aboutsummaryrefslogtreecommitdiffstats
path: root/src/glsl/linker.cpp
blob: be08156e0683b0f78034ff5ff32d96d692fb992b (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
/*
 * Copyright © 2010 Intel Corporation
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
 * DEALINGS IN THE SOFTWARE.
 */

/**
 * \file linker.cpp
 * GLSL linker implementation
 *
 * Given a set of shaders that are to be linked to generate a final program,
 * there are three distinct stages.
 *
 * In the first stage shaders are partitioned into groups based on the shader
 * type.  All shaders of a particular type (e.g., vertex shaders) are linked
 * together.
 *
 *   - Undefined references in each shader are resolve to definitions in
 *     another shader.
 *   - Types and qualifiers of uniforms, outputs, and global variables defined
 *     in multiple shaders with the same name are verified to be the same.
 *   - Initializers for uniforms and global variables defined
 *     in multiple shaders with the same name are verified to be the same.
 *
 * The result, in the terminology of the GLSL spec, is a set of shader
 * executables for each processing unit.
 *
 * After the first stage is complete, a series of semantic checks are performed
 * on each of the shader executables.
 *
 *   - Each shader executable must define a \c main function.
 *   - Each vertex shader executable must write to \c gl_Position.
 *   - Each fragment shader executable must write to either \c gl_FragData or
 *     \c gl_FragColor.
 *
 * In the final stage individual shader executables are linked to create a
 * complete exectuable.
 *
 *   - Types of uniforms defined in multiple shader stages with the same name
 *     are verified to be the same.
 *   - Initializers for uniforms defined in multiple shader stages with the
 *     same name are verified to be the same.
 *   - Types and qualifiers of outputs defined in one stage are verified to
 *     be the same as the types and qualifiers of inputs defined with the same
 *     name in a later stage.
 *
 * \author Ian Romanick <ian.d.romanick@intel.com>
 */

#include "main/core.h"
#include "glsl_symbol_table.h"
#include "ir.h"
#include "program.h"
#include "program/hash_table.h"
#include "linker.h"
#include "ir_optimization.h"

extern "C" {
#include "main/shaderobj.h"
}

#define ALIGN(value, alignment)  (((value) + alignment - 1) & ~(alignment - 1))

/**
 * Visitor that determines whether or not a variable is ever written.
 */
class find_assignment_visitor : public ir_hierarchical_visitor {
public:
   find_assignment_visitor(const char *name)
      : name(name), found(false)
   {
      /* empty */
   }

   virtual ir_visitor_status visit_enter(ir_assignment *ir)
   {
      ir_variable *const var = ir->lhs->variable_referenced();

      if (strcmp(name, var->name) == 0) {
	 found = true;
	 return visit_stop;
      }

      return visit_continue_with_parent;
   }

   virtual ir_visitor_status visit_enter(ir_call *ir)
   {
      exec_list_iterator sig_iter = ir->callee->parameters.iterator();
      foreach_iter(exec_list_iterator, iter, *ir) {
	 ir_rvalue *param_rval = (ir_rvalue *)iter.get();
	 ir_variable *sig_param = (ir_variable *)sig_iter.get();

	 if (sig_param->mode == ir_var_out ||
	     sig_param->mode == ir_var_inout) {
	    ir_variable *var = param_rval->variable_referenced();
	    if (var && strcmp(name, var->name) == 0) {
	       found = true;
	       return visit_stop;
	    }
	 }
	 sig_iter.next();
      }

      if (ir->return_deref != NULL) {
	 ir_variable *const var = ir->return_deref->variable_referenced();

	 if (strcmp(name, var->name) == 0) {
	    found = true;
	    return visit_stop;
	 }
      }

      return visit_continue_with_parent;
   }

   bool variable_found()
   {
      return found;
   }

private:
   const char *name;       /**< Find writes to a variable with this name. */
   bool found;             /**< Was a write to the variable found? */
};


