summaryrefslogtreecommitdiff
path: root/js/src/irregexp/RegExpEngine.cpp
blob: 0fb783170e5dcf0612ebefa69dbe0421dc17f652 (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
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 4 -*- */

// Copyright 2012 the V8 project authors. All rights reserved.
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
//     * Redistributions of source code must retain the above copyright
//       notice, this list of conditions and the following disclaimer.
//     * Redistributions in binary form must reproduce the above
//       copyright notice, this list of conditions and the following
//       disclaimer in the documentation and/or other materials provided
//       with the distribution.
//     * Neither the name of Google Inc. nor the names of its
//       contributors may be used to endorse or promote products derived
//       from this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

#include "irregexp/RegExpEngine.h"

#include "irregexp/NativeRegExpMacroAssembler.h"
#include "irregexp/RegExpCharacters.h"
#include "irregexp/RegExpMacroAssembler.h"
#include "jit/ExecutableAllocator.h"
#include "jit/JitCommon.h"

using namespace js;
using namespace js::irregexp;

using mozilla::DebugOnly;
using mozilla::Maybe;

#define DEFINE_ACCEPT(Type)                                          \
    void Type##Node::Accept(NodeVisitor* visitor) {                  \
        visitor->Visit##Type(this);                                  \
    }
FOR_EACH_NODE_TYPE(DEFINE_ACCEPT)
#undef DEFINE_ACCEPT

void LoopChoiceNode::Accept(NodeVisitor* visitor) {
    visitor->VisitLoopChoice(this);
}

static const int kMaxLookaheadForBoyerMoore = 8;

RegExpNode::RegExpNode(LifoAlloc* alloc)
  : replacement_(nullptr), trace_count_(0), alloc_(alloc)
{
    bm_info_[0] = bm_info_[1] = nullptr;
}

bool
RegExpCharacterClass::is_standard(LifoAlloc* alloc)
{
    // TODO(lrn): Remove need for this function, by not throwing away information
    // along the way.
    if (is_negated_)
        return false;
    if (set_.is_standard())
        return true;
    if (CharacterRange::CompareRanges(set_.ranges(alloc), kSpaceRanges, kSpaceRangeCount)) {
        set_.set_standard_set_type('s');
        return true;
    }
    if (CharacterRange::CompareInverseRanges(set_.ranges(alloc), kSpaceRanges, kSpaceRangeCount)) {
        set_.set_standard_set_type('S');
        return true;
    }
    if (CharacterRange::CompareInverseRanges(set_.ranges(alloc),
                             kLineTerminatorRanges,
                             kLineTerminatorRangeCount)) {
        set_.set_standard_set_type('.');
        return true;
    }
    if (CharacterRange::CompareRanges(set_.ranges(alloc),
                      kLineTerminatorRanges,
                      kLineTerminatorRangeCount)) {
        set_.set_standard_set_type('n');
        return true;
    }
    if (CharacterRange::CompareRanges(set_.ranges(alloc), kWordRanges, kWordRangeCount)) {
        set_.set_standard_set_type('w');
        return true;
    }
    if (CharacterRange::CompareInverseRanges(set_.ranges(alloc), kWordRanges, kWordRangeCount)) {
        set_.set_standard_set_type('W');
        return true;
    }
    return false;
}

// -------------------------------------------------------------------
// SeqRegExpNode

class VisitMarker
{
  public:
    explicit VisitMarker(NodeInfo* info)
      : info_(info)
    {
        MOZ_ASSERT(!info->visited);
        info->visited = true;
    }
    ~VisitMarker() {
        info_->visited = false;
    }
  private:
    NodeInfo* info_;
};

bool
SeqRegExpNode::FillInBMInfo(int offset,
                            int budget,
                            BoyerMooreLookahead* bm,
                            bool not_at_start)
{
    if (!bm->CheckOverRecursed())
        return false;
    if (!on_success_->FillInBMInfo(offset, budget - 1, bm, not_at_start))
        return false;
    if (offset == 0)
        set_bm_info(not_at_start, bm);
    return true;
}

RegExpNode*
SeqRegExpNode::FilterASCII(int depth, bool ignore_case, bool unicode)
{
    if (info()->replacement_calculated)
        return replacement();

    if (depth < 0)
        return this;

    MOZ_ASSERT(!info()->visited);
    VisitMarker marker(info());
    return FilterSuccessor(depth - 1, ignore_case, unicode);
}

RegExpNode*
SeqRegExpNode::FilterSuccessor(int depth, bool ignore_case, bool unicode)
{
    RegExpNode* next = on_success_->FilterASCII(depth - 1, ignore_case, unicode);
    if (next == nullptr)
        return set_replacement(nullptr);

    on_success_ = next;
    return set_replacement(this);
}

// -------------------------------------------------------------------
// ActionNode

int
ActionNode::EatsAtLeast(int still_to_find, int budget, bool not_at_start)
{
    if (budget <= 0)
        return 0;
    if (action_type_ == POSITIVE_SUBMATCH_SUCCESS)
        return 0;  // Rewinds input!
    return on_success()->EatsAtLeast(still_to_find,
                                     budget - 1,
                                     not_at_start);
}

bool
ActionNode::FillInBMInfo(int offset,
                         int budget,
                         BoyerMooreLookahead* bm,
                         bool not_at_start)
{
    if (!bm->CheckOverRecursed())
        return false;

    if (action_type_ == BEGIN_SUBMATCH) {
        bm->SetRest(offset);
    } else if (action_type_ != POSITIVE_SUBMATCH_SUCCESS) {
        if (!on_success()->FillInBMInfo(offset, budget - 1, bm, not_at_start))
            return false;
    }
    SaveBMInfo(bm, not_at_start, offset);

    return true;
}

/* static */ ActionNode*
ActionNode::SetRegister(int reg,
                        int val,
                        RegExpNode* on_success)
{
    ActionNode* result = on_success->alloc()->newInfallible<ActionNode>(SET_REGISTER, on_success);
    result->data_.u_store_register.reg = reg;
    result->data_.u_store_register.value = val;
    return result;
}

/* static */ ActionNode*
ActionNode::IncrementRegister(int reg, RegExpNode* on_success)
{
    ActionNode* result = on_success->alloc()->newInfallible<ActionNode>(INCREMENT_REGISTER, on_success);
    result->data_.u_increment_register.reg = reg;
    return result;
}

/* static */ ActionNode*
ActionNode::StorePosition(int reg, bool is_capture, RegExpNode* on_success)
{
    ActionNode* result = on_success->alloc()->newInfallible<ActionNode>(STORE_POSITION, on_success);
    result->data_.u_position_register.reg = reg;
    result->data_.u_position_register.is_capture = is_capture;
    return result;
}

/* static */ ActionNode*
ActionNode::ClearCaptures(Interval range, RegExpNode* on_success)
{
    ActionNode* result = on_success->alloc()->newInfallible<ActionNode>(CLEAR_CAPTURES, on_success);
    result->data_.u_clear_captures.range_from = range.from();
    result->data_.u_clear_captures.range_to = range.to();
    return result;
}

/* static */ ActionNode*
ActionNode::BeginSubmatch(int stack_pointer_reg, int position_reg, RegExpNode* on_success)
{
    ActionNode* result = on_success->alloc()->newInfallible<ActionNode>(BEGIN_SUBMATCH, on_success);
    result->data_.u_submatch.stack_pointer_register = stack_pointer_reg;
    result->data_.u_submatch.current_position_register = position_reg;
    return result;
}

/* static */ ActionNode*
ActionNode::PositiveSubmatchSuccess(int stack_pointer_reg,
                                    int restore_reg,
                                    int clear_capture_count,
                                    int clear_capture_from,
                                    RegExpNode* on_success)
{
    ActionNode* result = on_success->alloc()->newInfallible<ActionNode>(POSITIVE_SUBMATCH_SUCCESS, on_success);
    result->data_.u_submatch.stack_pointer_register = stack_pointer_reg;
    result->data_.u_submatch.current_position_register = restore_reg;
    result->data_.u_submatch.clear_register_count = clear_capture_count;
    result->data_.u_submatch.clear_register_from = clear_capture_from;
    return result;
}

/* static */ ActionNode*
ActionNode::EmptyMatchCheck(int start_register,
                            int repetition_register,
                            int repetition_limit,
                            RegExpNode* on_success)
{
    ActionNode* result = on_success->alloc()->newInfallible<ActionNode>(EMPTY_MATCH_CHECK, on_success);
    result->data_.u_empty_match_check.start_register = start_register;
    result->data_.u_empty_match_check.repetition_register = repetition_register;
    result->data_.u_empty_match_check.repetition_limit = repetition_limit;
    return result;
}

// -------------------------------------------------------------------
// TextNode

int
TextNode::EatsAtLeast(int still_to_find, int budget, bool not_at_start)
{
    if (read_backward())
        return 0;
    int answer = Length();
    if (answer >= still_to_find)
        return answer;
    if (budget <= 0)
        return answer;

    // We are not at start after this node so we set the last argument to 'true'.
    return answer + on_success()->EatsAtLeast(still_to_find - answer,
                                              budget - 1,
                                              true);
}

int
TextNode::GreedyLoopTextLength()
{
    return Length();
}

RegExpNode*
TextNode::FilterASCII(int depth, bool ignore_case, bool unicode)
{
    if (info()->replacement_calculated)
        return replacement();

    if (depth < 0)
        return this;

    MOZ_ASSERT(!info()->visited);
    VisitMarker marker(info());
    int element_count = elements().length();
    for (int i = 0; i < element_count; i++) {
        TextElement elm = elements()[i];
        if (elm.text_type() == TextElement::ATOM) {
            CharacterVector& quarks = const_cast<CharacterVector&>(elm.atom()->data());
            for (size_t j = 0; j < quarks.length(); j++) {
                uint16_t c = quarks[j];
                if (c <= kMaxOneByteCharCode)
                    continue;
                if (!ignore_case)
                    return set_replacement(nullptr);

                // Here, we need to check for characters whose upper and lower cases
                // are outside the Latin-1 range.
                char16_t converted = ConvertNonLatin1ToLatin1(c, unicode);
                if (converted == 0) {
                    // Character is outside Latin-1 completely
                    return set_replacement(nullptr);
                }

                // Convert quark to Latin-1 in place.
                quarks[j] = converted;
            }
        } else {
            MOZ_ASSERT(elm.text_type() == TextElement::CHAR_CLASS);
            RegExpCharacterClass* cc = elm.char_class();

            CharacterRangeVector& ranges = cc->ranges(alloc());
            if (!CharacterRange::IsCanonical(ranges))
                CharacterRange::Canonicalize(ranges);

            // Now they are in order so we only need to look at the first.
            int range_count = ranges.length();
            if (cc->is_negated()) {
                if (range_count != 0 &&
                    ranges[0].from() == 0 &&
                    ranges[0].to() >= kMaxOneByteCharCode)
                {
                    // This will be handled in a later filter.
                    if (ignore_case && CharacterRange::RangesContainLatin1Equivalents(ranges, unicode))
                        continue;
                    return set_replacement(nullptr);
                }
            } else {
                if (range_count == 0 ||
                    ranges[0].from() > kMaxOneByteCharCode)
                {
                    // This will be handled in a later filter.
                    if (ignore_case && CharacterRange::RangesContainLatin1Equivalents(ranges, unicode))
                        continue;
                    return set_replacement(nullptr);
                }
            }
        }
    }
    return FilterSuccessor(depth - 1, ignore_case, unicode);
}

void
TextNode::CalculateOffsets()
{
    int element_count = elements().length();

    // Set up the offsets of the elements relative to the start.  This is a fixed
    // quantity since a TextNode can only contain fixed-width things.
    int cp_offset = 0;
    for (int i = 0; i < element_count; i++) {
        TextElement& elm = elements()[i];
        elm.set_cp_offset(cp_offset);
        cp_offset += elm.length();
    }
}

void TextNode::MakeCaseIndependent(bool is_ascii, bool unicode)
{
    int element_count = elements().length();
    for (int i = 0; i < element_count; i++) {
        TextElement elm = elements()[i];
        if (elm.text_type() == TextElement::CHAR_CLASS) {
            RegExpCharacterClass* cc = elm.char_class();

            // None of the standard character classes is different in the case
            // independent case and it slows us down if we don't know that.
            if (cc->is_standard(alloc()))
                continue;

            CharacterRangeVector& ranges = cc->ranges(alloc());
            int range_count = ranges.length();
            for (int j = 0; j < range_count; j++)
                ranges[j].AddCaseEquivalents(is_ascii, unicode, &ranges);
        }
    }
}

// -------------------------------------------------------------------
// AssertionNode

int
AssertionNode::EatsAtLeast(int still_to_find, int budget, bool not_at_start)
{
    if (budget <= 0)
        return 0;

    // If we know we are not at the start and we are asked "how many characters
    // will you match if you succeed?" then we can answer anything since false
    // implies false.  So lets just return the max answer (still_to_find) since
    // that won't prevent us from preloading a lot of characters for the other
    // branches in the node graph.
    if (assertion_type() == AT_START && not_at_start)
        return still_to_find;

    return on_success()->EatsAtLeast(still_to_find, budget - 1, not_at_start);
}

bool
AssertionNode::FillInBMInfo(int offset, int budget, BoyerMooreLookahead* bm, bool not_at_start)
{
    if (!bm->CheckOverRecursed())
        return false;

    // Match the behaviour of EatsAtLeast on this node.
    if (assertion_type() == AT_START && not_at_start)
        return true;

    if (!on_success()->FillInBMInfo(offset, budget - 1, bm, not_at_start))
        return false;
    SaveBMInfo(bm, not_at_start, offset);
    return true;
}

// -------------------------------------------------------------------
// BackReferenceNode

int
BackReferenceNode::EatsAtLeast(int still_to_find, int budget, bool not_at_start)
{
    if (read_backward())
        return 0;
    if (budget <= 0)
        return 0;
    return on_success()->EatsAtLeast(still_to_find, budget - 1, not_at_start);
}

bool
BackReferenceNode::FillInBMInfo(int offset, int budget, BoyerMooreLookahead* bm, bool not_at_start)
{
    // Working out the set of characters that a backreference can match is too
    // hard, so we just say that any character can match.
    bm->SetRest(offset);
    SaveBMInfo(bm, not_at_start, offset);
    return true;
}

// -------------------------------------------------------------------
// ChoiceNode

int
ChoiceNode::EatsAtLeastHelper(int still_to_find,
                              int budget,
                              RegExpNode* ignore_this_node,
                              bool not_at_start)
{
    if (budget <= 0)
        return 0;

    int min = 100;
    size_t choice_count = alternatives().length();
    budget = (budget - 1) / choice_count;
    for (size_t i = 0; i < choice_count; i++) {
        RegExpNode* node = alternatives()[i].node();
        if (node == ignore_this_node) continue;
        int node_eats_at_least =
            node->EatsAtLeast(still_to_find, budget, not_at_start);
        if (node_eats_at_least < min)
            min = node_eats_at_least;
        if (min == 0)
            return 0;
    }
    return min;
}

int
ChoiceNode::EatsAtLeast(int still_to_find, int budget, bool not_at_start)
{
    return EatsAtLeastHelper(still_to_find,
                             budget,
                             nullptr,
                             not_at_start);
}

void
ChoiceNode::GetQuickCheckDetails(QuickCheckDetails* details,
                                 RegExpCompiler* compiler,
                                 int characters_filled_in,
                                 bool not_at_start)
{
    not_at_start = (not_at_start || not_at_start_);
    int choice_count = alternatives().length();
    MOZ_ASSERT(choice_count > 0);
    alternatives()[0].node()->GetQuickCheckDetails(details,
                                                   compiler,
                                                   characters_filled_in,
                                                   not_at_start);
    for (int i = 1; i < choice_count; i++) {
        QuickCheckDetails new_details(details->characters());
        RegExpNode* node = alternatives()[i].node();
        node->GetQuickCheckDetails(&new_details, compiler,
                                   characters_filled_in,
                                   not_at_start);
        // Here we merge the quick match details of the two branches.
        details->Merge(&new_details, characters_filled_in);
    }
}

bool
ChoiceNode::FillInBMInfo(int offset,
                         int budget,
                         BoyerMooreLookahead* bm,
                         bool not_at_start)
{
    if (!bm->CheckOverRecursed())
        return false;

    const GuardedAlternativeVector& alts = alternatives();
    budget = (budget - 1) / alts.length();
    for (size_t i = 0; i < alts.length(); i++) {
        const GuardedAlternative& alt = alts[i];
        if (alt.guards() != nullptr && alt.guards()->length() != 0) {
            bm->SetRest(offset);  // Give up trying to fill in info.
            SaveBMInfo(bm, not_at_start, offset);
            return true;
        }
        if (!alt.node()->FillInBMInfo(offset, budget, bm, not_at_start))
            return false;
    }
    SaveBMInfo(bm, not_at_start, offset);
    return true;
}

RegExpNode*
ChoiceNode::FilterASCII(int depth, bool ignore_case, bool unicode)
{
    if (info()->replacement_calculated)
        return replacement();
    if (depth < 0)
        return this;
    if (info()->visited)
        return this;
    VisitMarker marker(info());
    int choice_count = alternatives().length();

    for (int i = 0; i < choice_count; i++) {
        const GuardedAlternative alternative = alternatives()[i];
        if (alternative.guards() != nullptr && alternative.guards()->length() != 0) {
            set_replacement(this);
            return this;
        }
    }

    int surviving = 0;
    RegExpNode* survivor = nullptr;
    for (int i = 0; i < choice_count; i++) {
        GuardedAlternative alternative = alternatives()[i];
        RegExpNode* replacement =
            alternative.node()->FilterASCII(depth - 1, ignore_case, unicode);
        MOZ_ASSERT(replacement != this);  // No missing EMPTY_MATCH_CHECK.
        if (replacement != nullptr) {
            alternatives()[i].set_node(replacement);
            surviving++;
            survivor = replacement;
        }
    }
    if (surviving < 2)
        return set_replacement(survivor);

    set_replacement(this);
    if (surviving == choice_count)
        return this;

    // Only some of the nodes survived the filtering.  We need to rebuild the
    // alternatives list.
    GuardedAlternativeVector new_alternatives(*alloc());
    new_alternatives.reserve(surviving);
    for (int i = 0; i < choice_count; i++) {
        RegExpNode* replacement =
            alternatives()[i].node()->FilterASCII(depth - 1, ignore_case, unicode);
        if (replacement != nullptr) {
            alternatives()[i].set_node(replacement);
            new_alternatives.append(alternatives()[i]);
        }
    }

    alternatives_ = Move(new_alternatives);
    return this;
}

// -------------------------------------------------------------------
// NegativeLookaheadChoiceNode

bool
NegativeLookaheadChoiceNode::FillInBMInfo(int offset,
                                          int budget,
                                          BoyerMooreLookahead* bm,
                                          bool not_at_start)
{
    if (!bm->CheckOverRecursed())
        return false;

    if (!alternatives()[1].node()->FillInBMInfo(offset, budget - 1, bm, not_at_start))
        return false;
    if (offset == 0)
        set_bm_info(not_at_start, bm);
    return true;
}

int
NegativeLookaheadChoiceNode::EatsAtLeast(int still_to_find, int budget, bool not_at_start)
{
    if (budget <= 0)
        return 0;

    // Alternative 0 is the negative lookahead, alternative 1 is what comes
    // afterwards.
    RegExpNode* node = alternatives()[1].node();
    return node->EatsAtLeast(still_to_find, budget - 1, not_at_start);
}

void
NegativeLookaheadChoiceNode::GetQuickCheckDetails(QuickCheckDetails* details,
                                                  RegExpCompiler* compiler,
                                                  int filled_in,
                                                  bool not_at_start)
{
    // Alternative 0 is the negative lookahead, alternative 1 is what comes
    // afterwards.
    RegExpNode* node = alternatives()[1].node();
    return node->GetQuickCheckDetails(details, compiler, filled_in, not_at_start);
}

