self.assertEqual(s.rstrip(), SUBSTR.rstrip())
        del s
        s = SUBSTR.rjust(size)
        self.assertEqual(len(s), size)
        # Type-specific optimization
        if isinstance(s, (str, bytes)):
            stripped = s.rstrip()
            self.assertTrue(stripped is s)

    # The test takes about size bytes to build a string, and then about
    # sqrt(size) substrings of sqrt(size) in size and a list to
    # hold sqrt(size) items. It's close but just over 2x size.
    @bigmemtest(size=_2G, memuse=2.1)
    def test_split_small(self, size):
        _ = self.from_latin1
        # Crudely calculate an estimate so that the result of s.split won't
        # take up an inordinate amount of memory
        chunksize = int(size ** 0.5 + 2)
        SUBSTR = _('a') + _(' ') * chunksize
        s = SUBSTR * chunksize
        l = s.split()
        self.assertEqual(len(l), chunksize)
        expected = _('a')
        for item in l:
            self.assertEqual(item, expected)
        del l
        l = s.split(_('a'))
        self.assertEqual(len(l), chunksize + 1)
        expected = _(' ') * chunksize
        for item in filter(None, l):
            self.assertEqual(item, expected)

    # Allocates a string of twice size (and briefly two) and a list of
    # size.  Because of internal affairs, the s.split() call produces a
    # list of size times the same one-character string, so we only
    # suffer for the list size. (Otherwise, it'd cost another 48 times
    # size in bytes!) Nevertheless, a list of size takes
    # 8*size bytes.
    @bigmemtest(size=_2G + 5, memuse=2 * ascii_char_size + 8)
    def test_split_large(self, size):
        _ = self.from_latin1
        s = _(' a') * size + _(' ')
        l = s.split()
        self.assertEqual(len(l), size)
        self.assertEqual(set(l), set([_('a')]))
        del l
        l = s.split(_('a'))
        self.assertEqual(len(l), size + 1)
        self.assertEqual(set(l), set([_(' ')]))

    @bigmemtest(size=_2G, memuse=2.1)
    def test_splitlines(self, size):
        _ = self.from_latin1
        # Crudely calculate an estimate so that the result of s.split won't
        # take up an inordinate amount of memory
        chunksize = int(size ** 0.5 + 2) // 2
        SUBSTR = _(' ') * chunksize + _('\n') + _(' ') * chunksize + _('\r\n')
        s = SUBSTR * (chunksize * 2)
        l = s.splitlines()
        self.assertEqual(len(l), chunksize * 4)
        expected = _(' ') * chunksize
        for item in l:
            self.assertEqual(item, expected)

    @bigmemtest(size=_2G, memuse=2)
    def test_startswith(self, size):
        _ = self.from_latin1
        SUBSTR = _(' abc def ghi')
        s = _('-') * size + SUBSTR
        self.assertTrue(s.startswith(s))
        self.assertTrue(s.startswith(_('-') * size))
        self.assertFalse(s.startswith(SUBSTR))

    @bigmemtest(size=_2G, memuse=1)
    def test_strip(self, size):
        _ = self.from_latin1
        SUBSTR = _('   abc def ghi   ')
        s = SUBSTR.rjust(size)
        self.assertEqual(len(s), size)
        self.assertEqual(s.strip(), SUBSTR.strip())
        del s
        s = SUBSTR.ljust(size)
        self.assertEqual(len(s), size)
        self.assertEqual(s.strip(), SUBSTR.strip())

    def _test_swapcase(self, size):
        _ = self.from_latin1
        SUBSTR = _("aBcDeFG12.'\xa9\x00")
        sublen = len(SUBSTR)
        repeats = size // sublen + 2
        s = SUBSTR * repeats
        s = s.swapcase()
        self.assertEqual(len(s), sublen * repeats)
        self.assertEqual(s[:sublen * 3], SUBSTR.swapcase() * 3)
        self.assertEqual(s[-sublen * 3:], SUBSTR.swapcase() * 3)

    def _test_title(self, size):
        _ = self.from_latin1
        SUBSTR = _('SpaaHAaaAaham')
        s = SUBSTR * (size // len(SUBSTR) + 2)
        s = s.title()
        self.assertTrue(s.startswith((SUBSTR * 3).title()))
        self.assertTrue(s.endswith(SUBSTR.lower() * 3))