/**
 * Visitor that determines whether or not a variable is ever read.
 */
class find_deref_visitor : public ir_hierarchical_visitor {
public:
   find_deref_visitor(const char *name)
      : name(name), found(false)
   {
      /* empty */
   }

   virtual ir_visitor_status visit(ir_dereference_variable *ir)
   {
      if (strcmp(this->name, ir->var->name) == 0) {
	 this->found = true;
	 return visit_stop;
      }

      return visit_continue;
   }

   bool variable_found() const
   {
      return this->found;
   }

private:
   const char *name;       /**< Find writes to a variable with this name. */
   bool found;             /**< Was a write to the variable found? */
};


void
linker_error(gl_shader_program *prog, const char *fmt, ...)
{
   va_list ap;

   ralloc_strcat(&prog->InfoLog, "error: ");
   va_start(ap, fmt);
   ralloc_vasprintf_append(&prog->InfoLog, fmt, ap);
   va_end(ap);

   prog->LinkStatus = false;
}


void
linker_warning(gl_shader_program *prog, const char *fmt, ...)
{
   va_list ap;

   ralloc_strcat(&prog->InfoLog, "error: ");
   va_start(ap, fmt);
   ralloc_vasprintf_append(&prog->InfoLog, fmt, ap);
   va_end(ap);

}


void
link_invalidate_variable_locations(gl_shader *sh, int input_base,
                                   int output_base)
{
   foreach_list(node, sh->ir) {
      ir_variable *const var = ((ir_instruction *) node)->as_variable();

      if (var == NULL)
         continue;

      int base;
      switch (var->mode) {
      case ir_var_in:
         base = input_base;
         break;
      case ir_var_out:
         base = output_base;
         break;
      default:
         continue;
      }

      /* Only assign locations for generic attributes / varyings / etc.
       */
      if ((var->location >= base) && !var->explicit_location)
         var->location = -1;

      if ((var->location == -1) && !var->explicit_location) {
         var->is_unmatched_generic_inout = 1;
         var->location_frac = 0;
      } else {
         var->is_unmatched_generic_inout = 0;
      }
   }
}


/**
 * Determine the number of attribute slots required for a particular type
 *
 * This code is here because it implements the language rules of a specific
 * GLSL version.  Since it's a property of the language and not a property of
 * types in general, it doesn't really belong in glsl_type.
 */
unsigned
count_attribute_slots(const glsl_type *t)
{
   /* From page 31 (page 37 of the PDF) of the GLSL 1.50 spec:
    *
    *     "A scalar input counts the same amount against this limit as a vec4,
    *     so applications may want to consider packing groups of four
    *     unrelated float inputs together into a vector to better utilize the
    *     capabilities of the underlying hardware. A matrix input will use up
    *     multiple locations.  The number of locations used will equal the
    *     number of columns in the matrix."
    *
    * The spec does not explicitly say how arrays are counted.  However, it
    * should be safe to assume the total number of slots consumed by an array
    * is the number of entries in the array multiplied by the number of slots
    * consumed by a single element of the array.
    */

   if (t->is_array())
      return t->array_size() * count_attribute_slots(t->element_type());

   if (t->is_matrix())
      return t->matrix_columns;

   return 1;
}


/**
 * Verify that a vertex shader executable meets all semantic requirements.
 *
 * Also sets prog->Vert.UsesClipDistance and prog->Vert.ClipDistanceArraySize
 * as a side effect.
 *
 * \param shader  Vertex shader executable to be verified
 */
bool
validate_vertex_shader_executable(struct gl_shader_program *prog,
				  struct gl_shader *shader)
{
   if (shader == NULL)
      return true;

   /* From the GLSL 1.10 spec, page 48:
    *
    *     "The variable gl_Position is available only in the vertex
    *      language and is intended for writing the homogeneous vertex
    *      position. All executions of a well-formed vertex shader
    *      executable must write a value into this variable. [...] The
    *      variable gl_Position is available only in the vertex
    *      language and is intended for writing the homogeneous vertex
    *      position. All executions of a well-formed vertex shader
    *      executable must write a value into this variable."
    *
    * while in GLSL 1.40 this text is changed to:
    *
    *     "The variable gl_Position is available only in the vertex
    *      language and is intended for writing the homogeneous vertex
    *      position. It can be written at any time during shader
    *      execution. It may also be read back by a vertex shader
    *      after being written. This value will be used by primitive
    *      assembly, clipping, culling, and other fixed functionality
    *      operations, if present, that operate on primitives after
    *      vertex processing has occurred. Its value is undefined if
    *      the vertex shader executable does not write gl_Position."
    *
    * GLSL ES 3.00 is similar to GLSL 1.40--failing to write to gl_Position is
    * not an error.
    */
   if (prog->Version < (prog->IsES ? 300 : 140)) {
      find_assignment_visitor find("gl_Position");
      find.run(shader->ir);
      if (!find.variable_found()) {
	 linker_error(prog, "vertex shader does not write to `gl_Position'\n");
	 return false;
      }
   }