RegExpNode*
NegativeLookaheadChoiceNode::FilterASCII(int depth, bool ignore_case, bool unicode)
{
    if (info()->replacement_calculated)
        return replacement();
    if (depth < 0)
        return this;
    if (info()->visited)
        return this;

    VisitMarker marker(info());

    // Alternative 0 is the negative lookahead, alternative 1 is what comes
    // afterwards.
    RegExpNode* node = alternatives()[1].node();
    RegExpNode* replacement = node->FilterASCII(depth - 1, ignore_case, unicode);

    if (replacement == nullptr)
        return set_replacement(nullptr);
    alternatives()[1].set_node(replacement);

    RegExpNode* neg_node = alternatives()[0].node();
    RegExpNode* neg_replacement = neg_node->FilterASCII(depth - 1, ignore_case, unicode);

    // If the negative lookahead is always going to fail then
    // we don't need to check it.
    if (neg_replacement == nullptr)
        return set_replacement(replacement);

    alternatives()[0].set_node(neg_replacement);
    return set_replacement(this);
}

// -------------------------------------------------------------------
// LoopChoiceNode

void
GuardedAlternative::AddGuard(LifoAlloc* alloc, Guard* guard)
{
    if (guards_ == nullptr)
        guards_ = alloc->newInfallible<GuardVector>(*alloc);
    guards_->append(guard);
}

void
LoopChoiceNode::AddLoopAlternative(GuardedAlternative alt)
{
    MOZ_ASSERT(loop_node_ == nullptr);
    AddAlternative(alt);
    loop_node_ = alt.node();
}


void
LoopChoiceNode::AddContinueAlternative(GuardedAlternative alt)
{
    MOZ_ASSERT(continue_node_ == nullptr);
    AddAlternative(alt);
    continue_node_ = alt.node();
}

int
LoopChoiceNode::EatsAtLeast(int still_to_find,  int budget, bool not_at_start)
{
    return EatsAtLeastHelper(still_to_find,
                             budget - 1,
                             loop_node_,
                             not_at_start);
}

void
LoopChoiceNode::GetQuickCheckDetails(QuickCheckDetails* details,
                                     RegExpCompiler* compiler,
                                     int characters_filled_in,
                                     bool not_at_start)
{
    if (body_can_be_zero_length_ || info()->visited)
        return;
    VisitMarker marker(info());
    return ChoiceNode::GetQuickCheckDetails(details,
                                            compiler,
                                            characters_filled_in,
                                            not_at_start);
}

bool
LoopChoiceNode::FillInBMInfo(int offset,
                             int budget,
                             BoyerMooreLookahead* bm,
                             bool not_at_start)
{
    if (body_can_be_zero_length_ || budget <= 0) {
        bm->SetRest(offset);
        SaveBMInfo(bm, not_at_start, offset);
        return true;
    }
    if (!ChoiceNode::FillInBMInfo(offset, budget - 1, bm, not_at_start))
        return false;
    SaveBMInfo(bm, not_at_start, offset);
    return true;
}

RegExpNode*
LoopChoiceNode::FilterASCII(int depth, bool ignore_case, bool unicode)
{
    if (info()->replacement_calculated)
        return replacement();
    if (depth < 0)
        return this;
    if (info()->visited)
        return this;

    {
        VisitMarker marker(info());

        RegExpNode* continue_replacement =
            continue_node_->FilterASCII(depth - 1, ignore_case, unicode);

        // If we can't continue after the loop then there is no sense in doing the
        // loop.
        if (continue_replacement == nullptr)
            return set_replacement(nullptr);
    }

    return ChoiceNode::FilterASCII(depth - 1, ignore_case, unicode);
}

// -------------------------------------------------------------------
// Analysis

void
Analysis::EnsureAnalyzed(RegExpNode* that)
{
    JS_CHECK_RECURSION(cx, failASCII("Stack overflow"); return);

    if (that->info()->been_analyzed || that->info()->being_analyzed)
        return;
    that->info()->being_analyzed = true;
    that->Accept(this);
    that->info()->being_analyzed = false;
    that->info()->been_analyzed = true;
}

void
Analysis::VisitEnd(EndNode* that)
{
    // nothing to do
}

void
Analysis::VisitText(TextNode* that)
{
    if (ignore_case_)
        that->MakeCaseIndependent(is_ascii_, unicode_);
    EnsureAnalyzed(that->on_success());
    if (!has_failed()) {
        that->CalculateOffsets();
    }
}

void
Analysis::VisitAction(ActionNode* that)
{
    RegExpNode* target = that->on_success();
    EnsureAnalyzed(target);

    if (!has_failed()) {
        // If the next node is interested in what it follows then this node
        // has to be interested too so it can pass the information on.
        that->info()->AddFromFollowing(target->info());
    }
}

void
Analysis::VisitChoice(ChoiceNode* that)
{
    NodeInfo* info = that->info();

    for (size_t i = 0; i < that->alternatives().length(); i++) {
        RegExpNode* node = that->alternatives()[i].node();
        EnsureAnalyzed(node);
        if (has_failed()) return;

        // Anything the following nodes need to know has to be known by
        // this node also, so it can pass it on.
        info->AddFromFollowing(node->info());
    }
}

void
Analysis::VisitLoopChoice(LoopChoiceNode* that)
{
    NodeInfo* info = that->info();
    for (size_t i = 0; i < that->alternatives().length(); i++) {
        RegExpNode* node = that->alternatives()[i].node();
        if (node != that->loop_node()) {
            EnsureAnalyzed(node);
            if (has_failed()) return;
            info->AddFromFollowing(node->info());
        }
    }

    // Check the loop last since it may need the value of this node
    // to get a correct result.
    EnsureAnalyzed(that->loop_node());
    if (!has_failed())
        info->AddFromFollowing(that->loop_node()->info());
}

void
Analysis::VisitBackReference(BackReferenceNode* that)
{
    EnsureAnalyzed(that->on_success());
}

void
Analysis::VisitAssertion(AssertionNode* that)
{
    EnsureAnalyzed(that->on_success());
}

// -------------------------------------------------------------------
// Implementation of the Irregexp regular expression engine.
//
// The Irregexp regular expression engine is intended to be a complete
// implementation of ECMAScript regular expressions.  It generates either
// bytecodes or native code.

//   The Irregexp regexp engine is structured in three steps.
//   1) The parser generates an abstract syntax tree.  See RegExpAST.cpp.
//   2) From the AST a node network is created.  The nodes are all
//      subclasses of RegExpNode.  The nodes represent states when
//      executing a regular expression.  Several optimizations are
//      performed on the node network.
//   3) From the nodes we generate either byte codes or native code
//      that can actually execute the regular expression (perform
//      the search).  The code generation step is described in more
//      detail below.

// Code generation.
//
//   The nodes are divided into four main categories.
//   * Choice nodes
//        These represent places where the regular expression can
//        match in more than one way.  For example on entry to an
//        alternation (foo|bar) or a repetition (*, +, ? or {}).
//   * Action nodes
//        These represent places where some action should be
//        performed.  Examples include recording the current position
//        in the input string to a register (in order to implement
//        captures) or other actions on register for example in order
//        to implement the counters needed for {} repetitions.
//   * Matching nodes
//        These attempt to match some element part of the input string.
//        Examples of elements include character classes, plain strings
//        or back references.
//   * End nodes
//        These are used to implement the actions required on finding
//        a successful match or failing to find a match.
//
//   The code generated (whether as byte codes or native code) maintains
//   some state as it runs.  This consists of the following elements:
//
//   * The capture registers.  Used for string captures.
//   * Other registers.  Used for counters etc.
//   * The current position.
//   * The stack of backtracking information.  Used when a matching node
//     fails to find a match and needs to try an alternative.
//
// Conceptual regular expression execution model:
//
//   There is a simple conceptual model of regular expression execution
//   which will be presented first.  The actual code generated is a more
//   efficient simulation of the simple conceptual model:
//
//   * Choice nodes are implemented as follows:
//     For each choice except the last {
//       push current position
//       push backtrack code location
//       <generate code to test for choice>
//       backtrack code location:
//       pop current position
//     }
//     <generate code to test for last choice>
//
//   * Actions nodes are generated as follows
//     <push affected registers on backtrack stack>
//     <generate code to perform action>
//     push backtrack code location
//     <generate code to test for following nodes>
//     backtrack code location:
//     <pop affected registers to restore their state>
//     <pop backtrack location from stack and go to it>
//
//   * Matching nodes are generated as follows:
//     if input string matches at current position
//       update current position
//       <generate code to test for following nodes>
//     else
//       <pop backtrack location from stack and go to it>
//
//   Thus it can be seen that the current position is saved and restored
//   by the choice nodes, whereas the registers are saved and restored by
//   by the action nodes that manipulate them.
//
//   The other interesting aspect of this model is that nodes are generated
//   at the point where they are needed by a recursive call to Emit().  If
//   the node has already been code generated then the Emit() call will
//   generate a jump to the previously generated code instead.  In order to
//   limit recursion it is possible for the Emit() function to put the node
//   on a work list for later generation and instead generate a jump.  The
//   destination of the jump is resolved later when the code is generated.
//
// Actual regular expression code generation.
//
//   Code generation is actually more complicated than the above.  In order
//   to improve the efficiency of the generated code some optimizations are
//   performed
//
//   * Choice nodes have 1-character lookahead.
//     A choice node looks at the following character and eliminates some of
//     the choices immediately based on that character.  This is not yet
//     implemented.
//   * Simple greedy loops store reduced backtracking information.
//     A quantifier like /.*foo/m will greedily match the whole input.  It will
//     then need to backtrack to a point where it can match "foo".  The naive
//     implementation of this would push each character position onto the
//     backtracking stack, then pop them off one by one.  This would use space
//     proportional to the length of the input string.  However since the "."
//     can only match in one way and always has a constant length (in this case
//     of 1) it suffices to store the current position on the top of the stack
//     once.  Matching now becomes merely incrementing the current position and
//     backtracking becomes decrementing the current position and checking the
//     result against the stored current position.  This is faster and saves
//     space.
//   * The current state is virtualized.
//     This is used to defer expensive operations until it is clear that they
//     are needed and to generate code for a node more than once, allowing
//     specialized an efficient versions of the code to be created. This is
//     explained in the section below.
//
// Execution state virtualization.
//
//   Instead of emitting code, nodes that manipulate the state can record their
//   manipulation in an object called the Trace.  The Trace object can record a
//   current position offset, an optional backtrack code location on the top of
//   the virtualized backtrack stack and some register changes.  When a node is
//   to be emitted it can flush the Trace or update it.  Flushing the Trace
//   will emit code to bring the actual state into line with the virtual state.
//   Avoiding flushing the state can postpone some work (e.g. updates of capture
//   registers).  Postponing work can save time when executing the regular
//   expression since it may be found that the work never has to be done as a
//   failure to match can occur.  In addition it is much faster to jump to a
//   known backtrack code location than it is to pop an unknown backtrack
//   location from the stack and jump there.
//
//   The virtual state found in the Trace affects code generation.  For example
//   the virtual state contains the difference between the actual current
//   position and the virtual current position, and matching code needs to use
//   this offset to attempt a match in the correct location of the input
//   string.  Therefore code generated for a non-trivial trace is specialized
//   to that trace.  The code generator therefore has the ability to generate
//   code for each node several times.  In order to limit the size of the
//   generated code there is an arbitrary limit on how many specialized sets of
//   code may be generated for a given node.  If the limit is reached, the
//   trace is flushed and a generic version of the code for a node is emitted.
//   This is subsequently used for that node.  The code emitted for non-generic
//   trace is not recorded in the node and so it cannot currently be reused in
//   the event that code generation is requested for an identical trace.

/* static */ TextElement
TextElement::Atom(RegExpAtom* atom)
{
    return TextElement(ATOM, atom);
}

/* static */ TextElement
TextElement::CharClass(RegExpCharacterClass* char_class)
{
    return TextElement(CHAR_CLASS, char_class);
}

int
TextElement::length() const
{
    switch (text_type()) {
      case ATOM:
        return atom()->length();
      case CHAR_CLASS:
        return 1;
    }
    MOZ_CRASH("Bad text type");
}

class FrequencyCollator
{
  public:
    FrequencyCollator() : total_samples_(0) {
        for (int i = 0; i < RegExpMacroAssembler::kTableSize; i++) {
            frequencies_[i] = CharacterFrequency(i);
        }
    }

    void CountCharacter(int character) {
        int index = (character & RegExpMacroAssembler::kTableMask);
        frequencies_[index].Increment();
        total_samples_++;
    }

    // Does not measure in percent, but rather per-128 (the table size from the
    // regexp macro assembler).
    int Frequency(int in_character) {
        MOZ_ASSERT((in_character & RegExpMacroAssembler::kTableMask) == in_character);
        if (total_samples_ < 1) return 1;  // Division by zero.
        int freq_in_per128 =
            (frequencies_[in_character].counter() * 128) / total_samples_;
        return freq_in_per128;
    }

  private:
    class CharacterFrequency {
      public:
        CharacterFrequency() : counter_(0), character_(-1) { }
        explicit CharacterFrequency(int character)
          : counter_(0), character_(character)
        {}

        void Increment() { counter_++; }
        int counter() { return counter_; }
        int character() { return character_; }

     private:
        int counter_;
        int character_;
    };

  private:
    CharacterFrequency frequencies_[RegExpMacroAssembler::kTableSize];
    int total_samples_;
};

class irregexp::RegExpCompiler
{
  public:
    RegExpCompiler(JSContext* cx, LifoAlloc* alloc, int capture_count,
                   bool ignore_case, bool is_ascii, bool match_only, bool unicode);

    int AllocateRegister() {
        if (next_register_ >= RegExpMacroAssembler::kMaxRegister) {
            reg_exp_too_big_ = true;
            return next_register_;
        }
        return next_register_++;
    }

    RegExpCode Assemble(JSContext* cx,
                        RegExpMacroAssembler* assembler,
                        RegExpNode* start,
                        int capture_count);

    inline void AddWork(RegExpNode* node) {
        AutoEnterOOMUnsafeRegion oomUnsafe;
        if (!work_list_.append(node))
            oomUnsafe.crash("AddWork");
    }

    static const int kImplementationOffset = 0;
    static const int kNumberOfRegistersOffset = 0;
    static const int kCodeOffset = 1;

    RegExpMacroAssembler* macro_assembler() { return macro_assembler_; }
    EndNode* accept() { return accept_; }

    static const int kMaxRecursion = 100;
    inline int recursion_depth() { return recursion_depth_; }
    inline void IncrementRecursionDepth() { recursion_depth_++; }
    inline void DecrementRecursionDepth() { recursion_depth_--; }

    void SetRegExpTooBig() { reg_exp_too_big_ = true; }

    inline bool ignore_case() { return ignore_case_; }
    inline bool ascii() { return ascii_; }
    inline bool unicode() { return unicode_; }
    FrequencyCollator* frequency_collator() { return &frequency_collator_; }

    int current_expansion_factor() { return current_expansion_factor_; }
    void set_current_expansion_factor(int value) {
        current_expansion_factor_ = value;
    }

    bool read_backward() { return read_backward_; }
    void set_read_backward(bool value) { read_backward_ = value; }

    JSContext* cx() const { return cx_; }
    LifoAlloc* alloc() const { return alloc_; }

    static const int kNoRegister = -1;

  private:
    EndNode* accept_;
    int next_register_;
    Vector<RegExpNode*, 4, SystemAllocPolicy> work_list_;
    int recursion_depth_;
    RegExpMacroAssembler* macro_assembler_;
    bool ignore_case_;
    bool ascii_;
    bool match_only_;
    bool unicode_;
    bool reg_exp_too_big_;
    int current_expansion_factor_;
    bool read_backward_;
    FrequencyCollator frequency_collator_;
    JSContext* cx_;
    LifoAlloc* alloc_;
};

class RecursionCheck
{
  public:
    explicit RecursionCheck(RegExpCompiler* compiler) : compiler_(compiler) {
        compiler->IncrementRecursionDepth();
    }
    ~RecursionCheck() { compiler_->DecrementRecursionDepth(); }

  private:
    RegExpCompiler* compiler_;
};

// Attempts to compile the regexp using an Irregexp code generator.  Returns
// a fixed array or a null handle depending on whether it succeeded.
RegExpCompiler::RegExpCompiler(JSContext* cx, LifoAlloc* alloc, int capture_count,
                               bool ignore_case, bool ascii, bool match_only, bool unicode)
  : next_register_(2 * (capture_count + 1)),
    recursion_depth_(0),
    ignore_case_(ignore_case),
    ascii_(ascii),
    match_only_(match_only),
    unicode_(unicode),
    reg_exp_too_big_(false),
    current_expansion_factor_(1),
    read_backward_(false),
    frequency_collator_(),
    cx_(cx),
    alloc_(alloc)
{
    accept_ = alloc->newInfallible<EndNode>(alloc, EndNode::ACCEPT);
    MOZ_ASSERT(next_register_ - 1 <= RegExpMacroAssembler::kMaxRegister);
}

RegExpCode
RegExpCompiler::Assemble(JSContext* cx,
                         RegExpMacroAssembler* assembler,
                         RegExpNode* start,
                         int capture_count)
{
    macro_assembler_ = assembler;
    macro_assembler_->set_slow_safe(false);

    // The LifoAlloc used by the regexp compiler is infallible and is currently
    // expected to crash on OOM. Thus we have to disable the assertions made to
    // prevent us from allocating any new chunk in the LifoAlloc. This is needed
    // because the jit::MacroAssembler turns these assertions on by default.
    LifoAlloc::AutoFallibleScope fallibleAllocator(alloc());

    jit::Label fail;
    macro_assembler_->PushBacktrack(&fail);
    Trace new_trace;
    start->Emit(this, &new_trace);
    macro_assembler_->BindBacktrack(&fail);
    macro_assembler_->Fail();

    while (!work_list_.empty())
        work_list_.popCopy()->Emit(this, &new_trace);

    RegExpCode code = macro_assembler_->GenerateCode(cx, match_only_);
    if (code.empty())
        return RegExpCode();

    if (reg_exp_too_big_) {
        code.destroy();
        JS_ReportErrorASCII(cx, "regexp too big");
        return RegExpCode();
    }

    return code;
}

template <typename CharT>
static void
SampleChars(FrequencyCollator* collator, const CharT* chars, size_t length)
{
    // Sample some characters from the middle of the string.
    static const int kSampleSize = 128;

    int chars_sampled = 0;
    int half_way = (int(length) - kSampleSize) / 2;
    for (size_t i = Max(0, half_way);
         i < length && chars_sampled < kSampleSize;
         i++, chars_sampled++)
    {
        collator->CountCharacter(chars[i]);
    }
}

static bool
IsNativeRegExpEnabled(JSContext* cx)
{
#ifdef JS_CODEGEN_NONE
    return false;
#else
    return cx->options().nativeRegExp();
#endif
}

RegExpCode
irregexp::CompilePattern(JSContext* cx, HandleRegExpShared shared, RegExpCompileData* data,
                         HandleLinearString sample, bool is_global, bool ignore_case,
                         bool is_ascii, bool match_only, bool force_bytecode, bool sticky,
                         bool unicode, RegExpShared::JitCodeTables& tables)
{
    if ((data->capture_count + 1) * 2 - 1 > RegExpMacroAssembler::kMaxRegister) {
        JS_ReportErrorASCII(cx, "regexp too big");
        return RegExpCode();
    }

    LifoAlloc& alloc = cx->tempLifoAlloc();
    RegExpCompiler compiler(cx, &alloc, data->capture_count, ignore_case, is_ascii, match_only,
                            unicode);

    // Sample some characters from the middle of the string.
    if (sample->hasLatin1Chars()) {
        JS::AutoCheckCannotGC nogc;
        SampleChars(compiler.frequency_collator(), sample->latin1Chars(nogc), sample->length());
    } else {
        JS::AutoCheckCannotGC nogc;
        SampleChars(compiler.frequency_collator(), sample->twoByteChars(nogc), sample->length());
    }