    @bigmemtest(size=_2G, memuse=2)
    def test_translate(self, size):
        _ = self.from_latin1
        SUBSTR = _('aZz.z.Aaz.')
        trans = bytes.maketrans(b'.aZ', b'-!$')
        sublen = len(SUBSTR)
        repeats = size // sublen + 2
        s = SUBSTR * repeats
        s = s.translate(trans)
        self.assertEqual(len(s), repeats * sublen)
        self.assertEqual(s[:sublen], SUBSTR.translate(trans))
        self.assertEqual(s[-sublen:], SUBSTR.translate(trans))
        self.assertEqual(s.count(_('.')), 0)
        self.assertEqual(s.count(_('!')), repeats * 2)
        self.assertEqual(s.count(_('z')), repeats * 3)

    @bigmemtest(size=_2G + 5, memuse=2)
    def test_upper(self, size):
        _ = self.from_latin1
        s = _('a') * size
        s = s.upper()
        self.assertEqual(len(s), size)
        self.assertEqual(s.count(_('A')), size)

    @bigmemtest(size=_2G + 20, memuse=1)
    def test_zfill(self, size):
        _ = self.from_latin1
        SUBSTR = _('-568324723598234')
        s = SUBSTR.zfill(size)
        self.assertTrue(s.endswith(_('0') + SUBSTR[1:]))
        self.assertTrue(s.startswith(_('-0')))
        self.assertEqual(len(s), size)
        self.assertEqual(s.count(_('0')), size - len(SUBSTR))

    # This test is meaningful even with size < 2G, as long as the
    # doubled string is > 2G (but it tests more if both are > 2G :)
    @bigmemtest(size=_1G + 2, memuse=3)
    def test_concat(self, size):
        _ = self.from_latin1
        s = _('.') * size
        self.assertEqual(len(s), size)
        s = s + s
        self.assertEqual(len(s), size * 2)
        self.assertEqual(s.count(_('.')), size * 2)

    # This test is meaningful even with size < 2G, as long as the
    # repeated string is > 2G (but it tests more if both are > 2G :)
    @bigmemtest(size=_1G + 2, memuse=3)
    def test_repeat(self, size):
        _ = self.from_latin1
        s = _('.') * size
        self.assertEqual(len(s), size)
        s = s * 2
        self.assertEqual(len(s), size * 2)
        self.assertEqual(s.count(_('.')), size * 2)

    @bigmemtest(size=_2G + 20, memuse=2)
    def test_slice_and_getitem(self, size):
        _ = self.from_latin1
        SUBSTR = _('0123456789')
        sublen = len(SUBSTR)
        s = SUBSTR * (size // sublen)
        stepsize = len(s) // 100
        stepsize = stepsize - (stepsize % sublen)
        for i in range(0, len(s) - stepsize, stepsize):
            self.assertEqual(s[i], SUBSTR[0])
            self.assertEqual(s[i:i + sublen], SUBSTR)
            self.assertEqual(s[i:i + sublen:2], SUBSTR[::2])
            if i > 0:
                self.assertEqual(s[i + sublen - 1:i - 1:-3],
                                 SUBSTR[sublen::-3])
        # Make sure we do some slicing and indexing near the end of the
        # string, too.
        self.assertEqual(s[len(s) - 1], SUBSTR[-1])
        self.assertEqual(s[-1], SUBSTR[-1])
        self.assertEqual(s[len(s) - 10], SUBSTR[0])
        self.assertEqual(s[-sublen], SUBSTR[0])
        self.assertEqual(s[len(s):], _(''))
        self.assertEqual(s[len(s) - 1:], SUBSTR[-1:])
        self.assertEqual(s[-1:], SUBSTR[-1:])
        self.assertEqual(s[len(s) - sublen:], SUBSTR)
        self.assertEqual(s[-sublen:], SUBSTR)
        self.assertEqual(len(s[:]), len(s))
        self.assertEqual(len(s[:len(s) - 5]), len(s) - 5)
        self.assertEqual(len(s[5:-5]), len(s) - 10)

        self.assertRaises(IndexError, operator.getitem, s, len(s))
        self.assertRaises(IndexError, operator.getitem, s, len(s) + 1)
        self.assertRaises(IndexError, operator.getitem, s, len(s) + 1<<31)