   prog->Vert.ClipDistanceArraySize = 0;

   if (!prog->IsES && prog->Version >= 130) {
      /* From section 7.1 (Vertex Shader Special Variables) of the
       * GLSL 1.30 spec:
       *
       *   "It is an error for a shader to statically write both
       *   gl_ClipVertex and gl_ClipDistance."
       *
       * This does not apply to GLSL ES shaders, since GLSL ES defines neither
       * gl_ClipVertex nor gl_ClipDistance.
       */
      find_assignment_visitor clip_vertex("gl_ClipVertex");
      find_assignment_visitor clip_distance("gl_ClipDistance");

      clip_vertex.run(shader->ir);
      clip_distance.run(shader->ir);
      if (clip_vertex.variable_found() && clip_distance.variable_found()) {
         linker_error(prog, "vertex shader writes to both `gl_ClipVertex' "
                      "and `gl_ClipDistance'\n");
         return false;
      }
      prog->Vert.UsesClipDistance = clip_distance.variable_found();
      ir_variable *clip_distance_var =
         shader->symbols->get_variable("gl_ClipDistance");
      if (clip_distance_var)
         prog->Vert.ClipDistanceArraySize = clip_distance_var->type->length;
   }

   return true;
}


/**
 * Verify that a fragment shader executable meets all semantic requirements
 *
 * \param shader  Fragment shader executable to be verified
 */
bool
validate_fragment_shader_executable(struct gl_shader_program *prog,
				    struct gl_shader *shader)
{
   if (shader == NULL)
      return true;

   find_assignment_visitor frag_color("gl_FragColor");
   find_assignment_visitor frag_data("gl_FragData");

   frag_color.run(shader->ir);
   frag_data.run(shader->ir);

   if (frag_color.variable_found() && frag_data.variable_found()) {
      linker_error(prog,  "fragment shader writes to both "
		   "`gl_FragColor' and `gl_FragData'\n");
      return false;
   }

   return true;
}


/**
 * Generate a string describing the mode of a variable
 */
static const char *
mode_string(const ir_variable *var)
{
   switch (var->mode) {
   case ir_var_auto:
      return (var->read_only) ? "global constant" : "global variable";

   case ir_var_uniform: return "uniform";
   case ir_var_in:      return "shader input";
   case ir_var_out:     return "shader output";
   case ir_var_inout:   return "shader inout";

   case ir_var_const_in:
   case ir_var_temporary:
   default:
      assert(!"Should not get here.");
      return "invalid variable";
   }
}


/**
 * Perform validation of global variables used across multiple shaders
 */
bool
cross_validate_globals(struct gl_shader_program *prog,
		       struct gl_shader **shader_list,
		       unsigned num_shaders,
		       bool uniforms_only)
{
   /* Examine all of the uniforms in all of the shaders and cross validate
    * them.
    */
   glsl_symbol_table variables;
   for (unsigned i = 0; i < num_shaders; i++) {
      if (shader_list[i] == NULL)
	 continue;

      foreach_list(node, shader_list[i]->ir) {
	 ir_variable *const var = ((ir_instruction *) node)->as_variable();

	 if (var == NULL)
	    continue;

	 if (uniforms_only && (var->mode != ir_var_uniform))
	    continue;

	 /* Don't cross validate temporaries that are at global scope.  These
	  * will eventually get pulled into the shaders 'main'.
	  */
	 if (var->mode == ir_var_temporary)
	    continue;

	 /* If a global with this name has already been seen, verify that the
	  * new instance has the same type.  In addition, if the globals have
	  * initializers, the values of the initializers must be the same.
	  */
	 ir_variable *const existing = variables.get_variable(var->name);
	 if (existing != NULL) {
	    if (var->type != existing->type) {
	       /* Consider the types to be "the same" if both types are arrays
		* of the same type and one of the arrays is implicitly sized.
		* In addition, set the type of the linked variable to the
		* explicitly sized array.
		*/
	       if (var->type->is_array()
		   && existing->type->is_array()
		   && (var->type->fields.array == existing->type->fields.array)
		   && ((var->type->length == 0)
		       || (existing->type->length == 0))) {
		  if (var->type->length != 0) {
		     existing->type = var->type;
		  }
	       } else {
		  linker_error(prog, "%s `%s' declared as type "
			       "`%s' and type `%s'\n",
			       mode_string(var),
			       var->name, var->type->name,
			       existing->type->name);
		  return false;
	       }
	    }

	    if (var->explicit_location) {
	       if (existing->explicit_location
		   && (var->location != existing->location)) {
		     linker_error(prog, "explicit locations for %s "
				  "`%s' have differing values\n",
				  mode_string(var), var->name);
		     return false;
	       }

	       existing->location = var->location;
	       existing->explicit_location = true;
	    }