    // Wrap the body of the regexp in capture #0.
    RegExpNode* captured_body = RegExpCapture::ToNode(data->tree,
                                                      0,
                                                      &compiler,
                                                      compiler.accept());
    RegExpNode* node = captured_body;
    bool is_end_anchored = data->tree->IsAnchoredAtEnd();
    bool is_start_anchored = sticky || data->tree->IsAnchoredAtStart();
    int max_length = data->tree->max_match();
    if (!is_start_anchored) {
        // Add a .*? at the beginning, outside the body capture, unless
        // this expression is anchored at the beginning.
        RegExpNode* loop_node =
            RegExpQuantifier::ToNode(0,
                                     RegExpTree::kInfinity,
                                     false,
                                     alloc.newInfallible<RegExpCharacterClass>('*'),
                                     &compiler,
                                     captured_body,
                                     data->contains_anchor);

        if (data->contains_anchor) {
            // Unroll loop once, to take care of the case that might start
            // at the start of input.
            ChoiceNode* first_step_node = alloc.newInfallible<ChoiceNode>(&alloc, 2);
            RegExpNode* char_class =
                alloc.newInfallible<TextNode>(alloc.newInfallible<RegExpCharacterClass>('*'), false, loop_node);
            first_step_node->AddAlternative(GuardedAlternative(captured_body));
            first_step_node->AddAlternative(GuardedAlternative(char_class));
            node = first_step_node;
        } else {
            node = loop_node;
        }
    }
    if (is_ascii) {
        node = node->FilterASCII(RegExpCompiler::kMaxRecursion, ignore_case, unicode);
        // Do it again to propagate the new nodes to places where they were not
        // put because they had not been calculated yet.
        if (node != nullptr) {
            node = node->FilterASCII(RegExpCompiler::kMaxRecursion, ignore_case, unicode);
        }
    }

    if (node == nullptr)
        node = alloc.newInfallible<EndNode>(&alloc, EndNode::BACKTRACK);

    Analysis analysis(cx, ignore_case, is_ascii, unicode);
    analysis.EnsureAnalyzed(node);
    if (analysis.has_failed()) {
        JS_ReportErrorASCII(cx, "%s", analysis.errorMessage());
        return RegExpCode();
    }

    Maybe<jit::JitContext> ctx;
    Maybe<NativeRegExpMacroAssembler> native_assembler;
    Maybe<InterpretedRegExpMacroAssembler> interpreted_assembler;

    RegExpMacroAssembler* assembler;
    if (IsNativeRegExpEnabled(cx) &&
        !force_bytecode &&
        jit::CanLikelyAllocateMoreExecutableMemory() &&
        shared->getSource()->length() < 32 * 1024)
    {
        NativeRegExpMacroAssembler::Mode mode =
            is_ascii ? NativeRegExpMacroAssembler::ASCII
                     : NativeRegExpMacroAssembler::CHAR16;

        ctx.emplace(cx, (jit::TempAllocator*) nullptr);
        native_assembler.emplace(&alloc, cx->runtime(), mode, (data->capture_count + 1) * 2, tables);
        assembler = native_assembler.ptr();
    } else {
        interpreted_assembler.emplace(&alloc, (data->capture_count + 1) * 2);
        assembler = interpreted_assembler.ptr();
    }

    // Inserted here, instead of in Assembler, because it depends on information
    // in the AST that isn't replicated in the Node structure.
    static const int kMaxBacksearchLimit = 1024;
    if (is_end_anchored &&
        !is_start_anchored &&
        max_length < kMaxBacksearchLimit) {
        assembler->SetCurrentPositionFromEnd(max_length);
    }

    if (is_global) {
        assembler->set_global_mode((data->tree->min_match() > 0)
                                   ? RegExpMacroAssembler::GLOBAL_NO_ZERO_LENGTH_CHECK
                                   : RegExpMacroAssembler::GLOBAL);
    }

    return compiler.Assemble(cx, assembler, node, data->capture_count);
}

template <typename CharT>
RegExpRunStatus
irregexp::ExecuteCode(JSContext* cx, jit::JitCode* codeBlock, const CharT* chars, size_t start,
                      size_t length, MatchPairs* matches, size_t* endIndex)
{
    typedef void (*RegExpCodeSignature)(InputOutputData*);

    InputOutputData data(chars, chars + length, start, matches, endIndex);

    RegExpCodeSignature function = reinterpret_cast<RegExpCodeSignature>(codeBlock->raw());

    {
        JS::AutoSuppressGCAnalysis nogc;
        CALL_GENERATED_1(function, &data);
    }

    return (RegExpRunStatus) data.result;
}

template RegExpRunStatus
irregexp::ExecuteCode(JSContext* cx, jit::JitCode* codeBlock, const Latin1Char* chars, size_t start,
                      size_t length, MatchPairs* matches, size_t* endIndex);

template RegExpRunStatus
irregexp::ExecuteCode(JSContext* cx, jit::JitCode* codeBlock, const char16_t* chars, size_t start,
                      size_t length, MatchPairs* matches, size_t* endIndex);

// -------------------------------------------------------------------
// Tree to graph conversion

RegExpNode*
RegExpAtom::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
{
    TextElementVector* elms =
        compiler->alloc()->newInfallible<TextElementVector>(*compiler->alloc());
    elms->append(TextElement::Atom(this));
    return compiler->alloc()->newInfallible<TextNode>(elms, compiler->read_backward(), on_success);
}

RegExpNode*
RegExpText::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
{
    return compiler->alloc()->newInfallible<TextNode>(&elements_, compiler->read_backward(), on_success);
}

RegExpNode*
RegExpCharacterClass::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
{
    return compiler->alloc()->newInfallible<TextNode>(this, compiler->read_backward(), on_success);
}

RegExpNode*
RegExpDisjunction::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
{
    const RegExpTreeVector& alternatives = this->alternatives();
    size_t length = alternatives.length();
    ChoiceNode* result = compiler->alloc()->newInfallible<ChoiceNode>(compiler->alloc(), length);
    for (size_t i = 0; i < length; i++) {
        GuardedAlternative alternative(alternatives[i]->ToNode(compiler, on_success));
        result->AddAlternative(alternative);
    }
    return result;
}

RegExpNode*
RegExpQuantifier::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
{
    return ToNode(min(),
                  max(),
                  is_greedy(),
                  body(),
                  compiler,
                  on_success);
}

// Scoped object to keep track of how much we unroll quantifier loops in the
// regexp graph generator.
class RegExpExpansionLimiter
{
  public:
    static const int kMaxExpansionFactor = 6;
    RegExpExpansionLimiter(RegExpCompiler* compiler, int factor)
      : compiler_(compiler),
        saved_expansion_factor_(compiler->current_expansion_factor()),
        ok_to_expand_(saved_expansion_factor_ <= kMaxExpansionFactor)
    {
        MOZ_ASSERT(factor > 0);
        if (ok_to_expand_) {
            if (factor > kMaxExpansionFactor) {
                // Avoid integer overflow of the current expansion factor.
                ok_to_expand_ = false;
                compiler->set_current_expansion_factor(kMaxExpansionFactor + 1);
            } else {
                int new_factor = saved_expansion_factor_ * factor;
                ok_to_expand_ = (new_factor <= kMaxExpansionFactor);
                compiler->set_current_expansion_factor(new_factor);
            }
        }
    }

    ~RegExpExpansionLimiter() {
        compiler_->set_current_expansion_factor(saved_expansion_factor_);
    }

    bool ok_to_expand() { return ok_to_expand_; }

  private:
    RegExpCompiler* compiler_;
    int saved_expansion_factor_;
    bool ok_to_expand_;
};

/* static */ RegExpNode*
RegExpQuantifier::ToNode(int min,
                         int max,
                         bool is_greedy,
                         RegExpTree* body,
                         RegExpCompiler* compiler,
                         RegExpNode* on_success,
                         bool not_at_start /* = false */)
{
    // x{f, t} becomes this:
    //
    //             (r++)<-.
    //               |     `
    //               |     (x)
    //               v     ^
    //      (r=0)-->(?)---/ [if r < t]
    //               |
    //   [if r >= f] \----> ...
    //

    // 15.10.2.5 RepeatMatcher algorithm.
    // The parser has already eliminated the case where max is 0.  In the case
    // where max_match is zero the parser has removed the quantifier if min was
    // > 0 and removed the atom if min was 0.  See AddQuantifierToAtom.

    // If we know that we cannot match zero length then things are a little
    // simpler since we don't need to make the special zero length match check
    // from step 2.1.  If the min and max are small we can unroll a little in
    // this case.
    static const int kMaxUnrolledMinMatches = 3;  // Unroll (foo)+ and (foo){3,}
    static const int kMaxUnrolledMaxMatches = 3;  // Unroll (foo)? and (foo){x,3}

    if (max == 0)
        return on_success;  // This can happen due to recursion.

    bool body_can_be_empty = (body->min_match() == 0);
    int body_start_reg = RegExpCompiler::kNoRegister;
    Interval capture_registers = body->CaptureRegisters();
    bool needs_capture_clearing = !capture_registers.is_empty();
    LifoAlloc* alloc = compiler->alloc();

    if (body_can_be_empty) {
        body_start_reg = compiler->AllocateRegister();
    } else if (!needs_capture_clearing) {
        // Only unroll if there are no captures and the body can't be
        // empty.
        {
            RegExpExpansionLimiter limiter(compiler, min + ((max != min) ? 1 : 0));
            if (min > 0 && min <= kMaxUnrolledMinMatches && limiter.ok_to_expand()) {
                int new_max = (max == kInfinity) ? max : max - min;
                // Recurse once to get the loop or optional matches after the fixed
                // ones.
                RegExpNode* answer = ToNode(0, new_max, is_greedy, body, compiler, on_success, true);
                // Unroll the forced matches from 0 to min.  This can cause chains of
                // TextNodes (which the parser does not generate).  These should be
                // combined if it turns out they hinder good code generation.
                for (int i = 0; i < min; i++)
                    answer = body->ToNode(compiler, answer);
                return answer;
            }
        }
        if (max <= kMaxUnrolledMaxMatches && min == 0) {
            MOZ_ASSERT(max > 0);  // Due to the 'if' above.
            RegExpExpansionLimiter limiter(compiler, max);
            if (limiter.ok_to_expand()) {
                // Unroll the optional matches up to max.
                RegExpNode* answer = on_success;
                for (int i = 0; i < max; i++) {
                    ChoiceNode* alternation = alloc->newInfallible<ChoiceNode>(alloc, 2);
                    if (is_greedy) {
                        alternation->AddAlternative(GuardedAlternative(body->ToNode(compiler, answer)));
                        alternation->AddAlternative(GuardedAlternative(on_success));
                    } else {
                        alternation->AddAlternative(GuardedAlternative(on_success));
                        alternation->AddAlternative(GuardedAlternative(body->ToNode(compiler, answer)));
                    }
                    answer = alternation;
                    if (not_at_start && !compiler->read_backward()) {
                        alternation->set_not_at_start();
                    }
                }
                return answer;
            }
        }
    }
    bool has_min = min > 0;
    bool has_max = max < RegExpTree::kInfinity;
    bool needs_counter = has_min || has_max;
    int reg_ctr = needs_counter
        ? compiler->AllocateRegister()
        : RegExpCompiler::kNoRegister;
    LoopChoiceNode* center = alloc->newInfallible<LoopChoiceNode>(alloc, body->min_match() == 0,
                                                                  compiler->read_backward());
    if (not_at_start && !compiler->read_backward())
        center->set_not_at_start();
    RegExpNode* loop_return = needs_counter
        ? static_cast<RegExpNode*>(ActionNode::IncrementRegister(reg_ctr, center))
        : static_cast<RegExpNode*>(center);
    if (body_can_be_empty) {
        // If the body can be empty we need to check if it was and then
        // backtrack.
        loop_return = ActionNode::EmptyMatchCheck(body_start_reg,
                                                  reg_ctr,
                                                  min,
                                                  loop_return);
    }
    RegExpNode* body_node = body->ToNode(compiler, loop_return);
    if (body_can_be_empty) {
        // If the body can be empty we need to store the start position
        // so we can bail out if it was empty.
        body_node = ActionNode::StorePosition(body_start_reg, false, body_node);
    }
    if (needs_capture_clearing) {
        // Before entering the body of this loop we need to clear captures.
        body_node = ActionNode::ClearCaptures(capture_registers, body_node);
    }
    GuardedAlternative body_alt(body_node);
    if (has_max) {
        Guard* body_guard = alloc->newInfallible<Guard>(reg_ctr, Guard::LT, max);
        body_alt.AddGuard(alloc, body_guard);
    }
    GuardedAlternative rest_alt(on_success);
    if (has_min) {
        Guard* rest_guard = alloc->newInfallible<Guard>(reg_ctr, Guard::GEQ, min);
        rest_alt.AddGuard(alloc, rest_guard);
    }
    if (is_greedy) {
        center->AddLoopAlternative(body_alt);
        center->AddContinueAlternative(rest_alt);
    } else {
        center->AddContinueAlternative(rest_alt);
        center->AddLoopAlternative(body_alt);
    }
    if (needs_counter)
        return ActionNode::SetRegister(reg_ctr, 0, center);
    return center;
}

RegExpNode*
RegExpAssertion::ToNode(RegExpCompiler* compiler,
                        RegExpNode* on_success)
{
    NodeInfo info;
    LifoAlloc* alloc = compiler->alloc();

    switch (assertion_type()) {
      case START_OF_LINE:
        return AssertionNode::AfterNewline(on_success);
      case START_OF_INPUT:
        return AssertionNode::AtStart(on_success);
      case BOUNDARY:
        return AssertionNode::AtBoundary(on_success);
      case NON_BOUNDARY:
        return AssertionNode::AtNonBoundary(on_success);
      case END_OF_INPUT:
        return AssertionNode::AtEnd(on_success);
      case END_OF_LINE: {
        // Compile $ in multiline regexps as an alternation with a positive
        // lookahead in one side and an end-of-input on the other side.
        // We need two registers for the lookahead.
        int stack_pointer_register = compiler->AllocateRegister();
        int position_register = compiler->AllocateRegister();
        // The ChoiceNode to distinguish between a newline and end-of-input.
        ChoiceNode* result = alloc->newInfallible<ChoiceNode>(alloc, 2);
        // Create a newline atom.
        CharacterRangeVector* newline_ranges = alloc->newInfallible<CharacterRangeVector>(*alloc);
        CharacterRange::AddClassEscape(alloc, 'n', newline_ranges);
        RegExpCharacterClass* newline_atom = alloc->newInfallible<RegExpCharacterClass>('n');
        TextNode* newline_matcher =
            alloc->newInfallible<TextNode>(newline_atom, false,
                ActionNode::PositiveSubmatchSuccess(stack_pointer_register,
                                                    position_register,
                                                    0,  // No captures inside.
                                                    -1,  // Ignored if no captures.
                                                    on_success));
        // Create an end-of-input matcher.
        RegExpNode* end_of_line =
            ActionNode::BeginSubmatch(stack_pointer_register, position_register, newline_matcher);

        // Add the two alternatives to the ChoiceNode.
        GuardedAlternative eol_alternative(end_of_line);
        result->AddAlternative(eol_alternative);
        GuardedAlternative end_alternative(AssertionNode::AtEnd(on_success));
        result->AddAlternative(end_alternative);
        return result;
      }
      case NOT_AFTER_LEAD_SURROGATE:
        return AssertionNode::NotAfterLeadSurrogate(on_success);
      case NOT_IN_SURROGATE_PAIR:
        return AssertionNode::NotInSurrogatePair(on_success);
      default:
        MOZ_CRASH("Bad assertion type");
    }
    return on_success;
}

RegExpNode*
RegExpBackReference::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
{
    return compiler->alloc()->newInfallible<BackReferenceNode>(RegExpCapture::StartRegister(index()),
                                                               RegExpCapture::EndRegister(index()),
                                                               compiler->read_backward(),
                                                               on_success);
}

RegExpNode*
RegExpEmpty::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
{
    return on_success;
}

RegExpNode*
RegExpLookaround::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
{
    int stack_pointer_register = compiler->AllocateRegister();
    int position_register = compiler->AllocateRegister();

    const int registers_per_capture = 2;
    const int register_of_first_capture = 2;
    int register_count = capture_count_ * registers_per_capture;
    int register_start =
        register_of_first_capture + capture_from_ * registers_per_capture;

    RegExpNode* result;
    bool was_reading_backward = compiler->read_backward();
    compiler->set_read_backward(type() == LOOKBEHIND);

    if (is_positive()) {
        RegExpNode* bodyNode =
            body()->ToNode(compiler,
                           ActionNode::PositiveSubmatchSuccess(stack_pointer_register,
                                                               position_register,
                                                               register_count,
                                                               register_start,
                                                               on_success));
        result = ActionNode::BeginSubmatch(stack_pointer_register,
                                           position_register,
                                           bodyNode);
    } else {
        // We use a ChoiceNode for a negative lookahead because it has most of
        // the characteristics we need. It has the body of the lookahead as its
        // first alternative and the expression after the lookahead of the second
        // alternative. If the first alternative succeeds then the
        // NegativeSubmatchSuccess will unwind the stack including everything the
        // choice node set up and backtrack. If the first alternative fails then
        // the second alternative is tried, which is exactly the desired result
        // for a negative lookahead. The NegativeLookaheadChoiceNode is a special
        // ChoiceNode that knows to ignore the first exit when calculating quick
        // checks.
        LifoAlloc* alloc = compiler->alloc();

        RegExpNode* success =
            alloc->newInfallible<NegativeSubmatchSuccess>(alloc,
                                                          stack_pointer_register,
                                                          position_register,
                                                          register_count,
                                                          register_start);
        GuardedAlternative body_alt(body()->ToNode(compiler, success));

        ChoiceNode* choice_node =
            alloc->newInfallible<NegativeLookaheadChoiceNode>(alloc, body_alt, GuardedAlternative(on_success));

        result = ActionNode::BeginSubmatch(stack_pointer_register,
                                         position_register,
                                         choice_node);
    }
    compiler->set_read_backward(was_reading_backward);
    return result;
}

RegExpNode*
RegExpCapture::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
{
    return ToNode(body(), index(), compiler, on_success);
}

/* static */ RegExpNode*
RegExpCapture::ToNode(RegExpTree* body,
                      int index,
                      RegExpCompiler* compiler,
                      RegExpNode* on_success)
{
    MOZ_ASSERT(body);
    int start_reg = RegExpCapture::StartRegister(index);
    int end_reg = RegExpCapture::EndRegister(index);
    if (compiler->read_backward()) {
        std::swap(start_reg, end_reg);
    }
    RegExpNode* store_end = ActionNode::StorePosition(end_reg, true, on_success);
    RegExpNode* body_node = body->ToNode(compiler, store_end);
    return ActionNode::StorePosition(start_reg, true, body_node);
}

RegExpNode*
RegExpAlternative::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
{
    const RegExpTreeVector& children = nodes();
    RegExpNode* current = on_success;
    if (compiler->read_backward()) {
        for (int i = 0; i < children.length(); i++) {
            current = children[i]->ToNode(compiler, current);
        }
    } else {
        for (int i = children.length() - 1; i >= 0; i--) {
            current = children[i]->ToNode(compiler, current);
        }
    }
    return current;
}

// -------------------------------------------------------------------
// BoyerMooreLookahead

ContainedInLattice
irregexp::AddRange(ContainedInLattice containment,
                   const int* ranges,
                   int ranges_length,
                   Interval new_range)
{
    MOZ_ASSERT((ranges_length & 1) == 1);
    MOZ_ASSERT(ranges[ranges_length - 1] == kMaxUtf16CodeUnit + 1);
    if (containment == kLatticeUnknown) return containment;
    bool inside = false;
    int last = 0;
    for (int i = 0; i < ranges_length; inside = !inside, last = ranges[i], i++) {
        // Consider the range from last to ranges[i].
        // We haven't got to the new range yet.
        if (ranges[i] <= new_range.from())
            continue;