    @bigmemtest(size=_2G, memuse=2)
    def test_contains(self, size):
        _ = self.from_latin1
        SUBSTR = _('0123456789')
        edge = _('-') * (size // 2)
        s = _('').join([edge, SUBSTR, edge])
        del edge
        self.assertTrue(SUBSTR in s)
        self.assertFalse(SUBSTR * 2 in s)
        self.assertTrue(_('-') in s)
        self.assertFalse(_('a') in s)
        s += _('a')
        self.assertTrue(_('a') in s)

    @bigmemtest(size=_2G + 10, memuse=2)
    def test_compare(self, size):
        _ = self.from_latin1
        s1 = _('-') * size
        s2 = _('-') * size
        self.assertTrue(s1 == s2)
        del s2
        s2 = s1 + _('a')
        self.assertFalse(s1 == s2)
        del s2
        s2 = _('.') * size
        self.assertFalse(s1 == s2)

    @bigmemtest(size=_2G + 10, memuse=1)
    def test_hash(self, size):
        # Not sure if we can do any meaningful tests here...  Even if we
        # start relying on the exact algorithm used, the result will be
        # different depending on the size of the C 'long int'.  Even this
        # test is dodgy (there's no *guarantee* that the two things should
        # have a different hash, even if they, in the current
        # implementation, almost always do.)
        _ = self.from_latin1
        s = _('\x00') * size
        h1 = hash(s)
        del s
        s = _('\x00') * (size + 1)
        self.assertNotEqual(h1, hash(s))


class StrTest(unittest.TestCase, BaseStrTest):

    def from_latin1(self, s):
        return s

    def basic_encode_test(self, size, enc, c='.', expectedsize=None):
        if expectedsize is None:
            expectedsize = size
        try:
            s = c * size
            self.assertEqual(len(s.encode(enc)), expectedsize)
        finally:
            s = None

    def setUp(self):
        # HACK: adjust memory use of tests inherited from BaseStrTest
        # according to character size.
        self._adjusted = {}
        for name in dir(BaseStrTest):
            if not name.startswith('test_'):
                continue
            meth = getattr(type(self), name)
            try:
                memuse = meth.memuse
            except AttributeError:
                continue
            meth.memuse = ascii_char_size * memuse
            self._adjusted[name] = memuse

    def tearDown(self):
        for name, memuse in self._adjusted.items():
            getattr(type(self), name).memuse = memuse

    @bigmemtest(size=_2G, memuse=ucs4_char_size * 3)
    def test_capitalize(self, size):
        self._test_capitalize(size)

    @bigmemtest(size=_2G, memuse=ucs4_char_size * 3)
    def test_title(self, size):
        self._test_title(size)

    @bigmemtest(size=_2G, memuse=ucs4_char_size * 3)
    def test_swapcase(self, size):
        self._test_swapcase(size)

    # Many codecs convert to the legacy representation first, explaining
    # why we add 'ucs4_char_size' to the 'memuse' below.

    @bigmemtest(size=_2G + 2, memuse=ascii_char_size + 1)
    def test_encode(self, size):
        return self.basic_encode_test(size, 'utf-8')

    @bigmemtest(size=_4G // 6 + 2, memuse=ascii_char_size + ucs4_char_size + 1)
    def test_encode_raw_unicode_escape(self, size):
        try:
            return self.basic_encode_test(size, 'raw_unicode_escape')
        except MemoryError:
            pass # acceptable on 32-bit

    @bigmemtest(size=_4G // 5 + 70, memuse=ascii_char_size + ucs4_char_size + 1)
    def test_encode_utf7(self, size):
        try:
            return self.basic_encode_test(size, 'utf7')
        except MemoryError:
            pass # acceptable on 32-bit

    @bigmemtest(size=_4G // 4 + 5, memuse=ascii_char_size + ucs4_char_size + 4)
    def test_encode_utf32(self, size):
        try:
            return self.basic_encode_test(size, 'utf32', expectedsize=4 * size + 4)
        except MemoryError:
            pass # acceptable on 32-bit

    @bigmemtest(size=_2G - 1, memuse=ascii_char_size + 1)
    def test_encode_ascii(self, size):
        return self.basic_encode_test(size, 'ascii', c='A')