	    /* Validate layout qualifiers for gl_FragDepth.
	     *
	     * From the AMD/ARB_conservative_depth specs:
	     *
	     *    "If gl_FragDepth is redeclared in any fragment shader in a
	     *    program, it must be redeclared in all fragment shaders in
	     *    that program that have static assignments to
	     *    gl_FragDepth. All redeclarations of gl_FragDepth in all
	     *    fragment shaders in a single program must have the same set
	     *    of qualifiers."
	     */
	    if (strcmp(var->name, "gl_FragDepth") == 0) {
	       bool layout_declared = var->depth_layout != ir_depth_layout_none;
	       bool layout_differs =
		  var->depth_layout != existing->depth_layout;

	       if (layout_declared && layout_differs) {
		  linker_error(prog,
			       "All redeclarations of gl_FragDepth in all "
			       "fragment shaders in a single program must have "
			       "the same set of qualifiers.");
	       }

	       if (var->used && layout_differs) {
		  linker_error(prog,
			       "If gl_FragDepth is redeclared with a layout "
			       "qualifier in any fragment shader, it must be "
			       "redeclared with the same layout qualifier in "
			       "all fragment shaders that have assignments to "
			       "gl_FragDepth");
	       }
	    }

	    /* Page 35 (page 41 of the PDF) of the GLSL 4.20 spec says:
	     *
	     *     "If a shared global has multiple initializers, the
	     *     initializers must all be constant expressions, and they
	     *     must all have the same value. Otherwise, a link error will
	     *     result. (A shared global having only one initializer does
	     *     not require that initializer to be a constant expression.)"
	     *
	     * Previous to 4.20 the GLSL spec simply said that initializers
	     * must have the same value.  In this case of non-constant
	     * initializers, this was impossible to determine.  As a result,
	     * no vendor actually implemented that behavior.  The 4.20
	     * behavior matches the implemented behavior of at least one other
	     * vendor, so we'll implement that for all GLSL versions.
	     */
	    if (var->constant_initializer != NULL) {
	       if (existing->constant_initializer != NULL) {
		  if (!var->constant_initializer->has_value(existing->constant_initializer)) {
		     linker_error(prog, "initializers for %s "
				  "`%s' have differing values\n",
				  mode_string(var), var->name);
		     return false;
		  }
	       } else {
		  /* If the first-seen instance of a particular uniform did not
		   * have an initializer but a later instance does, copy the
		   * initializer to the version stored in the symbol table.
		   */
		  /* FINISHME: This is wrong.  The constant_value field should
		   * FINISHME: not be modified!  Imagine a case where a shader
		   * FINISHME: without an initializer is linked in two different
		   * FINISHME: programs with shaders that have differing
		   * FINISHME: initializers.  Linking with the first will
		   * FINISHME: modify the shader, and linking with the second
		   * FINISHME: will fail.
		   */
		  existing->constant_initializer =
		     var->constant_initializer->clone(ralloc_parent(existing),
						      NULL);
	       }
	    }

	    if (var->has_initializer) {
	       if (existing->has_initializer
		   && (var->constant_initializer == NULL
		       || existing->constant_initializer == NULL)) {
		  linker_error(prog,
			       "shared global variable `%s' has multiple "
			       "non-constant initializers.\n",
			       var->name);
		  return false;
	       }

	       /* Some instance had an initializer, so keep track of that.  In
		* this location, all sorts of initializers (constant or
		* otherwise) will propagate the existence to the variable
		* stored in the symbol table.
		*/
	       existing->has_initializer = true;
	    }

	    if (existing->invariant != var->invariant) {
	       linker_error(prog, "declarations for %s `%s' have "
			    "mismatching invariant qualifiers\n",
			    mode_string(var), var->name);
	       return false;
	    }
            if (existing->centroid != var->centroid) {
               linker_error(prog, "declarations for %s `%s' have "
			    "mismatching centroid qualifiers\n",
			    mode_string(var), var->name);
               return false;
            }
	 } else
	    variables.add_variable(var);
      }
   }

   return true;
}


/**
 * Perform validation of uniforms used across multiple shader stages
 */
bool
cross_validate_uniforms(struct gl_shader_program *prog)
{
   return cross_validate_globals(prog, prog->_LinkedShaders,
				 MESA_SHADER_TYPES, true);
}