        // New range is wholly inside last-ranges[i].  Note that new_range.to() is
        // inclusive, but the values in ranges are not.
        if (last <= new_range.from() && new_range.to() < ranges[i])
            return Combine(containment, inside ? kLatticeIn : kLatticeOut);

        return kLatticeUnknown;
    }
    return containment;
}

void
BoyerMoorePositionInfo::Set(int character)
{
    SetInterval(Interval(character, character));
}

void
BoyerMoorePositionInfo::SetInterval(const Interval& interval)
{
    s_ = AddRange(s_, kSpaceRanges, kSpaceRangeCount, interval);
    if (unicode_ignore_case_)
        w_ = AddRange(w_, kIgnoreCaseWordRanges, kIgnoreCaseWordRangeCount, interval);
    else
        w_ = AddRange(w_, kWordRanges, kWordRangeCount, interval);
    d_ = AddRange(d_, kDigitRanges, kDigitRangeCount, interval);
    surrogate_ =
        AddRange(surrogate_, kSurrogateRanges, kSurrogateRangeCount, interval);
    if (interval.to() - interval.from() >= kMapSize - 1) {
        if (map_count_ != kMapSize) {
            map_count_ = kMapSize;
            for (int i = 0; i < kMapSize; i++)
                map_[i] = true;
        }
        return;
    }
    for (int i = interval.from(); i <= interval.to(); i++) {
        int mod_character = (i & kMask);
        if (!map_[mod_character]) {
            map_count_++;
            map_[mod_character] = true;
        }
        if (map_count_ == kMapSize)
            return;
    }
}

void
BoyerMoorePositionInfo::SetAll()
{
    s_ = w_ = d_ = kLatticeUnknown;
    if (map_count_ != kMapSize) {
        map_count_ = kMapSize;
        for (int i = 0; i < kMapSize; i++)
            map_[i] = true;
    }
}

BoyerMooreLookahead::BoyerMooreLookahead(LifoAlloc* alloc, size_t length, RegExpCompiler* compiler)
  : length_(length), compiler_(compiler), bitmaps_(*alloc)
{
    bool unicode_ignore_case = compiler->unicode() && compiler->ignore_case();
    max_char_ = MaximumCharacter(compiler->ascii());

    bitmaps_.reserve(length);
    for (size_t i = 0; i < length; i++)
        bitmaps_.append(alloc->newInfallible<BoyerMoorePositionInfo>(alloc, unicode_ignore_case));
}

// Find the longest range of lookahead that has the fewest number of different
// characters that can occur at a given position.  Since we are optimizing two
// different parameters at once this is a tradeoff.
bool BoyerMooreLookahead::FindWorthwhileInterval(int* from, int* to) {
  int biggest_points = 0;
  // If more than 32 characters out of 128 can occur it is unlikely that we can
  // be lucky enough to step forwards much of the time.
  const int kMaxMax = 32;
  for (int max_number_of_chars = 4;
       max_number_of_chars < kMaxMax;
       max_number_of_chars *= 2) {
    biggest_points =
        FindBestInterval(max_number_of_chars, biggest_points, from, to);
  }
  if (biggest_points == 0) return false;
  return true;
}

// Find the highest-points range between 0 and length_ where the character
// information is not too vague.  'Too vague' means that there are more than
// max_number_of_chars that can occur at this position.  Calculates the number
// of points as the product of width-of-the-range and
// probability-of-finding-one-of-the-characters, where the probability is
// calculated using the frequency distribution of the sample subject string.
int
BoyerMooreLookahead::FindBestInterval(int max_number_of_chars, int old_biggest_points,
                                      int* from, int* to)
{
    int biggest_points = old_biggest_points;
    static const int kSize = RegExpMacroAssembler::kTableSize;
    for (int i = 0; i < length_; ) {
        while (i < length_ && Count(i) > max_number_of_chars) i++;
        if (i == length_) break;
        int remembered_from = i;
        bool union_map[kSize];
        for (int j = 0; j < kSize; j++) union_map[j] = false;
        while (i < length_ && Count(i) <= max_number_of_chars) {
            BoyerMoorePositionInfo* map = bitmaps_[i];
            for (int j = 0; j < kSize; j++) union_map[j] |= map->at(j);
            i++;
        }
        int frequency = 0;
        for (int j = 0; j < kSize; j++) {
            if (union_map[j]) {
                // Add 1 to the frequency to give a small per-character boost for
                // the cases where our sampling is not good enough and many
                // characters have a frequency of zero.  This means the frequency
                // can theoretically be up to 2*kSize though we treat it mostly as
                // a fraction of kSize.
                frequency += compiler_->frequency_collator()->Frequency(j) + 1;
            }
        }
        // We use the probability of skipping times the distance we are skipping to
        // judge the effectiveness of this.  Actually we have a cut-off:  By
        // dividing by 2 we switch off the skipping if the probability of skipping
        // is less than 50%.  This is because the multibyte mask-and-compare
        // skipping in quickcheck is more likely to do well on this case.
        bool in_quickcheck_range = ((i - remembered_from < 4) ||
                                    (compiler_->ascii() ? remembered_from <= 4 : remembered_from <= 2));
        // Called 'probability' but it is only a rough estimate and can actually
        // be outside the 0-kSize range.
        int probability = (in_quickcheck_range ? kSize / 2 : kSize) - frequency;
        int points = (i - remembered_from) * probability;
        if (points > biggest_points) {
            *from = remembered_from;
            *to = i - 1;
            biggest_points = points;
        }
    }
    return biggest_points;
}

// Take all the characters that will not prevent a successful match if they
// occur in the subject string in the range between min_lookahead and
// max_lookahead (inclusive) measured from the current position.  If the
// character at max_lookahead offset is not one of these characters, then we
// can safely skip forwards by the number of characters in the range.
int BoyerMooreLookahead::GetSkipTable(int min_lookahead,
                                      int max_lookahead,
                                      uint8_t* boolean_skip_table)
{
    const int kSize = RegExpMacroAssembler::kTableSize;

    const int kSkipArrayEntry = 0;
    const int kDontSkipArrayEntry = 1;

    for (int i = 0; i < kSize; i++)
        boolean_skip_table[i] = kSkipArrayEntry;
    int skip = max_lookahead + 1 - min_lookahead;

    for (int i = max_lookahead; i >= min_lookahead; i--) {
        BoyerMoorePositionInfo* map = bitmaps_[i];
        for (int j = 0; j < kSize; j++) {
            if (map->at(j))
                boolean_skip_table[j] = kDontSkipArrayEntry;
        }
    }

    return skip;
}

// See comment on the implementation of GetSkipTable.
bool
BoyerMooreLookahead::EmitSkipInstructions(RegExpMacroAssembler* masm)
{
    const int kSize = RegExpMacroAssembler::kTableSize;

    int min_lookahead = 0;
    int max_lookahead = 0;

    if (!FindWorthwhileInterval(&min_lookahead, &max_lookahead))
        return false;

    bool found_single_character = false;
    int single_character = 0;
    for (int i = max_lookahead; i >= min_lookahead; i--) {
        BoyerMoorePositionInfo* map = bitmaps_[i];
        if (map->map_count() > 1 ||
            (found_single_character && map->map_count() != 0)) {
            found_single_character = false;
            break;
        }
        for (int j = 0; j < kSize; j++) {
            if (map->at(j)) {
                found_single_character = true;
                single_character = j;
                break;
            }
        }
    }

    int lookahead_width = max_lookahead + 1 - min_lookahead;

    if (found_single_character && lookahead_width == 1 && max_lookahead < 3) {
        // The mask-compare can probably handle this better.
        return false;
    }

    if (found_single_character) {
        jit::Label cont, again;
        masm->Bind(&again);
        masm->LoadCurrentCharacter(max_lookahead, &cont, true);
        if (max_char_ > kSize) {
            masm->CheckCharacterAfterAnd(single_character,
                                         RegExpMacroAssembler::kTableMask,
                                         &cont);
        } else {
            masm->CheckCharacter(single_character, &cont);
        }
        masm->AdvanceCurrentPosition(lookahead_width);
        masm->JumpOrBacktrack(&again);
        masm->Bind(&cont);
        return true;
    }

    RegExpShared::JitCodeTable boolean_skip_table;
    {
        AutoEnterOOMUnsafeRegion oomUnsafe;
        boolean_skip_table.reset(static_cast<uint8_t*>(js_malloc(kSize)));
        if (!boolean_skip_table)
            oomUnsafe.crash("Table malloc");
    }

    int skip_distance = GetSkipTable(min_lookahead, max_lookahead, boolean_skip_table.get());
    MOZ_ASSERT(skip_distance != 0);

    jit::Label cont, again;
    masm->Bind(&again);
    masm->LoadCurrentCharacter(max_lookahead, &cont, true);
    masm->CheckBitInTable(Move(boolean_skip_table), &cont);
    masm->AdvanceCurrentPosition(skip_distance);
    masm->JumpOrBacktrack(&again);
    masm->Bind(&cont);

    return true;
}

bool
BoyerMooreLookahead::CheckOverRecursed()
{
    JS_CHECK_RECURSION(compiler()->cx(), compiler()->SetRegExpTooBig(); return false);
    return true;
}

// -------------------------------------------------------------------
// Trace

bool Trace::DeferredAction::Mentions(int that)
{
    if (action_type() == ActionNode::CLEAR_CAPTURES) {
        Interval range = static_cast<DeferredClearCaptures*>(this)->range();
        return range.Contains(that);
    }
    return reg() == that;
}

bool Trace::mentions_reg(int reg)
{
    for (DeferredAction* action = actions_; action != nullptr; action = action->next()) {
        if (action->Mentions(reg))
            return true;
    }
    return false;
}

bool
Trace::GetStoredPosition(int reg, int* cp_offset)
{
    MOZ_ASSERT(0 == *cp_offset);
    for (DeferredAction* action = actions_; action != nullptr; action = action->next()) {
        if (action->Mentions(reg)) {
            if (action->action_type() == ActionNode::STORE_POSITION) {
                *cp_offset = static_cast<DeferredCapture*>(action)->cp_offset();
                return true;
            }
            return false;
        }
    }
    return false;
}

int
Trace::FindAffectedRegisters(LifoAlloc* alloc, OutSet* affected_registers)
{
    int max_register = RegExpCompiler::kNoRegister;
    for (DeferredAction* action = actions_; action != nullptr; action = action->next()) {
        if (action->action_type() == ActionNode::CLEAR_CAPTURES) {
            Interval range = static_cast<DeferredClearCaptures*>(action)->range();
            for (int i = range.from(); i <= range.to(); i++)
                affected_registers->Set(alloc, i);
            if (range.to() > max_register) max_register = range.to();
        } else {
            affected_registers->Set(alloc, action->reg());
            if (action->reg() > max_register) max_register = action->reg();
        }
    }
    return max_register;
}

void
Trace::RestoreAffectedRegisters(RegExpMacroAssembler* assembler,
                                int max_register,
                                OutSet& registers_to_pop,
                                OutSet& registers_to_clear)
{
    for (int reg = max_register; reg >= 0; reg--) {
        if (registers_to_pop.Get(reg)) assembler->PopRegister(reg);
        else if (registers_to_clear.Get(reg)) {
            int clear_to = reg;
            while (reg > 0 && registers_to_clear.Get(reg - 1))
                reg--;
            assembler->ClearRegisters(reg, clear_to);
        }
    }
}

enum DeferredActionUndoType {
    DEFER_IGNORE,
    DEFER_RESTORE,
    DEFER_CLEAR
};

void
Trace::PerformDeferredActions(LifoAlloc* alloc,
                              RegExpMacroAssembler* assembler,
                              int max_register,
                              OutSet& affected_registers,
                              OutSet* registers_to_pop,
                              OutSet* registers_to_clear)
{
    // The "+1" is to avoid a push_limit of zero if stack_limit_slack() is 1.
    const int push_limit = (assembler->stack_limit_slack() + 1) / 2;
    static const int kNoStore = INT32_MIN;

    // Count pushes performed to force a stack limit check occasionally.
    int pushes = 0;

    for (int reg = 0; reg <= max_register; reg++) {
        if (!affected_registers.Get(reg))
            continue;

        // The chronologically first deferred action in the trace
        // is used to infer the action needed to restore a register
        // to its previous state (or not, if it's safe to ignore it).
        DeferredActionUndoType undo_action = DEFER_IGNORE;

        int value = 0;
        bool absolute = false;
        bool clear = false;
        int store_position = kNoStore;
        // This is a little tricky because we are scanning the actions in reverse
        // historical order (newest first).
        for (DeferredAction* action = actions_;
             action != nullptr;
             action = action->next()) {
            if (action->Mentions(reg)) {
                switch (action->action_type()) {
                  case ActionNode::SET_REGISTER: {
                    Trace::DeferredSetRegister* psr =
                        static_cast<Trace::DeferredSetRegister*>(action);
                    if (!absolute) {
                        value += psr->value();
                        absolute = true;
                    }
                    // SET_REGISTER is currently only used for newly introduced loop
                    // counters. They can have a significant previous value if they
                    // occour in a loop. TODO(lrn): Propagate this information, so
                    // we can set undo_action to IGNORE if we know there is no value to
                    // restore.
                    undo_action = DEFER_RESTORE;
                    MOZ_ASSERT(store_position == kNoStore);
                    MOZ_ASSERT(!clear);
                    break;
                  }
                  case ActionNode::INCREMENT_REGISTER:
                    if (!absolute) {
                        value++;
                    }
                    MOZ_ASSERT(store_position == kNoStore);
                    MOZ_ASSERT(!clear);
                    undo_action = DEFER_RESTORE;
                    break;
                  case ActionNode::STORE_POSITION: {
                    Trace::DeferredCapture* pc =
                        static_cast<Trace::DeferredCapture*>(action);
                    if (!clear && store_position == kNoStore) {
                        store_position = pc->cp_offset();
                    }

                    // For captures we know that stores and clears alternate.
                    // Other register, are never cleared, and if the occur
                    // inside a loop, they might be assigned more than once.
                    if (reg <= 1) {
                        // Registers zero and one, aka "capture zero", is
                        // always set correctly if we succeed. There is no
                        // need to undo a setting on backtrack, because we
                        // will set it again or fail.
                        undo_action = DEFER_IGNORE;
                    } else {
                        undo_action = pc->is_capture() ? DEFER_CLEAR : DEFER_RESTORE;
                    }
                    MOZ_ASSERT(!absolute);
                    MOZ_ASSERT(value == 0);
                    break;
                  }
                  case ActionNode::CLEAR_CAPTURES: {
                    // Since we're scanning in reverse order, if we've already
                    // set the position we have to ignore historically earlier
                    // clearing operations.
                    if (store_position == kNoStore) {
                        clear = true;
                    }
                    undo_action = DEFER_RESTORE;
                    MOZ_ASSERT(!absolute);
                    MOZ_ASSERT(value == 0);
                    break;
                  }
                  default:
                    MOZ_CRASH("Bad action");
                }
            }
        }
        // Prepare for the undo-action (e.g., push if it's going to be popped).
        if (undo_action == DEFER_RESTORE) {
            pushes++;
            RegExpMacroAssembler::StackCheckFlag stack_check =
                RegExpMacroAssembler::kNoStackLimitCheck;
            if (pushes == push_limit) {
                stack_check = RegExpMacroAssembler::kCheckStackLimit;
                pushes = 0;
            }

            assembler->PushRegister(reg, stack_check);
            registers_to_pop->Set(alloc, reg);
        } else if (undo_action == DEFER_CLEAR) {
            registers_to_clear->Set(alloc, reg);
        }
        // Perform the chronologically last action (or accumulated increment)
        // for the register.
        if (store_position != kNoStore) {
            assembler->WriteCurrentPositionToRegister(reg, store_position);
        } else if (clear) {
            assembler->ClearRegisters(reg, reg);
        } else if (absolute) {
            assembler->SetRegister(reg, value);
        } else if (value != 0) {
            assembler->AdvanceRegister(reg, value);
        }
    }
}

// This is called as we come into a loop choice node and some other tricky
// nodes.  It normalizes the state of the code generator to ensure we can
// generate generic code.
void Trace::Flush(RegExpCompiler* compiler, RegExpNode* successor)
{
    RegExpMacroAssembler* assembler = compiler->macro_assembler();

    MOZ_ASSERT(!is_trivial());

    if (actions_ == nullptr && backtrack() == nullptr) {
        // Here we just have some deferred cp advances to fix and we are back to
        // a normal situation.  We may also have to forget some information gained
        // through a quick check that was already performed.
        if (cp_offset_ != 0) assembler->AdvanceCurrentPosition(cp_offset_);
        // Create a new trivial state and generate the node with that.
        Trace new_state;
        successor->Emit(compiler, &new_state);
        return;
    }

    // Generate deferred actions here along with code to undo them again.
    OutSet affected_registers;

    if (backtrack() != nullptr) {
        // Here we have a concrete backtrack location.  These are set up by choice
        // nodes and so they indicate that we have a deferred save of the current
        // position which we may need to emit here.
        assembler->PushCurrentPosition();
    }

    int max_register = FindAffectedRegisters(compiler->alloc(), &affected_registers);
    OutSet registers_to_pop;
    OutSet registers_to_clear;
    PerformDeferredActions(compiler->alloc(),
                           assembler,
                           max_register,
                           affected_registers,
                           &registers_to_pop,
                           &registers_to_clear);
    if (cp_offset_ != 0)
        assembler->AdvanceCurrentPosition(cp_offset_);

    // Create a new trivial state and generate the node with that.
    jit::Label undo;
    assembler->PushBacktrack(&undo);
    Trace new_state;
    successor->Emit(compiler, &new_state);

    // On backtrack we need to restore state.
    assembler->BindBacktrack(&undo);
    RestoreAffectedRegisters(assembler,
                             max_register,
                             registers_to_pop,
                             registers_to_clear);
    if (backtrack() == nullptr) {
        assembler->Backtrack();
    } else {
        assembler->PopCurrentPosition();
        assembler->JumpOrBacktrack(backtrack());
    }
}

void
Trace::InvalidateCurrentCharacter()
{
    characters_preloaded_ = 0;
}

void
Trace::AdvanceCurrentPositionInTrace(int by, RegExpCompiler* compiler)
{
    // We don't have an instruction for shifting the current character register
    // down or for using a shifted value for anything so lets just forget that
    // we preloaded any characters into it.
    characters_preloaded_ = 0;
    // Adjust the offsets of the quick check performed information.  This
    // information is used to find out what we already determined about the
    // characters by means of mask and compare.
    quick_check_performed_.Advance(by, compiler->ascii());
    cp_offset_ += by;
    if (cp_offset_ > RegExpMacroAssembler::kMaxCPOffset) {
        compiler->SetRegExpTooBig();
        cp_offset_ = 0;
    }
    bound_checked_up_to_ = Max(0, bound_checked_up_to_ - by);
}

void
OutSet::Set(LifoAlloc* alloc, unsigned value)
{
    if (value < kFirstLimit) {
        first_ |= (1 << value);
    } else {
        if (remaining_ == nullptr)
            remaining_ = alloc->newInfallible<RemainingVector>(*alloc);

        for (size_t i = 0; i < remaining().length(); i++) {
            if (remaining()[i] == value)
                return;
        }
        remaining().append(value);
    }
}

bool
OutSet::Get(unsigned value)
{
    if (value < kFirstLimit)
        return (first_ & (1 << value)) != 0;
    if (remaining_ == nullptr)
        return false;
    for (size_t i = 0; i < remaining().length(); i++) {
        if (remaining()[i] == value)
            return true;
    }
    return false;
}

// -------------------------------------------------------------------
// Graph emitting

void
NegativeSubmatchSuccess::Emit(RegExpCompiler* compiler, Trace* trace)
{
    RegExpMacroAssembler* assembler = compiler->macro_assembler();

    // Omit flushing the trace. We discard the entire stack frame anyway.