    # str % (...) uses a Py_UCS4 intermediate representation

    @bigmemtest(size=_2G + 10, memuse=ascii_char_size * 2 + ucs4_char_size)
    def test_format(self, size):
        s = '-' * size
        sf = '%s' % (s,)
        self.assertTrue(s == sf)
        del sf
        sf = '..%s..' % (s,)
        self.assertEqual(len(sf), len(s) + 4)
        self.assertTrue(sf.startswith('..-'))
        self.assertTrue(sf.endswith('-..'))
        del s, sf

        size //= 2
        edge = '-' * size
        s = ''.join([edge, '%s', edge])
        del edge
        s = s % '...'
        self.assertEqual(len(s), size * 2 + 3)
        self.assertEqual(s.count('.'), 3)
        self.assertEqual(s.count('-'), size * 2)

    @bigmemtest(size=_2G + 10, memuse=ascii_char_size * 2)
    def test_repr_small(self, size):
        s = '-' * size
        s = repr(s)
        self.assertEqual(len(s), size + 2)
        self.assertEqual(s[0], "'")
        self.assertEqual(s[-1], "'")
        self.assertEqual(s.count('-'), size)
        del s
        # repr() will create a string four times as large as this 'binary
        # string', but we don't want to allocate much more than twice
        # size in total.  (We do extra testing in test_repr_large())
        size = size // 5 * 2
        s = '\x00' * size
        s = repr(s)
        self.assertEqual(len(s), size * 4 + 2)
        self.assertEqual(s[0], "'")
        self.assertEqual(s[-1], "'")
        self.assertEqual(s.count('\\'), size)
        self.assertEqual(s.count('0'), size * 2)

    @bigmemtest(size=_2G + 10, memuse=ascii_char_size * 5)
    def test_repr_large(self, size):
        s = '\x00' * size
        s = repr(s)
        self.assertEqual(len(s), size * 4 + 2)
        self.assertEqual(s[0], "'")
        self.assertEqual(s[-1], "'")
        self.assertEqual(s.count('\\'), size)
        self.assertEqual(s.count('0'), size * 2)

    # ascii() calls encode('ascii', 'backslashreplace'), which itself
    # creates a temporary Py_UNICODE representation in addition to the
    # original (Py_UCS2) one
    # There's also some overallocation when resizing the ascii() result
    # that isn't taken into account here.
    @bigmemtest(size=_2G // 5 + 1, memuse=ucs2_char_size +
                                          ucs4_char_size + ascii_char_size * 6)
    def test_unicode_repr(self, size):
        # Use an assigned, but not printable code point.
        # It is in the range of the low surrogates \uDC00-\uDFFF.
        char = "\uDCBA"
        s = char * size
        try:
            for f in (repr, ascii):
                r = f(s)
                self.assertEqual(len(r), 2 + (len(f(char)) - 2) * size)
                self.assertTrue(r.endswith(r"\udcba'"), r[-10:])
                r = None
        finally:
            r = s = None

    @bigmemtest(size=_2G // 5 + 1, memuse=ucs4_char_size * 2 + ascii_char_size * 10)
    def test_unicode_repr_wide(self, size):
        char = "\U0001DCBA"
        s = char * size
        try:
            for f in (repr, ascii):
                r = f(s)
                self.assertEqual(len(r), 2 + (len(f(char)) - 2) * size)
                self.assertTrue(r.endswith(r"\U0001dcba'"), r[-12:])
                r = None
        finally:
            r = s = None

    # The original test_translate is overriden here, so as to get the
    # correct size estimate: str.translate() uses an intermediate Py_UCS4
    # representation.

    @bigmemtest(size=_2G, memuse=ascii_char_size * 2 + ucs4_char_size)
    def test_translate(self, size):
        _ = self.from_latin1
        SUBSTR = _('aZz.z.Aaz.')
        trans = {
            ord(_('.')): _('-'),
            ord(_('a')): _('!'),
            ord(_('Z')): _('$'),
        }
        sublen = len(SUBSTR)
        repeats = size // sublen + 2
        s = SUBSTR * repeats
        s = s.translate(trans)
        self.assertEqual(len(s), repeats * sublen)
        self.assertEqual(s[:sublen], SUBSTR.translate(trans))
        self.assertEqual(s[-sublen:], SUBSTR.translate(trans))
        self.assertEqual(s.count(_('.')), 0)
        self.assertEqual(s.count(_('!')), repeats * 2)
        self.assertEqual(s.count(_('z')), repeats * 3)


class BytesTest(unittest.TestCase, BaseStrTest):

    def from_latin1(self, s):
        return s.encode("latin-1")

    @bigmemtest(size=_2G + 2, memuse=1 + ascii_char_size)
    def test_decode(self, size):
        s = self.from_latin1('.') * size
        self.assertEqual(len(s.decode('utf-8')), size)