/**
 * Accumulates the array of prog->UniformBlocks and checks that all
 * definitons of blocks agree on their contents.
 */
static bool
interstage_cross_validate_uniform_blocks(struct gl_shader_program *prog)
{
   unsigned max_num_uniform_blocks = 0;
   for (unsigned i = 0; i < MESA_SHADER_TYPES; i++) {
      if (prog->_LinkedShaders[i])
	 max_num_uniform_blocks += prog->_LinkedShaders[i]->NumUniformBlocks;
   }

   for (unsigned i = 0; i < MESA_SHADER_TYPES; i++) {
      struct gl_shader *sh = prog->_LinkedShaders[i];

      prog->UniformBlockStageIndex[i] = ralloc_array(prog, int,
						     max_num_uniform_blocks);
      for (unsigned int j = 0; j < max_num_uniform_blocks; j++)
	 prog->UniformBlockStageIndex[i][j] = -1;

      if (sh == NULL)
	 continue;

      for (unsigned int j = 0; j < sh->NumUniformBlocks; j++) {
	 int index = link_cross_validate_uniform_block(prog,
						       &prog->UniformBlocks,
						       &prog->NumUniformBlocks,
						       &sh->UniformBlocks[j]);

	 if (index == -1) {
	    linker_error(prog, "uniform block `%s' has mismatching definitions",
			 sh->UniformBlocks[j].Name);
	    return false;
	 }

	 prog->UniformBlockStageIndex[i][index] = j;
      }
   }

   return true;
}

/**
 * Validate that outputs from one stage match inputs of another
 */
bool
cross_validate_outputs_to_inputs(struct gl_shader_program *prog,
				 gl_shader *producer, gl_shader *consumer)
{
   glsl_symbol_table parameters;
   /* FINISHME: Figure these out dynamically. */
   const char *const producer_stage = "vertex";
   const char *const consumer_stage = "fragment";

   /* Find all shader outputs in the "producer" stage.
    */
   foreach_list(node, producer->ir) {
      ir_variable *const var = ((ir_instruction *) node)->as_variable();

      /* FINISHME: For geometry shaders, this should also look for inout
       * FINISHME: variables.
       */
      if ((var == NULL) || (var->mode != ir_var_out))
	 continue;

      parameters.add_variable(var);
   }


   /* Find all shader inputs in the "consumer" stage.  Any variables that have
    * matching outputs already in the symbol table must have the same type and
    * qualifiers.
    */
   foreach_list(node, consumer->ir) {
      ir_variable *const input = ((ir_instruction *) node)->as_variable();

      /* FINISHME: For geometry shaders, this should also look for inout
       * FINISHME: variables.
       */
      if ((input == NULL) || (input->mode != ir_var_in))
	 continue;

      ir_variable *const output = parameters.get_variable(input->name);
      if (output != NULL) {
	 /* Check that the types match between stages.
	  */
	 if (input->type != output->type) {
	    /* There is a bit of a special case for gl_TexCoord.  This
	     * built-in is unsized by default.  Applications that variable
	     * access it must redeclare it with a size.  There is some
	     * language in the GLSL spec that implies the fragment shader
	     * and vertex shader do not have to agree on this size.  Other
	     * driver behave this way, and one or two applications seem to
	     * rely on it.
	     *
	     * Neither declaration needs to be modified here because the array
	     * sizes are fixed later when update_array_sizes is called.
	     *
	     * From page 48 (page 54 of the PDF) of the GLSL 1.10 spec:
	     *
	     *     "Unlike user-defined varying variables, the built-in
	     *     varying variables don't have a strict one-to-one
	     *     correspondence between the vertex language and the
	     *     fragment language."
	     */
	    if (!output->type->is_array()
		|| (strncmp("gl_", output->name, 3) != 0)) {
	       linker_error(prog,
			    "%s shader output `%s' declared as type `%s', "
			    "but %s shader input declared as type `%s'\n",
			    producer_stage, output->name,
			    output->type->name,
			    consumer_stage, input->type->name);
	       return false;
	    }
	 }