    if (!label()->bound()) {
        // We are completely independent of the trace, since we ignore it,
        // so this code can be used as the generic version.
        assembler->Bind(label());
    }

    // Throw away everything on the backtrack stack since the start
    // of the negative submatch and restore the character position.
    assembler->ReadCurrentPositionFromRegister(current_position_register_);
    assembler->ReadBacktrackStackPointerFromRegister(stack_pointer_register_);

    if (clear_capture_count_ > 0) {
        // Clear any captures that might have been performed during the success
        // of the body of the negative look-ahead.
        int clear_capture_end = clear_capture_start_ + clear_capture_count_ - 1;
        assembler->ClearRegisters(clear_capture_start_, clear_capture_end);
    }

    // Now that we have unwound the stack we find at the top of the stack the
    // backtrack that the BeginSubmatch node got.
    assembler->Backtrack();
}

void
EndNode::Emit(RegExpCompiler* compiler, Trace* trace)
{
    if (!trace->is_trivial()) {
        trace->Flush(compiler, this);
        return;
    }
    RegExpMacroAssembler* assembler = compiler->macro_assembler();
    if (!label()->bound()) {
        assembler->Bind(label());
    }
    switch (action_) {
    case ACCEPT:
        assembler->Succeed();
        return;
    case BACKTRACK:
        assembler->JumpOrBacktrack(trace->backtrack());
        return;
    case NEGATIVE_SUBMATCH_SUCCESS:
        // This case is handled in a different virtual method.
        MOZ_CRASH("Bad action: NEGATIVE_SUBMATCH_SUCCESS");
    }
    MOZ_CRASH("Bad action");
}

// Emit the code to check for a ^ in multiline mode (1-character lookbehind
// that matches newline or the start of input).
static void
EmitHat(RegExpCompiler* compiler, RegExpNode* on_success, Trace* trace)
{
    RegExpMacroAssembler* assembler = compiler->macro_assembler();

    // We will be loading the previous character into the current character
    // register.
    Trace new_trace(*trace);
    new_trace.InvalidateCurrentCharacter();

    // A positive (> 0) cp_offset means we've already successfully matched a
    // non-empty-width part of the pattern, and thus cannot be at or before the
    // start of the subject string. We can thus skip both at-start and
    // bounds-checks when loading the one-character lookbehind.
    const bool may_be_at_or_before_subject_string_start = new_trace.cp_offset() <= 0;

    jit::Label ok;
    if (may_be_at_or_before_subject_string_start) {
        // The start of input counts as a newline in this context, so skip to ok if
        // we are at the start.
        assembler->CheckAtStart(new_trace.cp_offset(), &ok);
    }

    // If we've already checked that we are not at the start of input, it's okay
    // to load the previous character without bounds checks.
    const bool can_skip_bounds_check = !may_be_at_or_before_subject_string_start;
    assembler->LoadCurrentCharacter(new_trace.cp_offset() -1, new_trace.backtrack(), can_skip_bounds_check);

    if (!assembler->CheckSpecialCharacterClass('n', new_trace.backtrack())) {
        // Newline means \n, \r, 0x2028 or 0x2029.
        if (!compiler->ascii())
            assembler->CheckCharacterAfterAnd(0x2028, 0xfffe, &ok);
        assembler->CheckCharacter('\n', &ok);
        assembler->CheckNotCharacter('\r', new_trace.backtrack());
    }
    assembler->Bind(&ok);
    on_success->Emit(compiler, &new_trace);
}

// Assert that the next character cannot be a part of a surrogate pair.
static void
EmitNotAfterLeadSurrogate(RegExpCompiler* compiler, RegExpNode* on_success, Trace* trace)
{
    RegExpMacroAssembler* assembler = compiler->macro_assembler();

    // We will be loading the previous character into the current character
    // register.
    Trace new_trace(*trace);
    new_trace.InvalidateCurrentCharacter();

    jit::Label ok;
    if (new_trace.cp_offset() <= 0) {
        assembler->CheckAtStart(new_trace.cp_offset(), &ok);
    }

    assembler->LoadCurrentCharacter(new_trace.cp_offset() - 1, new_trace.backtrack(), false);
    assembler->CheckCharacterInRange(unicode::LeadSurrogateMin, unicode::LeadSurrogateMax,
                                     new_trace.backtrack());

    assembler->Bind(&ok);
    on_success->Emit(compiler, &new_trace);
}

// Assert that the next character is not a trail surrogate that has a
// corresponding lead surrogate.
static void
EmitNotInSurrogatePair(RegExpCompiler* compiler, RegExpNode* on_success, Trace* trace)
{
    RegExpMacroAssembler* assembler = compiler->macro_assembler();

    jit::Label ok;
    assembler->CheckPosition(trace->cp_offset(), &ok);

    // We will be loading the next and previous characters into the current
    // character register.
    Trace new_trace(*trace);
    new_trace.InvalidateCurrentCharacter();

    if (new_trace.cp_offset() <= 0) {
        assembler->CheckAtStart(new_trace.cp_offset(), &ok);
    }

    // First check if next character is a trail surrogate.
    assembler->LoadCurrentCharacter(new_trace.cp_offset(), new_trace.backtrack(), false);
    assembler->CheckCharacterNotInRange(unicode::TrailSurrogateMin, unicode::TrailSurrogateMax,
                                        &ok);

    // Next check if previous character is a lead surrogate.
    // We already checked that we are not at the start of input so it must be
    // OK to load the previous character.
    assembler->LoadCurrentCharacter(new_trace.cp_offset() - 1, new_trace.backtrack(), false);
    assembler->CheckCharacterInRange(unicode::LeadSurrogateMin, unicode::LeadSurrogateMax,
                                     new_trace.backtrack());

    assembler->Bind(&ok);
    on_success->Emit(compiler, &new_trace);
}

// Check for [0-9A-Z_a-z].
static void
EmitWordCheck(RegExpMacroAssembler* assembler,
              jit::Label* word, jit::Label* non_word, bool fall_through_on_word,
              bool unicode_ignore_case)
{
    if (!unicode_ignore_case &&
        assembler->CheckSpecialCharacterClass(fall_through_on_word ? 'w' : 'W',
                                              fall_through_on_word ? non_word : word))
    {
        // Optimized implementation available.
        return;
    }

    if (unicode_ignore_case) {
        assembler->CheckCharacter(0x017F, word);
        assembler->CheckCharacter(0x212A, word);
    }

    assembler->CheckCharacterGT('z', non_word);
    assembler->CheckCharacterLT('0', non_word);
    assembler->CheckCharacterGT('a' - 1, word);
    assembler->CheckCharacterLT('9' + 1, word);
    assembler->CheckCharacterLT('A', non_word);
    assembler->CheckCharacterLT('Z' + 1, word);

    if (fall_through_on_word)
        assembler->CheckNotCharacter('_', non_word);
    else
        assembler->CheckCharacter('_', word);
}

// Emit the code to handle \b and \B (word-boundary or non-word-boundary).
void
AssertionNode::EmitBoundaryCheck(RegExpCompiler* compiler, Trace* trace)
{
    RegExpMacroAssembler* assembler = compiler->macro_assembler();
    Trace::TriBool next_is_word_character = Trace::UNKNOWN;
    bool not_at_start = (trace->at_start() == Trace::FALSE_VALUE);
    BoyerMooreLookahead* lookahead = bm_info(not_at_start);
    if (lookahead == nullptr) {
        int eats_at_least =
            Min(kMaxLookaheadForBoyerMoore, EatsAtLeast(kMaxLookaheadForBoyerMoore,
                                                        kRecursionBudget,
                                                        not_at_start));
        if (eats_at_least >= 1) {
            BoyerMooreLookahead* bm =
                alloc()->newInfallible<BoyerMooreLookahead>(alloc(), eats_at_least, compiler);
            FillInBMInfo(0, kRecursionBudget, bm, not_at_start);
            if (bm->at(0)->is_non_word())
                next_is_word_character = Trace::FALSE_VALUE;
            if (bm->at(0)->is_word()) next_is_word_character = Trace::TRUE_VALUE;
        }
    } else {
        if (lookahead->at(0)->is_non_word())
            next_is_word_character = Trace::FALSE_VALUE;
        if (lookahead->at(0)->is_word())
            next_is_word_character = Trace::TRUE_VALUE;
    }
    bool at_boundary = (assertion_type_ == AssertionNode::AT_BOUNDARY);
    if (next_is_word_character == Trace::UNKNOWN) {
        jit::Label before_non_word;
        jit::Label before_word;
        if (trace->characters_preloaded() != 1) {
            assembler->LoadCurrentCharacter(trace->cp_offset(), &before_non_word);
        }
        // Fall through on non-word.
        EmitWordCheck(assembler, &before_word, &before_non_word, false,
                      compiler->unicode() && compiler->ignore_case());
        // Next character is not a word character.
        assembler->Bind(&before_non_word);
        jit::Label ok;
        BacktrackIfPrevious(compiler, trace, at_boundary ? kIsNonWord : kIsWord);
        assembler->JumpOrBacktrack(&ok);

        assembler->Bind(&before_word);
        BacktrackIfPrevious(compiler, trace, at_boundary ? kIsWord : kIsNonWord);
        assembler->Bind(&ok);
    } else if (next_is_word_character == Trace::TRUE_VALUE) {
        BacktrackIfPrevious(compiler, trace, at_boundary ? kIsWord : kIsNonWord);
    } else {
        MOZ_ASSERT(next_is_word_character == Trace::FALSE_VALUE);
        BacktrackIfPrevious(compiler, trace, at_boundary ? kIsNonWord : kIsWord);
    }
}

void
AssertionNode::BacktrackIfPrevious(RegExpCompiler* compiler,
                                   Trace* trace,
                                   AssertionNode::IfPrevious backtrack_if_previous)
{
    RegExpMacroAssembler* assembler = compiler->macro_assembler();
    Trace new_trace(*trace);
    new_trace.InvalidateCurrentCharacter();

    jit::Label fall_through, dummy;

    jit::Label* non_word = backtrack_if_previous == kIsNonWord ? new_trace.backtrack() : &fall_through;
    jit::Label* word     = backtrack_if_previous == kIsNonWord ? &fall_through : new_trace.backtrack();

    if (new_trace.cp_offset() <= 0) {
        // The start of input counts as a non-word character, so the question is
        // decided if we are at the start.
        assembler->CheckAtStart(new_trace.cp_offset(), non_word);
    }
    // We already checked that we are not at the start of input so it must be
    // OK to load the previous character.
    assembler->LoadCurrentCharacter(new_trace.cp_offset() - 1, &dummy, false);
    EmitWordCheck(assembler, word, non_word, backtrack_if_previous == kIsNonWord,
                  compiler->unicode() && compiler->ignore_case());

    assembler->Bind(&fall_through);
    on_success()->Emit(compiler, &new_trace);
}

void
AssertionNode::GetQuickCheckDetails(QuickCheckDetails* details,
                                    RegExpCompiler* compiler,
                                    int filled_in,
                                    bool not_at_start)
{
    if (assertion_type_ == AT_START && not_at_start) {
        details->set_cannot_match();
        return;
    }
    return on_success()->GetQuickCheckDetails(details, compiler, filled_in, not_at_start);
}

void
AssertionNode::Emit(RegExpCompiler* compiler, Trace* trace)
{
    RegExpMacroAssembler* assembler = compiler->macro_assembler();
    switch (assertion_type_) {
      case AT_END: {
        jit::Label ok;
        assembler->CheckPosition(trace->cp_offset(), &ok);
        assembler->JumpOrBacktrack(trace->backtrack());
        assembler->Bind(&ok);
        break;
      }
      case AT_START: {
        if (trace->at_start() == Trace::FALSE_VALUE) {
            assembler->JumpOrBacktrack(trace->backtrack());
            return;
        }
        if (trace->at_start() == Trace::UNKNOWN) {
            assembler->CheckNotAtStart(trace->cp_offset(), trace->backtrack());
            Trace at_start_trace = *trace;
            at_start_trace.set_at_start(Trace::TRUE_VALUE);
            on_success()->Emit(compiler, &at_start_trace);
            return;
        }
      }
        break;
      case AFTER_NEWLINE:
        EmitHat(compiler, on_success(), trace);
        return;
      case AT_BOUNDARY:
      case AT_NON_BOUNDARY: {
        EmitBoundaryCheck(compiler, trace);
        return;
      }
      case NOT_AFTER_LEAD_SURROGATE:
        EmitNotAfterLeadSurrogate(compiler, on_success(), trace);
        return;
      case NOT_IN_SURROGATE_PAIR:
        EmitNotInSurrogatePair(compiler, on_success(), trace);
        return;
    }
    on_success()->Emit(compiler, trace);
}

static bool
DeterminedAlready(QuickCheckDetails* quick_check, int offset)
{
    if (quick_check == nullptr)
        return false;
    if (offset >= quick_check->characters())
        return false;
    return quick_check->positions(offset)->determines_perfectly;
}

static void
UpdateBoundsCheck(int index, int* checked_up_to)
{
    if (index > *checked_up_to)
        *checked_up_to = index;
}

static void
EmitBoundaryTest(RegExpMacroAssembler* masm,
                 int border,
                 jit::Label* fall_through,
                 jit::Label* above_or_equal,
                 jit::Label* below)
{
    if (below != fall_through) {
        masm->CheckCharacterLT(border, below);
        if (above_or_equal != fall_through)
            masm->JumpOrBacktrack(above_or_equal);
    } else {
        masm->CheckCharacterGT(border - 1, above_or_equal);
    }
}

static void
EmitDoubleBoundaryTest(RegExpMacroAssembler* masm,
                       int first,
                       int last,
                       jit::Label* fall_through,
                       jit::Label* in_range,
                       jit::Label* out_of_range)
{
    if (in_range == fall_through) {
        if (first == last)
            masm->CheckNotCharacter(first, out_of_range);
        else
            masm->CheckCharacterNotInRange(first, last, out_of_range);
    } else {
        if (first == last)
            masm->CheckCharacter(first, in_range);
        else
            masm->CheckCharacterInRange(first, last, in_range);
        if (out_of_range != fall_through)
            masm->JumpOrBacktrack(out_of_range);
    }
}

typedef InfallibleVector<int, 4> RangeBoundaryVector;

// even_label is for ranges[i] to ranges[i + 1] where i - start_index is even.
// odd_label is for ranges[i] to ranges[i + 1] where i - start_index is odd.
static void
EmitUseLookupTable(RegExpMacroAssembler* masm,
                   RangeBoundaryVector& ranges,
                   int start_index,
                   int end_index,
                   int min_char,
                   jit::Label* fall_through,
                   jit::Label* even_label,
                   jit::Label* odd_label)
{
    static const int kSize = RegExpMacroAssembler::kTableSize;
    static const int kMask = RegExpMacroAssembler::kTableMask;

    DebugOnly<int> base = (min_char & ~kMask);

    // Assert that everything is on one kTableSize page.
    for (int i = start_index; i <= end_index; i++)
        MOZ_ASSERT((ranges[i] & ~kMask) == base);
    MOZ_ASSERT(start_index == 0 || (ranges[start_index - 1] & ~kMask) <= base);

    char templ[kSize];
    jit::Label* on_bit_set;
    jit::Label* on_bit_clear;
    int bit;
    if (even_label == fall_through) {
        on_bit_set = odd_label;
        on_bit_clear = even_label;
        bit = 1;
    } else {
        on_bit_set = even_label;
        on_bit_clear = odd_label;
        bit = 0;
    }
    for (int i = 0; i < (ranges[start_index] & kMask) && i < kSize; i++)
        templ[i] = bit;
    int j = 0;
    bit ^= 1;
    for (int i = start_index; i < end_index; i++) {
        for (j = (ranges[i] & kMask); j < (ranges[i + 1] & kMask); j++) {
            templ[j] = bit;
        }
        bit ^= 1;
    }
    for (int i = j; i < kSize; i++) {
        templ[i] = bit;
    }

    // TODO(erikcorry): Cache these.
    RegExpShared::JitCodeTable ba;
    {
        AutoEnterOOMUnsafeRegion oomUnsafe;
        ba.reset(static_cast<uint8_t*>(js_malloc(kSize)));
        if (!ba)
            oomUnsafe.crash("Table malloc");
    }

    for (int i = 0; i < kSize; i++)
        ba[i] = templ[i];

    masm->CheckBitInTable(Move(ba), on_bit_set);
    if (on_bit_clear != fall_through)
        masm->JumpOrBacktrack(on_bit_clear);
}

static void
CutOutRange(RegExpMacroAssembler* masm,
            RangeBoundaryVector& ranges,
            int start_index,
            int end_index,
            int cut_index,
            jit::Label* even_label,
            jit::Label* odd_label)
{
    bool odd = (((cut_index - start_index) & 1) == 1);
    jit::Label* in_range_label = odd ? odd_label : even_label;
    jit::Label dummy;
    EmitDoubleBoundaryTest(masm,
                           ranges[cut_index],
                           ranges[cut_index + 1] - 1,
                           &dummy,
                           in_range_label,
                           &dummy);
    MOZ_ASSERT(!dummy.used());
    // Cut out the single range by rewriting the array.  This creates a new
    // range that is a merger of the two ranges on either side of the one we
    // are cutting out.  The oddity of the labels is preserved.
    for (int j = cut_index; j > start_index; j--)
        ranges[j] = ranges[j - 1];
    for (int j = cut_index + 1; j < end_index; j++)
        ranges[j] = ranges[j + 1];
}

// Unicode case.  Split the search space into kSize spaces that are handled
// with recursion.
static void
SplitSearchSpace(RangeBoundaryVector& ranges,
                 int start_index,
                 int end_index,
                 int* new_start_index,
                 int* new_end_index,
                 int* border)
{
    static const int kSize = RegExpMacroAssembler::kTableSize;
    static const int kMask = RegExpMacroAssembler::kTableMask;

    int first = ranges[start_index];
    int last = ranges[end_index] - 1;

    *new_start_index = start_index;
    *border = (ranges[start_index] & ~kMask) + kSize;
    while (*new_start_index < end_index) {
        if (ranges[*new_start_index] > *border)
            break;
        (*new_start_index)++;
    }
    // new_start_index is the index of the first edge that is beyond the
    // current kSize space.