    @bigmemtest(size=_2G, memuse=2)
    def test_capitalize(self, size):
        self._test_capitalize(size)

    @bigmemtest(size=_2G, memuse=2)
    def test_title(self, size):
        self._test_title(size)

    @bigmemtest(size=_2G, memuse=2)
    def test_swapcase(self, size):
        self._test_swapcase(size)


class BytearrayTest(unittest.TestCase, BaseStrTest):

    def from_latin1(self, s):
        return bytearray(s.encode("latin-1"))

    @bigmemtest(size=_2G + 2, memuse=1 + ascii_char_size)
    def test_decode(self, size):
        s = self.from_latin1('.') * size
        self.assertEqual(len(s.decode('utf-8')), size)

    @bigmemtest(size=_2G, memuse=2)
    def test_capitalize(self, size):
        self._test_capitalize(size)

    @bigmemtest(size=_2G, memuse=2)
    def test_title(self, size):
        self._test_title(size)

    @bigmemtest(size=_2G, memuse=2)
    def test_swapcase(self, size):
        self._test_swapcase(size)

    test_hash = None
    test_split_large = None

class TupleTest(unittest.TestCase):

    # Tuples have a small, fixed-sized head and an array of pointers to
    # data.  Since we're testing 64-bit addressing, we can assume that the
    # pointers are 8 bytes, and that thus that the tuples take up 8 bytes
    # per size.

    # As a side-effect of testing long tuples, these tests happen to test
    # having more than 2<<31 references to any given object. Hence the
    # use of different types of objects as contents in different tests.

    @bigmemtest(size=_2G + 2, memuse=16)
    def test_compare(self, size):
        t1 = ('',) * size
        t2 = ('',) * size
        self.assertTrue(t1 == t2)
        del t2
        t2 = ('',) * (size + 1)
        self.assertFalse(t1 == t2)
        del t2
        t2 = (1,) * size
        self.assertFalse(t1 == t2)

    # Test concatenating into a single tuple of more than 2G in length,
    # and concatenating a tuple of more than 2G in length separately, so
    # the smaller test still gets run even if there isn't memory for the
    # larger test (but we still let the tester know the larger test is
    # skipped, in verbose mode.)
    def basic_concat_test(self, size):
        t = ((),) * size
        self.assertEqual(len(t), size)
        t = t + t
        self.assertEqual(len(t), size * 2)

    @bigmemtest(size=_2G // 2 + 2, memuse=24)
    def test_concat_small(self, size):
        return self.basic_concat_test(size)

    @bigmemtest(size=_2G + 2, memuse=24)
    def test_concat_large(self, size):
        return self.basic_concat_test(size)

    @bigmemtest(size=_2G // 5 + 10, memuse=8 * 5)
    def test_contains(self, size):
        t = (1, 2, 3, 4, 5) * size
        self.assertEqual(len(t), size * 5)
        self.assertTrue(5 in t)
        self.assertFalse((1, 2, 3, 4, 5) in t)
        self.assertFalse(0 in t)

    @bigmemtest(size=_2G + 10, memuse=8)
    def test_hash(self, size):
        t1 = (0,) * size
        h1 = hash(t1)
        del t1
        t2 = (0,) * (size + 1)
        self.assertFalse(h1 == hash(t2))

    @bigmemtest(size=_2G + 10, memuse=8)
    def test_index_and_slice(self, size):
        t = (None,) * size
        self.assertEqual(len(t), size)
        self.assertEqual(t[-1], None)
        self.assertEqual(t[5], None)
        self.assertEqual(t[size - 1], None)
        self.assertRaises(IndexError, operator.getitem, t, size)
        self.assertEqual(t[:5], (None,) * 5)
        self.assertEqual(t[-5:], (None,) * 5)
        self.assertEqual(t[20:25], (None,) * 5)
        self.assertEqual(t[-25:-20], (None,) * 5)
        self.assertEqual(t[size - 5:], (None,) * 5)
        self.assertEqual(t[size - 5:size], (None,) * 5)
        self.assertEqual(t[size - 6:size - 2], (None,) * 4)
        self.assertEqual(t[size:size], ())
        self.assertEqual(t[size:size+5], ())

    # Like test_concat, split in two.
    def basic_test_repeat(self, size):
        t = ('',) * size
        self.assertEqual(len(t), size)
        t = t * 2
        self.assertEqual(len(t), size * 2)