	 /* Check that all of the qualifiers match between stages.
	  */
	 if (input->centroid != output->centroid) {
	    linker_error(prog,
			 "%s shader output `%s' %s centroid qualifier, "
			 "but %s shader input %s centroid qualifier\n",
			 producer_stage,
			 output->name,
			 (output->centroid) ? "has" : "lacks",
			 consumer_stage,
			 (input->centroid) ? "has" : "lacks");
	    return false;
	 }

	 if (input->invariant != output->invariant) {
	    linker_error(prog,
			 "%s shader output `%s' %s invariant qualifier, "
			 "but %s shader input %s invariant qualifier\n",
			 producer_stage,
			 output->name,
			 (output->invariant) ? "has" : "lacks",
			 consumer_stage,
			 (input->invariant) ? "has" : "lacks");
	    return false;
	 }

	 if (input->interpolation != output->interpolation) {
	    linker_error(prog,
			 "%s shader output `%s' specifies %s "
			 "interpolation qualifier, "
			 "but %s shader input specifies %s "
			 "interpolation qualifier\n",
			 producer_stage,
			 output->name,
			 output->interpolation_string(),
			 consumer_stage,
			 input->interpolation_string());
	    return false;
	 }
      }
   }

   return true;
}


/**
 * Populates a shaders symbol table with all global declarations
 */
static void
populate_symbol_table(gl_shader *sh)
{
   sh->symbols = new(sh) glsl_symbol_table;

   foreach_list(node, sh->ir) {
      ir_instruction *const inst = (ir_instruction *) node;
      ir_variable *var;
      ir_function *func;

      if ((func = inst->as_function()) != NULL) {
	 sh->symbols->add_function(func);
      } else if ((var = inst->as_variable()) != NULL) {
	 sh->symbols->add_variable(var);
      }
   }
}


/**
 * Remap variables referenced in an instruction tree
 *
 * This is used when instruction trees are cloned from one shader and placed in
 * another.  These trees will contain references to \c ir_variable nodes that
 * do not exist in the target shader.  This function finds these \c ir_variable
 * references and replaces the references with matching variables in the target
 * shader.
 *
 * If there is no matching variable in the target shader, a clone of the
 * \c ir_variable is made and added to the target shader.  The new variable is
 * added to \b both the instruction stream and the symbol table.
 *
 * \param inst         IR tree that is to be processed.
 * \param symbols      Symbol table containing global scope symbols in the
 *                     linked shader.
 * \param instructions Instruction stream where new variable declarations
 *                     should be added.
 */
void
remap_variables(ir_instruction *inst, struct gl_shader *target,
		hash_table *temps)
{
   class remap_visitor : public ir_hierarchical_visitor {
   public:
	 remap_visitor(struct gl_shader *target,
		    hash_table *temps)
      {
	 this->target = target;
	 this->symbols = target->symbols;
	 this->instructions = target->ir;
	 this->temps = temps;
      }

      virtual ir_visitor_status visit(ir_dereference_variable *ir)
      {
	 if (ir->var->mode == ir_var_temporary) {
	    ir_variable *var = (ir_variable *) hash_table_find(temps, ir->var);

	    assert(var != NULL);
	    ir->var = var;
	    return visit_continue;
	 }

	 ir_variable *const existing =
	    this->symbols->get_variable(ir->var->name);
	 if (existing != NULL)
	    ir->var = existing;
	 else {
	    ir_variable *copy = ir->var->clone(this->target, NULL);

	    this->symbols->add_variable(copy);
	    this->instructions->push_head(copy);
	    ir->var = copy;
	 }

	 return visit_continue;
      }

   private:
      struct gl_shader *target;
      glsl_symbol_table *symbols;
      exec_list *instructions;
      hash_table *temps;
   };

   remap_visitor v(target, temps);

   inst->accept(&v);
}


/**
 * Move non-declarations from one instruction stream to another
 *
 * The intended usage pattern of this function is to pass the pointer to the
 * head sentinel of a list (i.e., a pointer to the list cast to an \c exec_node
 * pointer) for \c last and \c false for \c make_copies on the first
 * call.  Successive calls pass the return value of the previous call for
 * \c last and \c true for \c make_copies.
 *
 * \param instructions Source instruction stream
 * \param last         Instruction after which new instructions should be
 *                     inserted in the target instruction stream
 * \param make_copies  Flag selecting whether instructions in \c instructions
 *                     should be copied (via \c ir_instruction::clone) into the
 *                     target list or moved.
 *
 * \return
 * The new "last" instruction in the target instruction stream.  This pointer
 * is suitable for use as the \c last parameter of a later call to this
 * function.
 */
exec_node *
move_non_declarations(exec_list *instructions, exec_node *last,
		      bool make_copies, gl_shader *target)
{
   hash_table *temps = NULL;