    // For very large search spaces we do a binary chop search of the non-ASCII
    // space instead of just going to the end of the current kSize space.  The
    // heuristics are complicated a little by the fact that any 128-character
    // encoding space can be quickly tested with a table lookup, so we don't
    // wish to do binary chop search at a smaller granularity than that.  A
    // 128-character space can take up a lot of space in the ranges array if,
    // for example, we only want to match every second character (eg. the lower
    // case characters on some Unicode pages).
    int binary_chop_index = (end_index + start_index) / 2;
    // The first test ensures that we get to the code that handles the ASCII
    // range with a single not-taken branch, speeding up this important
    // character range (even non-ASCII charset-based text has spaces and
    // punctuation).
    if (*border - 1 > kMaxOneByteCharCode &&  // ASCII case.
        end_index - start_index > (*new_start_index - start_index) * 2 &&
        last - first > kSize * 2 &&
        binary_chop_index > *new_start_index &&
        ranges[binary_chop_index] >= first + 2 * kSize)
    {
        int scan_forward_for_section_border = binary_chop_index;;
        int new_border = (ranges[binary_chop_index] | kMask) + 1;

        while (scan_forward_for_section_border < end_index) {
            if (ranges[scan_forward_for_section_border] > new_border) {
                *new_start_index = scan_forward_for_section_border;
                *border = new_border;
                break;
            }
            scan_forward_for_section_border++;
        }
    }

    MOZ_ASSERT(*new_start_index > start_index);
    *new_end_index = *new_start_index - 1;
    if (ranges[*new_end_index] == *border)
        (*new_end_index)--;
    if (*border >= ranges[end_index]) {
        *border = ranges[end_index];
        *new_start_index = end_index;  // Won't be used.
        *new_end_index = end_index - 1;
    }
}

// Gets a series of segment boundaries representing a character class.  If the
// character is in the range between an even and an odd boundary (counting from
// start_index) then go to even_label, otherwise go to odd_label.  We already
// know that the character is in the range of min_char to max_char inclusive.
// Either label can be nullptr indicating backtracking.  Either label can also be
// equal to the fall_through label.
static void
GenerateBranches(RegExpMacroAssembler* masm,
                 RangeBoundaryVector& ranges,
                 int start_index,
                 int end_index,
                 char16_t min_char,
                 char16_t max_char,
                 jit::Label* fall_through,
                 jit::Label* even_label,
                 jit::Label* odd_label)
{
    int first = ranges[start_index];
    int last = ranges[end_index] - 1;

    MOZ_ASSERT(min_char < first);

    // Just need to test if the character is before or on-or-after
    // a particular character.
    if (start_index == end_index) {
        EmitBoundaryTest(masm, first, fall_through, even_label, odd_label);
        return;
    }

    // Another almost trivial case:  There is one interval in the middle that is
    // different from the end intervals.
    if (start_index + 1 == end_index) {
        EmitDoubleBoundaryTest(masm, first, last, fall_through, even_label, odd_label);
        return;
    }

    // It's not worth using table lookup if there are very few intervals in the
    // character class.
    if (end_index - start_index <= 6) {
        // It is faster to test for individual characters, so we look for those
        // first, then try arbitrary ranges in the second round.
        static int kNoCutIndex = -1;
        int cut = kNoCutIndex;
        for (int i = start_index; i < end_index; i++) {
            if (ranges[i] == ranges[i + 1] - 1) {
                cut = i;
                break;
            }
        }
        if (cut == kNoCutIndex) cut = start_index;
        CutOutRange(masm, ranges, start_index, end_index, cut, even_label, odd_label);
        MOZ_ASSERT(end_index - start_index >= 2);
        GenerateBranches(masm,
                         ranges,
                         start_index + 1,
                         end_index - 1,
                         min_char,
                         max_char,
                         fall_through,
                         even_label,
                         odd_label);
        return;
    }

    // If there are a lot of intervals in the regexp, then we will use tables to
    // determine whether the character is inside or outside the character class.
    static const int kBits = RegExpMacroAssembler::kTableSizeBits;

    if ((max_char >> kBits) == (min_char >> kBits)) {
        EmitUseLookupTable(masm,
                           ranges,
                           start_index,
                           end_index,
                           min_char,
                           fall_through,
                           even_label,
                           odd_label);
        return;
    }

    if ((min_char >> kBits) != (first >> kBits)) {
        masm->CheckCharacterLT(first, odd_label);
        GenerateBranches(masm,
                         ranges,
                         start_index + 1,
                         end_index,
                         first,
                         max_char,
                         fall_through,
                         odd_label,
                         even_label);
        return;
    }

    int new_start_index = 0;
    int new_end_index = 0;
    int border = 0;

    SplitSearchSpace(ranges,
                     start_index,
                     end_index,
                     &new_start_index,
                     &new_end_index,
                     &border);

    jit::Label handle_rest;
    jit::Label* above = &handle_rest;
    if (border == last + 1) {
        // We didn't find any section that started after the limit, so everything
        // above the border is one of the terminal labels.
        above = (end_index & 1) != (start_index & 1) ? odd_label : even_label;
        MOZ_ASSERT(new_end_index == end_index - 1);
    }

    MOZ_ASSERT(start_index <= new_end_index);
    MOZ_ASSERT(new_start_index <= end_index);
    MOZ_ASSERT(start_index < new_start_index);
    MOZ_ASSERT(new_end_index < end_index);
    MOZ_ASSERT(new_end_index + 1 == new_start_index ||
               (new_end_index + 2 == new_start_index &&
                border == ranges[new_end_index + 1]));
    MOZ_ASSERT(min_char < border - 1);
    MOZ_ASSERT(border < max_char);
    MOZ_ASSERT(ranges[new_end_index] < border);
    MOZ_ASSERT(border < ranges[new_start_index] ||
               (border == ranges[new_start_index] &&
                new_start_index == end_index &&
                new_end_index == end_index - 1 &&
                border == last + 1));
    MOZ_ASSERT(new_start_index == 0 || border >= ranges[new_start_index - 1]);

    masm->CheckCharacterGT(border - 1, above);
    jit::Label dummy;
    GenerateBranches(masm,
                     ranges,
                     start_index,
                     new_end_index,
                     min_char,
                     border - 1,
                     &dummy,
                     even_label,
                     odd_label);
    if (handle_rest.used()) {
        masm->Bind(&handle_rest);
        bool flip = (new_start_index & 1) != (start_index & 1);
        GenerateBranches(masm,
                         ranges,
                         new_start_index,
                         end_index,
                         border,
                         max_char,
                         &dummy,
                         flip ? odd_label : even_label,
                         flip ? even_label : odd_label);
    }
}

static void
EmitCharClass(LifoAlloc* alloc,
              RegExpMacroAssembler* macro_assembler,
              RegExpCharacterClass* cc,
              bool ascii,
              jit::Label* on_failure,
              int cp_offset,
              bool check_offset,
              bool preloaded)
{
    CharacterRangeVector& ranges = cc->ranges(alloc);
    if (!CharacterRange::IsCanonical(ranges)) {
        CharacterRange::Canonicalize(ranges);
    }

    int max_char = MaximumCharacter(ascii);
    int range_count = ranges.length();

    int last_valid_range = range_count - 1;
    while (last_valid_range >= 0) {
        CharacterRange& range = ranges[last_valid_range];
        if (range.from() <= max_char) {
            break;
        }
        last_valid_range--;
    }

    if (last_valid_range < 0) {
        if (!cc->is_negated()) {
            macro_assembler->JumpOrBacktrack(on_failure);
        }
        if (check_offset) {
            macro_assembler->CheckPosition(cp_offset, on_failure);
        }
        return;
    }

    if (last_valid_range == 0 &&
        ranges[0].IsEverything(max_char)) {
        if (cc->is_negated()) {
            macro_assembler->JumpOrBacktrack(on_failure);
        } else {
            // This is a common case hit by non-anchored expressions.
            if (check_offset) {
                macro_assembler->CheckPosition(cp_offset, on_failure);
            }
        }
        return;
    }
    if (last_valid_range == 0 &&
        !cc->is_negated() &&
        ranges[0].IsEverything(max_char)) {
        // This is a common case hit by non-anchored expressions.
        if (check_offset) {
            macro_assembler->CheckPosition(cp_offset, on_failure);
        }
        return;
    }

    if (!preloaded) {
        macro_assembler->LoadCurrentCharacter(cp_offset, on_failure, check_offset);
    }

    if (cc->is_standard(alloc) &&
        macro_assembler->CheckSpecialCharacterClass(cc->standard_type(),
                                                    on_failure)) {
        return;
    }

    // A new list with ascending entries.  Each entry is a code unit
    // where there is a boundary between code units that are part of
    // the class and code units that are not.  Normally we insert an
    // entry at zero which goes to the failure label, but if there
    // was already one there we fall through for success on that entry.
    // Subsequent entries have alternating meaning (success/failure).
    RangeBoundaryVector* range_boundaries =
        alloc->newInfallible<RangeBoundaryVector>(*alloc);

    bool zeroth_entry_is_failure = !cc->is_negated();

    range_boundaries->reserve(last_valid_range);
    for (int i = 0; i <= last_valid_range; i++) {
        CharacterRange& range = ranges[i];
        if (range.from() == 0) {
            MOZ_ASSERT(i == 0);
            zeroth_entry_is_failure = !zeroth_entry_is_failure;
        } else {
            range_boundaries->append(range.from());
        }
        range_boundaries->append(range.to() + 1);
    }
    int end_index = range_boundaries->length() - 1;
    if ((*range_boundaries)[end_index] > max_char)
        end_index--;

    jit::Label fall_through;
    GenerateBranches(macro_assembler,
                     *range_boundaries,
                     0,  // start_index.
                     end_index,
                     0,  // min_char.
                     max_char,
                     &fall_through,
                     zeroth_entry_is_failure ? &fall_through : on_failure,
                     zeroth_entry_is_failure ? on_failure : &fall_through);
    macro_assembler->Bind(&fall_through);
}

typedef bool EmitCharacterFunction(RegExpCompiler* compiler,
                                   char16_t c,
                                   jit::Label* on_failure,
                                   int cp_offset,
                                   bool check,
                                   bool preloaded);

static inline bool
EmitSimpleCharacter(RegExpCompiler* compiler,
                    char16_t c,
                    jit::Label* on_failure,
                    int cp_offset,
                    bool check,
                    bool preloaded)
{
    RegExpMacroAssembler* assembler = compiler->macro_assembler();
    bool bound_checked = false;
    if (!preloaded) {
        assembler->LoadCurrentCharacter(cp_offset, on_failure, check);
        bound_checked = true;
    }
    assembler->CheckNotCharacter(c, on_failure);
    return bound_checked;
}

// Only emits non-letters (things that don't have case).  Only used for case
// independent matches.
static inline bool
EmitAtomNonLetter(RegExpCompiler* compiler,
                  char16_t c,
                  jit::Label* on_failure,
                  int cp_offset,
                  bool check,
                  bool preloaded)
{
    RegExpMacroAssembler* macro_assembler = compiler->macro_assembler();
    bool ascii = compiler->ascii();
    char16_t chars[kEcma262UnCanonicalizeMaxWidth];
    int length = GetCaseIndependentLetters(c, ascii, compiler->unicode(), chars);
    if (length < 1) {
        // This can't match.  Must be an ASCII subject and a non-ASCII character.
        // We do not need to do anything since the ASCII pass already handled this.
        return false;  // Bounds not checked.
    }
    bool checked = false;
    // We handle the length > 1 case in a later pass.
    if (length == 1) {
        if (ascii && c > kMaxOneByteCharCode) {
            // Can't match - see above.
            return false;  // Bounds not checked.
        }
        if (!preloaded) {
            macro_assembler->LoadCurrentCharacter(cp_offset, on_failure, check);
            checked = check;
        }
        macro_assembler->CheckNotCharacter(c, on_failure);
    }
    return checked;
}

static bool
ShortCutEmitCharacterPair(RegExpMacroAssembler* macro_assembler,
                          bool ascii,
                          char16_t c1,
                          char16_t c2,
                          jit::Label* on_failure)
{
    char16_t char_mask = MaximumCharacter(ascii);

    MOZ_ASSERT(c1 != c2);
    if (c1 > c2) {
        char16_t tmp = c1;
        c1 = c2;
        c2 = tmp;
    }

    char16_t exor = c1 ^ c2;
    // Check whether exor has only one bit set.
    if (((exor - 1) & exor) == 0) {
        // If c1 and c2 differ only by one bit.
        char16_t mask = char_mask ^ exor;
        macro_assembler->CheckNotCharacterAfterAnd(c1, mask, on_failure);
        return true;
    }

    char16_t diff = c2 - c1;
    if (((diff - 1) & diff) == 0 && c1 >= diff) {
        // If the characters differ by 2^n but don't differ by one bit then
        // subtract the difference from the found character, then do the or
        // trick.  We avoid the theoretical case where negative numbers are
        // involved in order to simplify code generation.
        char16_t mask = char_mask ^ diff;
        macro_assembler->CheckNotCharacterAfterMinusAnd(c1 - diff,
                                                        diff,
                                                        mask,
                                                        on_failure);
        return true;
    }
    return false;
}

// Only emits letters (things that have case).  Only used for case independent
// matches.
static inline bool
EmitAtomLetter(RegExpCompiler* compiler,
               char16_t c,
               jit::Label* on_failure,
               int cp_offset,
               bool check,
               bool preloaded)
{
    RegExpMacroAssembler* macro_assembler = compiler->macro_assembler();
    bool ascii = compiler->ascii();
    char16_t chars[kEcma262UnCanonicalizeMaxWidth];
    int length = GetCaseIndependentLetters(c, ascii, compiler->unicode(), chars);
    if (length <= 1) return false;
    // We may not need to check against the end of the input string
    // if this character lies before a character that matched.
    if (!preloaded)
        macro_assembler->LoadCurrentCharacter(cp_offset, on_failure, check);
    jit::Label ok;
    MOZ_ASSERT(kEcma262UnCanonicalizeMaxWidth == 4);
    switch (length) {
      case 2: {
        if (ShortCutEmitCharacterPair(macro_assembler,
                                      ascii,
                                      chars[0],
                                      chars[1],
                                      on_failure)) {
        } else {
            macro_assembler->CheckCharacter(chars[0], &ok);
            macro_assembler->CheckNotCharacter(chars[1], on_failure);
            macro_assembler->Bind(&ok);
        }
        break;
      }
      case 4:
        macro_assembler->CheckCharacter(chars[3], &ok);
        [[fallthrough]];
      case 3:
        macro_assembler->CheckCharacter(chars[0], &ok);
        macro_assembler->CheckCharacter(chars[1], &ok);
        macro_assembler->CheckNotCharacter(chars[2], on_failure);
        macro_assembler->Bind(&ok);
        break;
      default:
        MOZ_CRASH("Bad length");
    }
    return true;
}

// We call this repeatedly to generate code for each pass over the text node.
// The passes are in increasing order of difficulty because we hope one
// of the first passes will fail in which case we are saved the work of the
// later passes.  for example for the case independent regexp /%[asdfghjkl]a/
// we will check the '%' in the first pass, the case independent 'a' in the
// second pass and the character class in the last pass.
//
// The passes are done from right to left, so for example to test for /bar/
// we will first test for an 'r' with offset 2, then an 'a' with offset 1
// and then a 'b' with offset 0.  This means we can avoid the end-of-input
// bounds check most of the time.  In the example we only need to check for
// end-of-input when loading the putative 'r'.
//
// A slight complication involves the fact that the first character may already
// be fetched into a register by the previous node.  In this case we want to
// do the test for that character first.  We do this in separate passes.  The
// 'preloaded' argument indicates that we are doing such a 'pass'.  If such a
// pass has been performed then subsequent passes will have true in
// first_element_checked to indicate that that character does not need to be
// checked again.
//
// In addition to all this we are passed a Trace, which can
// contain an AlternativeGeneration object.  In this AlternativeGeneration
// object we can see details of any quick check that was already passed in
// order to get to the code we are now generating.  The quick check can involve
// loading characters, which means we do not need to recheck the bounds
// up to the limit the quick check already checked.  In addition the quick
// check can have involved a mask and compare operation which may simplify
// or obviate the need for further checks at some character positions.
void
TextNode::TextEmitPass(RegExpCompiler* compiler,
                       TextEmitPassType pass,
                       bool preloaded,
                       Trace* trace,
                       bool first_element_checked,
                       int* checked_up_to)
{
    RegExpMacroAssembler* assembler = compiler->macro_assembler();
    bool ascii = compiler->ascii();
    jit::Label* backtrack = trace->backtrack();
    QuickCheckDetails* quick_check = trace->quick_check_performed();
    int element_count = elements().length();
    int backward_offset = read_backward() ? -Length() : 0;
    for (int i = preloaded ? 0 : element_count - 1; i >= 0; i--) {
        TextElement elm = elements()[i];
        int cp_offset = trace->cp_offset() + elm.cp_offset() + backward_offset;
        if (elm.text_type() == TextElement::ATOM) {
            const CharacterVector& quarks = elm.atom()->data();
            for (int j = preloaded ? 0 : quarks.length() - 1; j >= 0; j--) {
                if (first_element_checked && i == 0 && j == 0) continue;
                if (DeterminedAlready(quick_check, elm.cp_offset() + j)) continue;
                EmitCharacterFunction* emit_function = nullptr;
                switch (pass) {
                  case NON_ASCII_MATCH:
                    MOZ_ASSERT(ascii);
                    if (quarks[j] > kMaxOneByteCharCode) {
                        assembler->JumpOrBacktrack(backtrack);
                        return;
                    }
                    break;
                  case NON_LETTER_CHARACTER_MATCH:
                    emit_function = &EmitAtomNonLetter;
                    break;
                  case SIMPLE_CHARACTER_MATCH:
                    emit_function = &EmitSimpleCharacter;
                    break;
                  case CASE_CHARACTER_MATCH:
                    emit_function = &EmitAtomLetter;
                    break;
                  default:
                    break;
                }
                if (emit_function != nullptr) {
                    bool bounds_check = *checked_up_to < cp_offset + j || read_backward();
                    bool bound_checked = emit_function(compiler,
                                                       quarks[j],
                                                       backtrack,
                                                       cp_offset + j,
                                                       bounds_check,
                                                       preloaded);
                    if (bound_checked) UpdateBoundsCheck(cp_offset + j, checked_up_to);
                }
            }
        } else {
            MOZ_ASSERT(TextElement::CHAR_CLASS == elm.text_type());
            if (pass == CHARACTER_CLASS_MATCH) {
                if (first_element_checked && i == 0) continue;
                if (DeterminedAlready(quick_check, elm.cp_offset())) continue;
                RegExpCharacterClass* cc = elm.char_class();
                bool bounds_check = *checked_up_to < cp_offset || read_backward();
                EmitCharClass(alloc(),
                              assembler,
                              cc,
                              ascii,
                              backtrack,
                              cp_offset,
                              bounds_check,
                              preloaded);
                UpdateBoundsCheck(cp_offset, checked_up_to);
            }
        }
    }
}

int
TextNode::Length()
{
    TextElement elm = elements()[elements().length() - 1];
    MOZ_ASSERT(elm.cp_offset() >= 0);
    return elm.cp_offset() + elm.length();
}

bool
TextNode::SkipPass(int int_pass, bool ignore_case)
{
    TextEmitPassType pass = static_cast<TextEmitPassType>(int_pass);
    if (ignore_case)
        return pass == SIMPLE_CHARACTER_MATCH;
    return pass == NON_LETTER_CHARACTER_MATCH || pass == CASE_CHARACTER_MATCH;
}

// This generates the code to match a text node.  A text node can contain
// straight character sequences (possibly to be matched in a case-independent
// way) and character classes.  For efficiency we do not do this in a single
// pass from left to right.  Instead we pass over the text node several times,
// emitting code for some character positions every time.  See the comment on
// TextEmitPass for details.
void
TextNode::Emit(RegExpCompiler* compiler, Trace* trace)
{
    LimitResult limit_result = LimitVersions(compiler, trace);
    if (limit_result == DONE) return;
    MOZ_ASSERT(limit_result == CONTINUE);

    if (trace->cp_offset() + Length() > RegExpMacroAssembler::kMaxCPOffset) {
        compiler->SetRegExpTooBig();
        return;
    }

    if (compiler->ascii()) {
        int dummy = 0;
        TextEmitPass(compiler, NON_ASCII_MATCH, false, trace, false, &dummy);
    }

    bool first_elt_done = false;
    int bound_checked_to = trace->cp_offset() - 1;
    bound_checked_to += trace->bound_checked_up_to();

    // If a character is preloaded into the current character register then
    // check that now.
    if (trace->characters_preloaded() == 1) {
        for (int pass = kFirstRealPass; pass <= kLastPass; pass++) {
            if (!SkipPass(pass, compiler->ignore_case())) {
                TextEmitPass(compiler,
                             static_cast<TextEmitPassType>(pass),
                             true,
                             trace,
                             false,
                             &bound_checked_to);
            }
        }
        first_elt_done = true;
    }

    for (int pass = kFirstRealPass; pass <= kLastPass; pass++) {
        if (!SkipPass(pass, compiler->ignore_case())) {
            TextEmitPass(compiler,
                         static_cast<TextEmitPassType>(pass),
                         false,
                         trace,
                         first_elt_done,
                         &bound_checked_to);
        }
    }