    @bigmemtest(size=_2G // 2 + 2, memuse=24)
    def test_repeat_small(self, size):
        return self.basic_test_repeat(size)

    @bigmemtest(size=_2G + 2, memuse=24)
    def test_repeat_large(self, size):
        return self.basic_test_repeat(size)

    @bigmemtest(size=_1G - 1, memuse=12)
    def test_repeat_large_2(self, size):
        return self.basic_test_repeat(size)

    @bigmemtest(size=_1G - 1, memuse=9)
    def test_from_2G_generator(self, size):
        self.skipTest("test needs much more memory than advertised, see issue5438")
        try:
            t = tuple(range(size))
        except MemoryError:
            pass # acceptable on 32-bit
        else:
            count = 0
            for item in t:
                self.assertEqual(item, count)
                count += 1
            self.assertEqual(count, size)

    @bigmemtest(size=_1G - 25, memuse=9)
    def test_from_almost_2G_generator(self, size):
        self.skipTest("test needs much more memory than advertised, see issue5438")
        try:
            t = tuple(range(size))
            count = 0
            for item in t:
                self.assertEqual(item, count)
                count += 1
            self.assertEqual(count, size)
        except MemoryError:
            pass # acceptable, expected on 32-bit

    # Like test_concat, split in two.
    def basic_test_repr(self, size):
        t = (0,) * size
        s = repr(t)
        # The repr of a tuple of 0's is exactly three times the tuple length.
        self.assertEqual(len(s), size * 3)
        self.assertEqual(s[:5], '(0, 0')
        self.assertEqual(s[-5:], '0, 0)')
        self.assertEqual(s.count('0'), size)

    @bigmemtest(size=_2G // 3 + 2, memuse=8 + 3 * ascii_char_size)
    def test_repr_small(self, size):
        return self.basic_test_repr(size)

    @bigmemtest(size=_2G + 2, memuse=8 + 3 * ascii_char_size)
    def test_repr_large(self, size):
        return self.basic_test_repr(size)

class ListTest(unittest.TestCase):

    # Like tuples, lists have a small, fixed-sized head and an array of
    # pointers to data, so 8 bytes per size. Also like tuples, we make the
    # lists hold references to various objects to test their refcount
    # limits.

    @bigmemtest(size=_2G + 2, memuse=16)
    def test_compare(self, size):
        l1 = [''] * size
        l2 = [''] * size
        self.assertTrue(l1 == l2)
        del l2
        l2 = [''] * (size + 1)
        self.assertFalse(l1 == l2)
        del l2
        l2 = [2] * size
        self.assertFalse(l1 == l2)

    # Test concatenating into a single list of more than 2G in length,
    # and concatenating a list of more than 2G in length separately, so
    # the smaller test still gets run even if there isn't memory for the
    # larger test (but we still let the tester know the larger test is
    # skipped, in verbose mode.)
    def basic_test_concat(self, size):
        l = [[]] * size
        self.assertEqual(len(l), size)
        l = l + l
        self.assertEqual(len(l), size * 2)

    @bigmemtest(size=_2G // 2 + 2, memuse=24)
    def test_concat_small(self, size):
        return self.basic_test_concat(size)

    @bigmemtest(size=_2G + 2, memuse=24)
    def test_concat_large(self, size):
        return self.basic_test_concat(size)

    def basic_test_inplace_concat(self, size):
        l = [sys.stdout] * size
        l += l
        self.assertEqual(len(l), size * 2)
        self.assertTrue(l[0] is l[-1])
        self.assertTrue(l[size - 1] is l[size + 1])

    @bigmemtest(size=_2G // 2 + 2, memuse=24)
    def test_inplace_concat_small(self, size):
        return self.basic_test_inplace_concat(size)

    @bigmemtest(size=_2G + 2, memuse=24)
    def test_inplace_concat_large(self, size):
        return self.basic_test_inplace_concat(size)

    @bigmemtest(size=_2G // 5 + 10, memuse=8 * 5)
    def test_contains(self, size):
        l = [1, 2, 3, 4, 5] * size
        self.assertEqual(len(l), size * 5)
        self.assertTrue(5 in l)
        self.assertFalse([1, 2, 3, 4, 5] in l)
        self.assertFalse(0 in l)

    @bigmemtest(size=_2G + 10, memuse=8)
    def test_hash(self, size):
        l = [0] * size
        self.assertRaises(TypeError, hash, l)