   if (make_copies)
      temps = hash_table_ctor(0, hash_table_pointer_hash,
			      hash_table_pointer_compare);

   foreach_list_safe(node, instructions) {
      ir_instruction *inst = (ir_instruction *) node;

      if (inst->as_function())
	 continue;

      ir_variable *var = inst->as_variable();
      if ((var != NULL) && (var->mode != ir_var_temporary))
	 continue;

      assert(inst->as_assignment()
             || inst->as_call()
             || inst->as_if() /* for initializers with the ?: operator */
	     || ((var != NULL) && (var->mode == ir_var_temporary)));

      if (make_copies) {
	 inst = inst->clone(target, NULL);

	 if (var != NULL)
	    hash_table_insert(temps, inst, var);
	 else
	    remap_variables(inst, target, temps);
      } else {
	 inst->remove();
      }

      last->insert_after(inst);
      last = inst;
   }

   if (make_copies)
      hash_table_dtor(temps);

   return last;
}

/**
 * Get the function signature for main from a shader
 */
static ir_function_signature *
get_main_function_signature(gl_shader *sh)
{
   ir_function *const f = sh->symbols->get_function("main");
   if (f != NULL) {
      exec_list void_parameters;

      /* Look for the 'void main()' signature and ensure that it's defined.
       * This keeps the linker from accidentally pick a shader that just
       * contains a prototype for main.
       *
       * We don't have to check for multiple definitions of main (in multiple
       * shaders) because that would have already been caught above.
       */
      ir_function_signature *sig = f->matching_signature(&void_parameters);
      if ((sig != NULL) && sig->is_defined) {
	 return sig;
      }
   }

   return NULL;
}


/**
 * This class is only used in link_intrastage_shaders() below but declaring
 * it inside that function leads to compiler warnings with some versions of
 * gcc.
 */
class array_sizing_visitor : public ir_hierarchical_visitor {
public:
   virtual ir_visitor_status visit(ir_variable *var)
   {
      if (var->type->is_array() && (var->type->length == 0)) {
         const glsl_type *type =
            glsl_type::get_array_instance(var->type->fields.array,
                                          var->max_array_access + 1);
         assert(type != NULL);
         var->type = type;
      }
      return visit_continue;
   }
};

/**
 * Combine a group of shaders for a single stage to generate a linked shader
 *
 * \note
 * If this function is supplied a single shader, it is cloned, and the new
 * shader is returned.
 */
static struct gl_shader *
link_intrastage_shaders(void *mem_ctx,
			struct gl_context *ctx,
			struct gl_shader_program *prog,
			struct gl_shader **shader_list,
			unsigned num_shaders)
{
   struct gl_uniform_block *uniform_blocks = NULL;
   unsigned num_uniform_blocks = 0;

   /* Check that global variables defined in multiple shaders are consistent.
    */
   if (!cross_validate_globals(prog, shader_list, num_shaders, false))
      return NULL;

   /* Check that uniform blocks between shaders for a stage agree. */
   for (unsigned i = 0; i < num_shaders; i++) {
      struct gl_shader *sh = shader_list[i];

      for (unsigned j = 0; j < shader_list[i]->NumUniformBlocks; j++) {
	 link_assign_uniform_block_offsets(shader_list[i]);

	 int index = link_cross_validate_uniform_block(mem_ctx,
						       &uniform_blocks,
						       &num_uniform_blocks,
						       &sh->UniformBlocks[j]);
	 if (index == -1) {
	    linker_error(prog, "uniform block `%s' has mismatching definitions",
			 sh->UniformBlocks[j].Name);
	    return NULL;
	 }
      }
   }

   /* Check that there is only a single definition of each function signature
    * across all shaders.
    */
   for (unsigned i = 0; i < (num_shaders - 1); i++) {
      foreach_list(node, shader_list[i]->ir) {
	 ir_function *const f = ((ir_instruction *) node)->as_function();

	 if (f == NULL)
	    continue;

	 for (unsigned j = i + 1; j < num_shaders; j++) {
	    ir_function *const other =
	       shader_list[j]->symbols->get_function(f->name);