    Trace successor_trace(*trace);
    // If we advance backward, we may end up at the start.
    successor_trace.AdvanceCurrentPositionInTrace(
        read_backward() ? -Length() : Length(), compiler);
    successor_trace.set_at_start(read_backward() ? Trace::UNKNOWN
                                                 : Trace::FALSE_VALUE);
    RecursionCheck rc(compiler);
    on_success()->Emit(compiler, &successor_trace);
}

void
LoopChoiceNode::Emit(RegExpCompiler* compiler, Trace* trace)
{
    RegExpMacroAssembler* macro_assembler = compiler->macro_assembler();
    if (trace->stop_node() == this) {
        int text_length =
            GreedyLoopTextLengthForAlternative(&alternatives()[0]);
        MOZ_ASSERT(text_length != kNodeIsTooComplexForGreedyLoops);
        // Update the counter-based backtracking info on the stack.  This is an
        // optimization for greedy loops (see below).
        MOZ_ASSERT(trace->cp_offset() == text_length);
        macro_assembler->AdvanceCurrentPosition(text_length);
        macro_assembler->JumpOrBacktrack(trace->loop_label());
        return;
    }
    MOZ_ASSERT(trace->stop_node() == nullptr);
    if (!trace->is_trivial()) {
        trace->Flush(compiler, this);
        return;
    }
    ChoiceNode::Emit(compiler, trace);
}

/* Code generation for choice nodes.
 *
 * We generate quick checks that do a mask and compare to eliminate a
 * choice.  If the quick check succeeds then it jumps to the continuation to
 * do slow checks and check subsequent nodes.  If it fails (the common case)
 * it falls through to the next choice.
 *
 * Here is the desired flow graph.  Nodes directly below each other imply
 * fallthrough.  Alternatives 1 and 2 have quick checks.  Alternative
 * 3 doesn't have a quick check so we have to call the slow check.
 * Nodes are marked Qn for quick checks and Sn for slow checks.  The entire
 * regexp continuation is generated directly after the Sn node, up to the
 * next JumpOrBacktrack if we decide to reuse some already generated code.  Some
 * nodes expect preload_characters to be preloaded into the current
 * character register.  R nodes do this preloading.  Vertices are marked
 * F for failures and S for success (possible success in the case of quick
 * nodes).  L, V, < and > are used as arrow heads.
 *
 * ----------> R
 *             |
 *             V
 *            Q1 -----> S1
 *             |   S   /
 *            F|      /
 *             |    F/
 *             |    /
 *             |   R
 *             |  /
 *             V L
 *            Q2 -----> S2
 *             |   S   /
 *            F|      /
 *             |    F/
 *             |    /
 *             |   R
 *             |  /
 *             V L
 *            S3
 *             |
 *            F|
 *             |
 *             R
 *             |
 * backtrack   V
 * <----------Q4
 *   \    F    |
 *    \        |S
 *     \   F   V
 *      \-----S4
 *
 * For greedy loops we reverse our expectation and expect to match rather
 * than fail. Therefore we want the loop code to look like this (U is the
 * unwind code that steps back in the greedy loop).  The following alternatives
 * look the same as above.
 *              _____
 *             /     \
 *             V     |
 * ----------> S1    |
 *            /|     |
 *           / |S    |
 *         F/  \_____/
 *         /
 *        |<-----------
 *        |            \
 *        V             \
 *        Q2 ---> S2     \
 *        |  S   /       |
 *       F|     /        |
 *        |   F/         |
 *        |   /          |
 *        |  R           |
 *        | /            |
 *   F    VL             |
 * <------U              |
 * back   |S             |
 *        \______________/
 */

// This class is used when generating the alternatives in a choice node.  It
// records the way the alternative is being code generated.
class irregexp::AlternativeGeneration
{
  public:
    AlternativeGeneration()
      : possible_success(),
        expects_preload(false),
        after(),
        quick_check_details()
    {}

    jit::Label possible_success;
    bool expects_preload;
    jit::Label after;
    QuickCheckDetails quick_check_details;
};

void
ChoiceNode::GenerateGuard(RegExpMacroAssembler* macro_assembler,
                          Guard* guard, Trace* trace)
{
    switch (guard->op()) {
      case Guard::LT:
        MOZ_ASSERT(!trace->mentions_reg(guard->reg()));
        macro_assembler->IfRegisterGE(guard->reg(),
                                      guard->value(),
                                      trace->backtrack());
        break;
      case Guard::GEQ:
        MOZ_ASSERT(!trace->mentions_reg(guard->reg()));
        macro_assembler->IfRegisterLT(guard->reg(),
                                      guard->value(),
                                      trace->backtrack());
        break;
    }
}

int
ChoiceNode::CalculatePreloadCharacters(RegExpCompiler* compiler, int eats_at_least)
{
    int preload_characters = Min(4, eats_at_least);
    if (compiler->macro_assembler()->CanReadUnaligned()) {
        bool ascii = compiler->ascii();
        if (ascii) {
            if (preload_characters > 4)
                preload_characters = 4;
            // We can't preload 3 characters because there is no machine instruction
            // to do that.  We can't just load 4 because we could be reading
            // beyond the end of the string, which could cause a memory fault.
            if (preload_characters == 3)
                preload_characters = 2;
        } else {
            if (preload_characters > 2)
                preload_characters = 2;
        }
    } else {
        if (preload_characters > 1)
            preload_characters = 1;
    }
    return preload_characters;
}

RegExpNode*
TextNode::GetSuccessorOfOmnivorousTextNode(RegExpCompiler* compiler)
{
    if (read_backward()) return nullptr;

    if (elements().length() != 1)
        return nullptr;

    TextElement elm = elements()[0];
    if (elm.text_type() != TextElement::CHAR_CLASS)
        return nullptr;

    RegExpCharacterClass* node = elm.char_class();
    CharacterRangeVector& ranges = node->ranges(alloc());

    if (!CharacterRange::IsCanonical(ranges))
        CharacterRange::Canonicalize(ranges);

    if (node->is_negated())
        return ranges.length() == 0 ? on_success() : nullptr;

    if (ranges.length() != 1)
        return nullptr;

    uint32_t max_char = MaximumCharacter(compiler->ascii());
    return ranges[0].IsEverything(max_char) ? on_success() : nullptr;
}

// Finds the fixed match length of a sequence of nodes that goes from
// this alternative and back to this choice node.  If there are variable
// length nodes or other complications in the way then return a sentinel
// value indicating that a greedy loop cannot be constructed.
int
ChoiceNode::GreedyLoopTextLengthForAlternative(GuardedAlternative* alternative)
{
    int length = 0;
    RegExpNode* node = alternative->node();
    // Later we will generate code for all these text nodes using recursion
    // so we have to limit the max number.
    int recursion_depth = 0;
    while (node != this) {
        if (recursion_depth++ > RegExpCompiler::kMaxRecursion) {
            return kNodeIsTooComplexForGreedyLoops;
        }
        int node_length = node->GreedyLoopTextLength();
        if (node_length == kNodeIsTooComplexForGreedyLoops) {
            return kNodeIsTooComplexForGreedyLoops;
        }
        length += node_length;
        SeqRegExpNode* seq_node = static_cast<SeqRegExpNode*>(node);
        node = seq_node->on_success();
    }
    return read_backward() ? -length : length;
}

// Creates a list of AlternativeGenerations.  If the list has a reasonable
// size then it is on the stack, otherwise the excess is on the heap.
class AlternativeGenerationList
{
  public:
    AlternativeGenerationList(LifoAlloc* alloc, size_t count)
      : alt_gens_(*alloc)
    {
        alt_gens_.reserve(count);
        for (size_t i = 0; i < count && i < kAFew; i++)
            alt_gens_.append(a_few_alt_gens_ + i);
        for (size_t i = kAFew; i < count; i++) {
            AutoEnterOOMUnsafeRegion oomUnsafe;
            AlternativeGeneration* gen = js_new<AlternativeGeneration>();
            if (!gen)
                oomUnsafe.crash("AlternativeGenerationList js_new");
            alt_gens_.append(gen);
        }
    }

    ~AlternativeGenerationList() {
        for (size_t i = kAFew; i < alt_gens_.length(); i++) {
            js_delete(alt_gens_[i]);
            alt_gens_[i] = nullptr;
        }
    }

    AlternativeGeneration* at(int i) {
        return alt_gens_[i];
    }

  private:
    static const size_t kAFew = 10;
    InfallibleVector<AlternativeGeneration*, 1> alt_gens_;
    AlternativeGeneration a_few_alt_gens_[kAFew];
};

void
ChoiceNode::Emit(RegExpCompiler* compiler, Trace* trace)
{
    RegExpMacroAssembler* macro_assembler = compiler->macro_assembler();
    size_t choice_count = alternatives().length();
#ifdef DEBUG
    for (size_t i = 0; i < choice_count - 1; i++) {
        const GuardedAlternative& alternative = alternatives()[i];
        const GuardVector* guards = alternative.guards();
        if (guards) {
            for (size_t j = 0; j < guards->length(); j++)
                MOZ_ASSERT(!trace->mentions_reg((*guards)[j]->reg()));
        }
    }
#endif

    LimitResult limit_result = LimitVersions(compiler, trace);
    if (limit_result == DONE) return;
    MOZ_ASSERT(limit_result == CONTINUE);

    int new_flush_budget = trace->flush_budget() / choice_count;
    if (trace->flush_budget() == 0 && trace->actions() != nullptr) {
        trace->Flush(compiler, this);
        return;
    }

    RecursionCheck rc(compiler);

    Trace* current_trace = trace;

    int text_length = GreedyLoopTextLengthForAlternative(&alternatives()[0]);
    bool greedy_loop = false;
    jit::Label greedy_loop_label;
    Trace counter_backtrack_trace;
    counter_backtrack_trace.set_backtrack(&greedy_loop_label);
    if (not_at_start()) counter_backtrack_trace.set_at_start(Trace::FALSE_VALUE);

    if (choice_count > 1 && text_length != kNodeIsTooComplexForGreedyLoops) {
        // Here we have special handling for greedy loops containing only text nodes
        // and other simple nodes.  These are handled by pushing the current
        // position on the stack and then incrementing the current position each
        // time around the switch.  On backtrack we decrement the current position
        // and check it against the pushed value.  This avoids pushing backtrack
        // information for each iteration of the loop, which could take up a lot of
        // space.
        greedy_loop = true;
        MOZ_ASSERT(trace->stop_node() == nullptr);
        macro_assembler->PushCurrentPosition();
        current_trace = &counter_backtrack_trace;
        jit::Label greedy_match_failed;
        Trace greedy_match_trace;
        if (not_at_start()) greedy_match_trace.set_at_start(Trace::FALSE_VALUE);
        greedy_match_trace.set_backtrack(&greedy_match_failed);
        jit::Label loop_label;
        macro_assembler->Bind(&loop_label);
        greedy_match_trace.set_stop_node(this);
        greedy_match_trace.set_loop_label(&loop_label);
        alternatives()[0].node()->Emit(compiler, &greedy_match_trace);
        macro_assembler->Bind(&greedy_match_failed);
    }

    jit::Label second_choice;  // For use in greedy matches.
    macro_assembler->Bind(&second_choice);

    size_t first_normal_choice = greedy_loop ? 1 : 0;

    bool not_at_start = current_trace->at_start() == Trace::FALSE_VALUE;
    const int kEatsAtLeastNotYetInitialized = -1;
    int eats_at_least = kEatsAtLeastNotYetInitialized;

    bool skip_was_emitted = false;

    if (!greedy_loop && choice_count == 2) {
        GuardedAlternative alt1 = alternatives()[1];
        if (!alt1.guards() || alt1.guards()->length() == 0) {
            RegExpNode* eats_anything_node = alt1.node();
            if (eats_anything_node->GetSuccessorOfOmnivorousTextNode(compiler) == this) {
                // At this point we know that we are at a non-greedy loop that will eat
                // any character one at a time.  Any non-anchored regexp has such a
                // loop prepended to it in order to find where it starts.  We look for
                // a pattern of the form ...abc... where we can look 6 characters ahead
                // and step forwards 3 if the character is not one of abc.  Abc need
                // not be atoms, they can be any reasonably limited character class or
                // small alternation.
                MOZ_ASSERT(trace->is_trivial());  // This is the case on LoopChoiceNodes.
                BoyerMooreLookahead* lookahead = bm_info(not_at_start);
                if (lookahead == nullptr) {
                    eats_at_least = Min(kMaxLookaheadForBoyerMoore,
                                        EatsAtLeast(kMaxLookaheadForBoyerMoore,
                                                    kRecursionBudget,
                                                    not_at_start));
                    if (eats_at_least >= 1) {
                        BoyerMooreLookahead* bm =
                            alloc()->newInfallible<BoyerMooreLookahead>(alloc(), eats_at_least, compiler);
                        GuardedAlternative alt0 = alternatives()[0];
                        alt0.node()->FillInBMInfo(0, kRecursionBudget, bm, not_at_start);
                        skip_was_emitted = bm->EmitSkipInstructions(macro_assembler);
                    }
                } else {
                    skip_was_emitted = lookahead->EmitSkipInstructions(macro_assembler);
                }
            }
        }
    }

    if (eats_at_least == kEatsAtLeastNotYetInitialized) {
        // Save some time by looking at most one machine word ahead.
        eats_at_least =
            EatsAtLeast(compiler->ascii() ? 4 : 2, kRecursionBudget, not_at_start);
    }
    int preload_characters = CalculatePreloadCharacters(compiler, eats_at_least);

    bool preload_is_current = !skip_was_emitted &&
        (current_trace->characters_preloaded() == preload_characters);
    bool preload_has_checked_bounds = preload_is_current;

    AlternativeGenerationList alt_gens(alloc(), choice_count);

    // For now we just call all choices one after the other.  The idea ultimately
    // is to use the Dispatch table to try only the relevant ones.
    for (size_t i = first_normal_choice; i < choice_count; i++) {
        bool is_last = i == choice_count - 1;
        bool fall_through_on_failure = !is_last;
        GuardedAlternative alternative = alternatives()[i];
        AlternativeGeneration* alt_gen = alt_gens.at(i);
        alt_gen->quick_check_details.set_characters(preload_characters);
        const GuardVector* guards = alternative.guards();
        Trace new_trace(*current_trace);
        new_trace.set_characters_preloaded(preload_is_current ?
                                           preload_characters :
                                           0);
        if (preload_has_checked_bounds) {
            new_trace.set_bound_checked_up_to(preload_characters);
        }
        new_trace.quick_check_performed()->Clear();
        if (not_at_start_) new_trace.set_at_start(Trace::FALSE_VALUE);
        alt_gen->expects_preload = preload_is_current;
        bool generate_full_check_inline = false;
        if (try_to_emit_quick_check_for_alternative(i == 0) &&
            alternative.node()->EmitQuickCheck(compiler,
                                               &new_trace,
                                               preload_has_checked_bounds,
                                               &alt_gen->possible_success,
                                               &alt_gen->quick_check_details,
                                               fall_through_on_failure)) {
            // Quick check was generated for this choice.
            preload_is_current = true;
            preload_has_checked_bounds = true;
            // On the last choice in the ChoiceNode we generated the quick
            // check to fall through on possible success.  So now we need to
            // generate the full check inline.
            if (!fall_through_on_failure) {
                macro_assembler->Bind(&alt_gen->possible_success);
                new_trace.set_quick_check_performed(&alt_gen->quick_check_details);
                new_trace.set_characters_preloaded(preload_characters);
                new_trace.set_bound_checked_up_to(preload_characters);
                generate_full_check_inline = true;
            }
        } else if (alt_gen->quick_check_details.cannot_match()) {
            if (!fall_through_on_failure && !greedy_loop) {
                macro_assembler->JumpOrBacktrack(trace->backtrack());
            }
            continue;
        } else {
            // No quick check was generated.  Put the full code here.
            // If this is not the first choice then there could be slow checks from
            // previous cases that go here when they fail.  There's no reason to
            // insist that they preload characters since the slow check we are about
            // to generate probably can't use it.
            if (i != first_normal_choice) {
                alt_gen->expects_preload = false;
                new_trace.InvalidateCurrentCharacter();
            }
            if (!is_last) {
                new_trace.set_backtrack(&alt_gen->after);
            }
            generate_full_check_inline = true;
        }
        if (generate_full_check_inline) {
            if (new_trace.actions() != nullptr)
                new_trace.set_flush_budget(new_flush_budget);
            if (guards) {
                for (size_t j = 0; j < guards->length(); j++)
                    GenerateGuard(macro_assembler, (*guards)[j], &new_trace);
            }
            alternative.node()->Emit(compiler, &new_trace);
            preload_is_current = false;
        }
        macro_assembler->Bind(&alt_gen->after);
    }
    if (greedy_loop) {
        macro_assembler->Bind(&greedy_loop_label);
        // If we have unwound to the bottom then backtrack.
        macro_assembler->CheckGreedyLoop(trace->backtrack());
        // Otherwise try the second priority at an earlier position.
        macro_assembler->AdvanceCurrentPosition(-text_length);
        macro_assembler->JumpOrBacktrack(&second_choice);
    }

    // At this point we need to generate slow checks for the alternatives where
    // the quick check was inlined.  We can recognize these because the associated
    // label was bound.
    for (size_t i = first_normal_choice; i < choice_count - 1; i++) {
        AlternativeGeneration* alt_gen = alt_gens.at(i);
        Trace new_trace(*current_trace);
        // If there are actions to be flushed we have to limit how many times
        // they are flushed.  Take the budget of the parent trace and distribute
        // it fairly amongst the children.
        if (new_trace.actions() != nullptr) {
            new_trace.set_flush_budget(new_flush_budget);
        }
        bool next_expects_preload =
            i == choice_count - 1 ? false : alt_gens.at(i + 1)->expects_preload;
        EmitOutOfLineContinuation(compiler,
                                  &new_trace,
                                  alternatives()[i],
                                  alt_gen,
                                  preload_characters,
                                  next_expects_preload);
    }
}

void
ChoiceNode::EmitOutOfLineContinuation(RegExpCompiler* compiler,
                                      Trace* trace,
                                      GuardedAlternative alternative,
                                      AlternativeGeneration* alt_gen,
                                      int preload_characters,
                                      bool next_expects_preload)
{
    if (!alt_gen->possible_success.used())
        return;

    RegExpMacroAssembler* macro_assembler = compiler->macro_assembler();
    macro_assembler->Bind(&alt_gen->possible_success);
    Trace out_of_line_trace(*trace);
    out_of_line_trace.set_characters_preloaded(preload_characters);
    out_of_line_trace.set_quick_check_performed(&alt_gen->quick_check_details);
    if (not_at_start_) out_of_line_trace.set_at_start(Trace::FALSE_VALUE);
    const GuardVector* guards = alternative.guards();
    if (next_expects_preload) {
        jit::Label reload_current_char;
        out_of_line_trace.set_backtrack(&reload_current_char);
        if (guards) {
            for (size_t j = 0; j < guards->length(); j++)
                GenerateGuard(macro_assembler, (*guards)[j], &out_of_line_trace);
        }
        alternative.node()->Emit(compiler, &out_of_line_trace);
        macro_assembler->Bind(&reload_current_char);
        // Reload the current character, since the next quick check expects that.
        // We don't need to check bounds here because we only get into this
        // code through a quick check which already did the checked load.
        macro_assembler->LoadCurrentCharacter(trace->cp_offset(),
                                              nullptr,
                                              false,
                                              preload_characters);
        macro_assembler->JumpOrBacktrack(&(alt_gen->after));
    } else {
        out_of_line_trace.set_backtrack(&(alt_gen->after));
        if (guards) {
            for (size_t j = 0; j < guards->length(); j++)
                GenerateGuard(macro_assembler, (*guards)[j], &out_of_line_trace);
        }
        alternative.node()->Emit(compiler, &out_of_line_trace);
    }
}

void
ActionNode::Emit(RegExpCompiler* compiler, Trace* trace)
{
    RegExpMacroAssembler* assembler = compiler->macro_assembler();
    LimitResult limit_result = LimitVersions(compiler, trace);
    if (limit_result == DONE) return;
    MOZ_ASSERT(limit_result == CONTINUE);