    @bigmemtest(size=_2G + 10, memuse=8)
    def test_index_and_slice(self, size):
        l = [None] * size
        self.assertEqual(len(l), size)
        self.assertEqual(l[-1], None)
        self.assertEqual(l[5], None)
        self.assertEqual(l[size - 1], None)
        self.assertRaises(IndexError, operator.getitem, l, size)
        self.assertEqual(l[:5], [None] * 5)
        self.assertEqual(l[-5:], [None] * 5)
        self.assertEqual(l[20:25], [None] * 5)
        self.assertEqual(l[-25:-20], [None] * 5)
        self.assertEqual(l[size - 5:], [None] * 5)
        self.assertEqual(l[size - 5:size], [None] * 5)
        self.assertEqual(l[size - 6:size - 2], [None] * 4)
        self.assertEqual(l[size:size], [])
        self.assertEqual(l[size:size+5], [])

        l[size - 2] = 5
        self.assertEqual(len(l), size)
        self.assertEqual(l[-3:], [None, 5, None])
        self.assertEqual(l.count(5), 1)
        self.assertRaises(IndexError, operator.setitem, l, size, 6)
        self.assertEqual(len(l), size)

        l[size - 7:] = [1, 2, 3, 4, 5]
        size -= 2
        self.assertEqual(len(l), size)
        self.assertEqual(l[-7:], [None, None, 1, 2, 3, 4, 5])

        l[:7] = [1, 2, 3, 4, 5]
        size -= 2
        self.assertEqual(len(l), size)
        self.assertEqual(l[:7], [1, 2, 3, 4, 5, None, None])

        del l[size - 1]
        size -= 1
        self.assertEqual(len(l), size)
        self.assertEqual(l[-1], 4)

        del l[-2:]
        size -= 2
        self.assertEqual(len(l), size)
        self.assertEqual(l[-1], 2)

        del l[0]
        size -= 1
        self.assertEqual(len(l), size)
        self.assertEqual(l[0], 2)

        del l[:2]
        size -= 2
        self.assertEqual(len(l), size)
        self.assertEqual(l[0], 4)

    # Like test_concat, split in two.
    def basic_test_repeat(self, size):
        l = [] * size
        self.assertFalse(l)
        l = [''] * size
        self.assertEqual(len(l), size)
        l = l * 2
        self.assertEqual(len(l), size * 2)

    @bigmemtest(size=_2G // 2 + 2, memuse=24)
    def test_repeat_small(self, size):
        return self.basic_test_repeat(size)

    @bigmemtest(size=_2G + 2, memuse=24)
    def test_repeat_large(self, size):
        return self.basic_test_repeat(size)

    def basic_test_inplace_repeat(self, size):
        l = ['']
        l *= size
        self.assertEqual(len(l), size)
        self.assertTrue(l[0] is l[-1])
        del l

        l = [''] * size
        l *= 2
        self.assertEqual(len(l), size * 2)
        self.assertTrue(l[size - 1] is l[-1])

    @bigmemtest(size=_2G // 2 + 2, memuse=16)
    def test_inplace_repeat_small(self, size):
        return self.basic_test_inplace_repeat(size)

    @bigmemtest(size=_2G + 2, memuse=16)
    def test_inplace_repeat_large(self, size):
        return self.basic_test_inplace_repeat(size)

    def basic_test_repr(self, size):
        l = [0] * size
        s = repr(l)
        # The repr of a list of 0's is exactly three times the list length.
        self.assertEqual(len(s), size * 3)
        self.assertEqual(s[:5], '[0, 0')
        self.assertEqual(s[-5:], '0, 0]')
        self.assertEqual(s.count('0'), size)

    @bigmemtest(size=_2G // 3 + 2, memuse=8 + 3 * ascii_char_size)
    def test_repr_small(self, size):
        return self.basic_test_repr(size)

    @bigmemtest(size=_2G + 2, memuse=8 + 3 * ascii_char_size)
    def test_repr_large(self, size):
        return self.basic_test_repr(size)

    # list overallocates ~1/8th of the total size (on first expansion) so
    # the single list.append call puts memuse at 9 bytes per size.
    @bigmemtest(size=_2G, memuse=9)
    def test_append(self, size):
        l = [object()] * size
        l.append(object())
        self.assertEqual(len(l), size+1)
        self.assertTrue(l[-3] is l[-2])
        self.assertFalse(l[-2] is l[-1])

    @bigmemtest(size=_2G // 5 + 2, memuse=8 * 5)
    def test_count(self, size):
        l = [1, 2, 3, 4, 5] * size
        self.assertEqual(l.count(1), size)
        self.assertEqual(l.count("1"), 0)

    def basic_test_extend(self, size):
        l = [object] * size
        l.extend(l)
        self.assertEqual(len(l), size * 2)
        self.assertTrue(l[0] is l[-1])
        self.assertTrue(l[size - 1] is l[size + 1])

    @bigmemtest(size=_2G // 2 + 2, memuse=16)
    def test_extend_small(self, size):
        return self.basic_test_extend(size)

    @bigmemtest(size=_2G + 2, memuse=16)
    def test_extend_large(self, size):
        return self.basic_test_extend(size)

    @bigmemtest(size=_2G // 5 + 2, memuse=8 * 5)
    def test_index(self, size):
        l = [1, 2, 3, 4, 5] * size
        size *= 5
        self.assertEqual(l.index(1), 0)
        self.assertEqual(l.index(5, size - 5), size - 1)
        self.assertEqual(l.index(5, size - 5, size), size - 1)
        self.assertRaises(ValueError, l.index, 1, size - 4, size)
        self.assertRaises(ValueError, l.index, 6)

    # This tests suffers from overallocation, just like test_append.
    @bigmemtest(size=_2G + 10, memuse=9)
    def test_insert(self, size):
        l = [1.0] * size
        l.insert(size - 1, "A")
        size += 1
        self.assertEqual(len(l), size)
        self.assertEqual(l[-3:], [1.0, "A", 1.0])

        l.insert(size + 1, "B")
        size += 1
        self.assertEqual(len(l), size)
        self.assertEqual(l[-3:], ["A", 1.0, "B"])

        l.insert(1, "C")
        size += 1
        self.assertEqual(len(l), size)
        self.assertEqual(l[:3], [1.0, "C", 1.0])
        self.assertEqual(l[size - 3:], ["A", 1.0, "B"])

    @bigmemtest(size=_2G // 5 + 4, memuse=8 * 5)
    def test_pop(self, size):
        l = ["a", "b", "c", "d", "e"] * size
        size *= 5
        self.assertEqual(len(l), size)

        item = l.pop()
        size -= 1
        self.assertEqual(len(l), size)
        self.assertEqual(item, "e")
        self.assertEqual(l[-2:], ["c", "d"])

        item = l.pop(0)
        size -= 1
        self.assertEqual(len(l), size)
        self.assertEqual(item, "a")
        self.assertEqual(l[:2], ["b", "c"])

        item = l.pop(size - 2)
        size -= 1
        self.assertEqual(len(l), size)
        self.assertEqual(item, "c")
        self.assertEqual(l[-2:], ["b", "d"])

    @bigmemtest(size=_2G + 10, memuse=8)
    def test_remove(self, size):
        l = [10] * size
        self.assertEqual(len(l), size)

        l.remove(10)
        size -= 1
        self.assertEqual(len(l), size)

        # Because of the earlier l.remove(), this append doesn't trigger
        # a resize.
        l.append(5)
        size += 1
        self.assertEqual(len(l), size)
        self.assertEqual(l[-2:], [10, 5])
        l.remove(5)
        size -= 1
        self.assertEqual(len(l), size)
        self.assertEqual(l[-2:], [10, 10])

    @bigmemtest(size=_2G // 5 + 2, memuse=8 * 5)
    def test_reverse(self, size):
        l = [1, 2, 3, 4, 5] * size
        l.reverse()
        self.assertEqual(len(l), size * 5)
        self.assertEqual(l[-5:], [5, 4, 3, 2, 1])
        self.assertEqual(l[:5], [5, 4, 3, 2, 1])

    @bigmemtest(size=_2G // 5 + 2, memuse=8 * 5)
    def test_sort(self, size):
        l = [1, 2, 3, 4, 5] * size
        l.sort()
        self.assertEqual(len(l), size * 5)
        self.assertEqual(l.count(1), size)
        self.assertEqual(l[:10], [1] * 10)
        self.assertEqual(l[-10:], [5] * 10)

def test_main():
    support.run_unittest(StrTest, BytesTest, BytearrayTest,
        TupleTest, ListTest)

if __name__ == '__main__':
    if len(sys.argv) > 1:
        support.set_memlimit(sys.argv[1])
    test_main()