    RecursionCheck rc(compiler);

    switch (action_type_) {
      case STORE_POSITION: {
        Trace::DeferredCapture
            new_capture(data_.u_position_register.reg,
                        data_.u_position_register.is_capture,
                        trace);
        Trace new_trace = *trace;
        new_trace.add_action(&new_capture);
        on_success()->Emit(compiler, &new_trace);
        break;
      }
      case INCREMENT_REGISTER: {
        Trace::DeferredIncrementRegister
            new_increment(data_.u_increment_register.reg);
        Trace new_trace = *trace;
        new_trace.add_action(&new_increment);
        on_success()->Emit(compiler, &new_trace);
        break;
      }
      case SET_REGISTER: {
        Trace::DeferredSetRegister
            new_set(data_.u_store_register.reg, data_.u_store_register.value);
        Trace new_trace = *trace;
        new_trace.add_action(&new_set);
        on_success()->Emit(compiler, &new_trace);
        break;
      }
      case CLEAR_CAPTURES: {
        Trace::DeferredClearCaptures
            new_capture(Interval(data_.u_clear_captures.range_from,
                                 data_.u_clear_captures.range_to));
        Trace new_trace = *trace;
        new_trace.add_action(&new_capture);
        on_success()->Emit(compiler, &new_trace);
        break;
      }
      case BEGIN_SUBMATCH:
        if (!trace->is_trivial()) {
            trace->Flush(compiler, this);
        } else {
            assembler->WriteCurrentPositionToRegister(data_.u_submatch.current_position_register, 0);
            assembler->WriteBacktrackStackPointerToRegister(data_.u_submatch.stack_pointer_register);
            on_success()->Emit(compiler, trace);
        }
        break;
      case EMPTY_MATCH_CHECK: {
        int start_pos_reg = data_.u_empty_match_check.start_register;
        int stored_pos = 0;
        int rep_reg = data_.u_empty_match_check.repetition_register;
        bool has_minimum = (rep_reg != RegExpCompiler::kNoRegister);
        bool know_dist = trace->GetStoredPosition(start_pos_reg, &stored_pos);
        if (know_dist && !has_minimum && stored_pos == trace->cp_offset()) {
            // If we know we haven't advanced and there is no minimum we
            // can just backtrack immediately.
            assembler->JumpOrBacktrack(trace->backtrack());
        } else if (know_dist && stored_pos < trace->cp_offset()) {
            // If we know we've advanced we can generate the continuation
            // immediately.
            on_success()->Emit(compiler, trace);
        } else if (!trace->is_trivial()) {
            trace->Flush(compiler, this);
        } else {
            jit::Label skip_empty_check;
            // If we have a minimum number of repetitions we check the current
            // number first and skip the empty check if it's not enough.
            if (has_minimum) {
                int limit = data_.u_empty_match_check.repetition_limit;
                assembler->IfRegisterLT(rep_reg, limit, &skip_empty_check);
            }
            // If the match is empty we bail out, otherwise we fall through
            // to the on-success continuation.
            assembler->IfRegisterEqPos(data_.u_empty_match_check.start_register,
                                       trace->backtrack());
            assembler->Bind(&skip_empty_check);
            on_success()->Emit(compiler, trace);
        }
        break;
      }
      case POSITIVE_SUBMATCH_SUCCESS: {
        if (!trace->is_trivial()) {
            trace->Flush(compiler, this);
            return;
        }
        assembler->ReadCurrentPositionFromRegister(data_.u_submatch.current_position_register);
        assembler->ReadBacktrackStackPointerFromRegister(data_.u_submatch.stack_pointer_register);
        int clear_register_count = data_.u_submatch.clear_register_count;
        if (clear_register_count == 0) {
            on_success()->Emit(compiler, trace);
            return;
        }
        int clear_registers_from = data_.u_submatch.clear_register_from;
        jit::Label clear_registers_backtrack;
        Trace new_trace = *trace;
        new_trace.set_backtrack(&clear_registers_backtrack);
        on_success()->Emit(compiler, &new_trace);

        assembler->Bind(&clear_registers_backtrack);
        int clear_registers_to = clear_registers_from + clear_register_count - 1;
        assembler->ClearRegisters(clear_registers_from, clear_registers_to);

        MOZ_ASSERT(trace->backtrack() == nullptr);
        assembler->Backtrack();
        return;
      }
      default:
        MOZ_CRASH("Bad action");
    }
}

void
BackReferenceNode::Emit(RegExpCompiler* compiler, Trace* trace)
{
    RegExpMacroAssembler* assembler = compiler->macro_assembler();
    if (!trace->is_trivial()) {
        trace->Flush(compiler, this);
        return;
    }

    LimitResult limit_result = LimitVersions(compiler, trace);
    if (limit_result == DONE) return;
    MOZ_ASSERT(limit_result == CONTINUE);

    RecursionCheck rc(compiler);

    MOZ_ASSERT(start_reg_ + 1 == end_reg_);
    if (compiler->ignore_case()) {
        assembler->CheckNotBackReferenceIgnoreCase(start_reg_,
                                                   read_backward(),
                                                   trace->backtrack(),
                                                   compiler->unicode());
    } else {
        assembler->CheckNotBackReference(start_reg_, read_backward(), trace->backtrack());
    }
    // We are going to advance backward, so we may end up at the start.
    if (read_backward()) trace->set_at_start(Trace::UNKNOWN);
    on_success()->Emit(compiler, trace);
}

RegExpNode::LimitResult
RegExpNode::LimitVersions(RegExpCompiler* compiler, Trace* trace)
{
    // If we are generating a greedy loop then don't stop and don't reuse code.
    if (trace->stop_node() != nullptr)
        return CONTINUE;

    RegExpMacroAssembler* macro_assembler = compiler->macro_assembler();
    if (trace->is_trivial()) {
        if (label()->bound()) {
            // We are being asked to generate a generic version, but that's already
            // been done so just go to it.
            macro_assembler->JumpOrBacktrack(label());
            return DONE;
        }
        if (compiler->recursion_depth() >= RegExpCompiler::kMaxRecursion) {
            // To avoid too deep recursion we push the node to the work queue and just
            // generate a goto here.
            compiler->AddWork(this);
            macro_assembler->JumpOrBacktrack(label());
            return DONE;
        }
        // Generate generic version of the node and bind the label for later use.
        macro_assembler->Bind(label());
        return CONTINUE;
    }

    // We are being asked to make a non-generic version.  Keep track of how many
    // non-generic versions we generate so as not to overdo it.
    trace_count_++;
    if (trace_count_ < kMaxCopiesCodeGenerated &&
        compiler->recursion_depth() <= RegExpCompiler::kMaxRecursion) {
        return CONTINUE;
    }

    // If we get here code has been generated for this node too many times or
    // recursion is too deep.  Time to switch to a generic version.  The code for
    // generic versions above can handle deep recursion properly.
    trace->Flush(compiler, this);
    return DONE;
}

bool
RegExpNode::EmitQuickCheck(RegExpCompiler* compiler,
                           Trace* trace,
                           bool preload_has_checked_bounds,
                           jit::Label* on_possible_success,
                           QuickCheckDetails* details,
                           bool fall_through_on_failure)
{
    if (details->characters() == 0) return false;
    GetQuickCheckDetails(
                         details, compiler, 0, trace->at_start() == Trace::FALSE_VALUE);
    if (details->cannot_match()) return false;
    if (!details->Rationalize(compiler->ascii())) return false;
    MOZ_ASSERT(details->characters() == 1 ||
               compiler->macro_assembler()->CanReadUnaligned());
    uint32_t mask = details->mask();
    uint32_t value = details->value();

    RegExpMacroAssembler* assembler = compiler->macro_assembler();

    if (trace->characters_preloaded() != details->characters()) {
        assembler->LoadCurrentCharacter(trace->cp_offset(),
                                        trace->backtrack(),
                                        !preload_has_checked_bounds,
                                        details->characters());
    }

    bool need_mask = true;

    if (details->characters() == 1) {
        // If number of characters preloaded is 1 then we used a byte or 16 bit
        // load so the value is already masked down.
        uint32_t char_mask = MaximumCharacter(compiler->ascii());
        if ((mask & char_mask) == char_mask) need_mask = false;
        mask &= char_mask;
    } else {
        // For 2-character preloads in ASCII mode or 1-character preloads in
        // TWO_BYTE mode we also use a 16 bit load with zero extend.
        if (details->characters() == 2 && compiler->ascii()) {
            if ((mask & 0xffff) == 0xffff) need_mask = false;
        } else if (details->characters() == 1 && !compiler->ascii()) {
            if ((mask & 0xffff) == 0xffff) need_mask = false;
        } else {
            if (mask == 0xffffffff) need_mask = false;
        }
    }

    if (fall_through_on_failure) {
        if (need_mask) {
            assembler->CheckCharacterAfterAnd(value, mask, on_possible_success);
        } else {
            assembler->CheckCharacter(value, on_possible_success);
        }
    } else {
        if (need_mask) {
            assembler->CheckNotCharacterAfterAnd(value, mask, trace->backtrack());
        } else {
            assembler->CheckNotCharacter(value, trace->backtrack());
        }
    }
    return true;
}

bool
TextNode::FillInBMInfo(int initial_offset,
                       int budget,
                       BoyerMooreLookahead* bm,
                       bool not_at_start)
{
    if (!bm->CheckOverRecursed())
        return false;

    if (initial_offset >= bm->length())
        return true;

    int offset = initial_offset;
    int max_char = bm->max_char();
    for (size_t i = 0; i < elements().length(); i++) {
        if (offset >= bm->length()) {
            if (initial_offset == 0)
                set_bm_info(not_at_start, bm);
            return true;
        }
        TextElement text = elements()[i];
        if (text.text_type() == TextElement::ATOM) {
            RegExpAtom* atom = text.atom();
            for (int j = 0; j < atom->length(); j++, offset++) {
                if (offset >= bm->length()) {
                    if (initial_offset == 0)
                        set_bm_info(not_at_start, bm);
                    return true;
                }
                char16_t character = atom->data()[j];
                if (bm->compiler()->ignore_case()) {
                    char16_t chars[kEcma262UnCanonicalizeMaxWidth];
                    int length = GetCaseIndependentLetters(character,
                                                           bm->max_char() == kMaxOneByteCharCode,
                                                           bm->compiler()->unicode(),
                                                           chars);
                    for (int j = 0; j < length; j++)
                        bm->Set(offset, chars[j]);
                } else {
                    if (character <= max_char) bm->Set(offset, character);
                }
            }
        } else {
            MOZ_ASSERT(TextElement::CHAR_CLASS == text.text_type());
            RegExpCharacterClass* char_class = text.char_class();
            const CharacterRangeVector& ranges = char_class->ranges(alloc());
            if (char_class->is_negated()) {
                bm->SetAll(offset);
            } else {
                for (size_t k = 0; k < ranges.length(); k++) {
                    const CharacterRange& range = ranges[k];
                    if (range.from() > max_char)
                        continue;
                    int to = Min(max_char, static_cast<int>(range.to()));
                    bm->SetInterval(offset, Interval(range.from(), to));
                }
            }
            offset++;
        }
    }
    if (offset >= bm->length()) {
        if (initial_offset == 0) set_bm_info(not_at_start, bm);
        return true;
    }
    if (!on_success()->FillInBMInfo(offset,
                                    budget - 1,
                                    bm,
                                    true))   // Not at start after a text node.
        return false;
    if (initial_offset == 0)
        set_bm_info(not_at_start, bm);
    return true;
}

// -------------------------------------------------------------------
// QuickCheckDetails

// Takes the left-most 1-bit and smears it out, setting all bits to its right.
static inline uint32_t
SmearBitsRight(uint32_t v)
{
    v |= v >> 1;
    v |= v >> 2;
    v |= v >> 4;
    v |= v >> 8;
    v |= v >> 16;
    return v;
}

// Here is the meat of GetQuickCheckDetails (see also the comment on the
// super-class in the .h file).
//
// We iterate along the text object, building up for each character a
// mask and value that can be used to test for a quick failure to match.
// The masks and values for the positions will be combined into a single
// machine word for the current character width in order to be used in
// generating a quick check.
void
TextNode::GetQuickCheckDetails(QuickCheckDetails* details,
                               RegExpCompiler* compiler,
                               int characters_filled_in,
                               bool not_at_start)
{
    // Do not collect any quick check details if the text node reads backward,
    // since it reads in the opposite direction than we use for quick checks.
    if (read_backward()) return;
    MOZ_ASSERT(characters_filled_in < details->characters());
    int characters = details->characters();
    int char_mask = MaximumCharacter(compiler->ascii());

    for (size_t k = 0; k < elements().length(); k++) {
        TextElement elm = elements()[k];
        if (elm.text_type() == TextElement::ATOM) {
            const CharacterVector& quarks = elm.atom()->data();
            for (size_t i = 0; i < (size_t) characters && i < quarks.length(); i++) {
                QuickCheckDetails::Position* pos =
                    details->positions(characters_filled_in);
                char16_t c = quarks[i];
                if (c > char_mask) {
                    // If we expect a non-ASCII character from an ASCII string,
                    // there is no way we can match. Not even case independent
                    // matching can turn an ASCII character into non-ASCII or
                    // vice versa.
                    details->set_cannot_match();
                    pos->determines_perfectly = false;
                    return;
                }
                if (compiler->ignore_case()) {
                    char16_t chars[kEcma262UnCanonicalizeMaxWidth];
                    size_t length = GetCaseIndependentLetters(c, compiler->ascii(),
                                                              compiler->unicode(), chars);
                    MOZ_ASSERT(length != 0);  // Can only happen if c > char_mask (see above).
                    if (length == 1) {
                        // This letter has no case equivalents, so it's nice and simple
                        // and the mask-compare will determine definitely whether we have
                        // a match at this character position.
                        pos->mask = char_mask;
                        pos->value = c;
                        pos->determines_perfectly = true;
                    } else {
                        uint32_t common_bits = char_mask;
                        uint32_t bits = chars[0];
                        for (size_t j = 1; j < length; j++) {
                            uint32_t differing_bits = ((chars[j] & common_bits) ^ bits);
                            common_bits ^= differing_bits;
                            bits &= common_bits;
                        }
                        // If length is 2 and common bits has only one zero in it then
                        // our mask and compare instruction will determine definitely
                        // whether we have a match at this character position.  Otherwise
                        // it can only be an approximate check.
                        uint32_t one_zero = (common_bits | ~char_mask);
                        if (length == 2 && ((~one_zero) & ((~one_zero) - 1)) == 0) {
                            pos->determines_perfectly = true;
                        }
                        pos->mask = common_bits;
                        pos->value = bits;
                    }
                } else {
                    // Don't ignore case.  Nice simple case where the mask-compare will
                    // determine definitely whether we have a match at this character
                    // position.
                    pos->mask = char_mask;
                    pos->value = c;
                    pos->determines_perfectly = true;
                }
                characters_filled_in++;
                MOZ_ASSERT(characters_filled_in <= details->characters());
                if (characters_filled_in == details->characters()) {
                    return;
                }
            }
        } else {
            QuickCheckDetails::Position* pos =
                details->positions(characters_filled_in);
            RegExpCharacterClass* tree = elm.char_class();
            const CharacterRangeVector& ranges = tree->ranges(alloc());
            if (tree->is_negated()) {
                // A quick check uses multi-character mask and compare.  There is no
                // useful way to incorporate a negative char class into this scheme
                // so we just conservatively create a mask and value that will always
                // succeed.
                pos->mask = 0;
                pos->value = 0;
            } else {
                size_t first_range = 0;
                while (ranges[first_range].from() > char_mask) {
                    first_range++;
                    if (first_range == ranges.length()) {
                        details->set_cannot_match();
                        pos->determines_perfectly = false;
                        return;
                    }
                }
                CharacterRange range = ranges[first_range];
                char16_t from = range.from();
                char16_t to = range.to();
                if (to > char_mask) {
                    to = char_mask;
                }
                uint32_t differing_bits = (from ^ to);
                // A mask and compare is only perfect if the differing bits form a
                // number like 00011111 with one single block of trailing 1s.
                if ((differing_bits & (differing_bits + 1)) == 0 &&
                    from + differing_bits == to) {
                    pos->determines_perfectly = true;
                }
                uint32_t common_bits = ~SmearBitsRight(differing_bits);
                uint32_t bits = (from & common_bits);
                for (size_t i = first_range + 1; i < ranges.length(); i++) {
                    CharacterRange range = ranges[i];
                    char16_t from = range.from();
                    char16_t to = range.to();
                    if (from > char_mask) continue;
                    if (to > char_mask) to = char_mask;
                    // Here we are combining more ranges into the mask and compare
                    // value.  With each new range the mask becomes more sparse and
                    // so the chances of a false positive rise.  A character class
                    // with multiple ranges is assumed never to be equivalent to a
                    // mask and compare operation.
                    pos->determines_perfectly = false;
                    uint32_t new_common_bits = (from ^ to);
                    new_common_bits = ~SmearBitsRight(new_common_bits);
                    common_bits &= new_common_bits;
                    bits &= new_common_bits;
                    uint32_t differing_bits = (from & common_bits) ^ bits;
                    common_bits ^= differing_bits;
                    bits &= common_bits;
                }
                pos->mask = common_bits;
                pos->value = bits;
            }
            characters_filled_in++;
            MOZ_ASSERT(characters_filled_in <= details->characters());
            if (characters_filled_in == details->characters()) {
                return;
            }
        }
    }
    MOZ_ASSERT(characters_filled_in != details->characters());
    if (!details->cannot_match()) {
        on_success()-> GetQuickCheckDetails(details,
                                            compiler,
                                            characters_filled_in,
                                            true);
    }
}

void
QuickCheckDetails::Clear()
{
    for (int i = 0; i < characters_; i++) {
        positions_[i].mask = 0;
        positions_[i].value = 0;
        positions_[i].determines_perfectly = false;
    }
    characters_ = 0;
}

void
QuickCheckDetails::Advance(int by, bool ascii)
{
    if (by >= characters_ || by < 0) {
        Clear();
        return;
    }
    for (int i = 0; i < characters_ - by; i++) {
        positions_[i] = positions_[by + i];
    }
    for (int i = characters_ - by; i < characters_; i++) {
        positions_[i].mask = 0;
        positions_[i].value = 0;
        positions_[i].determines_perfectly = false;
    }
    characters_ -= by;
    // We could change mask_ and value_ here but we would never advance unless
    // they had already been used in a check and they won't be used again because
    // it would gain us nothing.  So there's no point.
}

bool
QuickCheckDetails::Rationalize(bool is_ascii)
{
    bool found_useful_op = false;
    uint32_t char_mask = MaximumCharacter(is_ascii);

    mask_ = 0;
    value_ = 0;
    int char_shift = 0;
    for (int i = 0; i < characters_; i++) {
        Position* pos = &positions_[i];
        if ((pos->mask & kMaxOneByteCharCode) != 0)
            found_useful_op = true;
        mask_ |= (pos->mask & char_mask) << char_shift;
        value_ |= (pos->value & char_mask) << char_shift;
        char_shift += is_ascii ? 8 : 16;
    }
    return found_useful_op;
}

void QuickCheckDetails::Merge(QuickCheckDetails* other, int from_index)
{
    MOZ_ASSERT(characters_ == other->characters_);
    if (other->cannot_match_)
        return;
    if (cannot_match_) {
        *this = *other;
        return;
    }
    for (int i = from_index; i < characters_; i++) {
        QuickCheckDetails::Position* pos = positions(i);
        QuickCheckDetails::Position* other_pos = other->positions(i);
        if (pos->mask != other_pos->mask ||
            pos->value != other_pos->value ||
            !other_pos->determines_perfectly) {
            // Our mask-compare operation will be approximate unless we have the
            // exact same operation on both sides of the alternation.
            pos->determines_perfectly = false;
        }
        pos->mask &= other_pos->mask;
        pos->value &= pos->mask;
        other_pos->value &= pos->mask;
        char16_t differing_bits = (pos->value ^ other_pos->value);
        pos->mask &= ~differing_bits;
        pos->value &= pos->mask;
    }
}