	    /* If the other shader has no function (and therefore no function
	     * signatures) with the same name, skip to the next shader.
	     */
	    if (other == NULL)
	       continue;

	    foreach_iter (exec_list_iterator, iter, *f) {
	       ir_function_signature *sig =
		  (ir_function_signature *) iter.get();

	       if (!sig->is_defined || sig->is_builtin)
		  continue;

	       ir_function_signature *other_sig =
		  other->exact_matching_signature(& sig->parameters);

	       if ((other_sig != NULL) && other_sig->is_defined
		   && !other_sig->is_builtin) {
		  linker_error(prog, "function `%s' is multiply defined",
			       f->name);
		  return NULL;
	       }
	    }
	 }
      }
   }

   /* Find the shader that defines main, and make a clone of it.
    *
    * Starting with the clone, search for undefined references.  If one is
    * found, find the shader that defines it.  Clone the reference and add
    * it to the shader.  Repeat until there are no undefined references or
    * until a reference cannot be resolved.
    */
   gl_shader *main = NULL;
   for (unsigned i = 0; i < num_shaders; i++) {
      if (get_main_function_signature(shader_list[i]) != NULL) {
	 main = shader_list[i];
	 break;
      }
   }

   if (main == NULL) {
      linker_error(prog, "%s shader lacks `main'\n",
		   (shader_list[0]->Type == GL_VERTEX_SHADER)
		   ? "vertex" : "fragment");
      return NULL;
   }

   gl_shader *linked = ctx->Driver.NewShader(NULL, 0, main->Type);
   linked->ir = new(linked) exec_list;
   clone_ir_list(mem_ctx, linked->ir, main->ir);

   linked->UniformBlocks = uniform_blocks;
   linked->NumUniformBlocks = num_uniform_blocks;
   ralloc_steal(linked, linked->UniformBlocks);

   populate_symbol_table(linked);

   /* The a pointer to the main function in the final linked shader (i.e., the
    * copy of the original shader that contained the main function).
    */
   ir_function_signature *const main_sig = get_main_function_signature(linked);

   /* Move any instructions other than variable declarations or function
    * declarations into main.
    */
   exec_node *insertion_point =
      move_non_declarations(linked->ir, (exec_node *) &main_sig->body, false,
			    linked);

   for (unsigned i = 0; i < num_shaders; i++) {
      if (shader_list[i] == main)
	 continue;

      insertion_point = move_non_declarations(shader_list[i]->ir,
					      insertion_point, true, linked);
   }

   /* Resolve initializers for global variables in the linked shader.
    */
   unsigned num_linking_shaders = num_shaders;
   for (unsigned i = 0; i < num_shaders; i++)
      num_linking_shaders += shader_list[i]->num_builtins_to_link;

   gl_shader **linking_shaders =
      (gl_shader **) calloc(num_linking_shaders, sizeof(gl_shader *));

   memcpy(linking_shaders, shader_list,
	  sizeof(linking_shaders[0]) * num_shaders);

   unsigned idx = num_shaders;
   for (unsigned i = 0; i < num_shaders; i++) {
      memcpy(&linking_shaders[idx], shader_list[i]->builtins_to_link,
	     sizeof(linking_shaders[0]) * shader_list[i]->num_builtins_to_link);
      idx += shader_list[i]->num_builtins_to_link;
   }

   assert(idx == num_linking_shaders);

   if (!link_function_calls(prog, linked, linking_shaders,
			    num_linking_shaders)) {
      ctx->Driver.DeleteShader(ctx, linked);
      linked = NULL;
   }

   free(linking_shaders);

#ifdef DEBUG
   /* At this point linked should contain all of the linked IR, so
    * validate it to make sure nothing went wrong.
    */
   if (linked)
      validate_ir_tree(linked->ir);
#endif

   /* Make a pass over all variable declarations to ensure that arrays with
    * unspecified sizes have a size specified.  The size is inferred from the
    * max_array_access field.
    */
   if (linked != NULL) {
      array_sizing_visitor v;

      v.run(linked->ir);
   }

   return linked;
}

/**
 * Update the sizes of linked shader uniform arrays to the maximum
 * array index used.
 *
 * From page 81 (page 95 of the PDF) of the OpenGL 2.1 spec:
 *
 *     If one or more elements of an array are active,
 *     GetActiveUniform will return the name of the array in name,
 *     subject to the restrictions listed above. The type of the array
 *     is returned in type. The size parameter contains the highest
 *     array element index used, plus one. The compiler or linker
 *     determines the highest index used.  There will be only one
 *     active uniform reported by the GL per uniform array.

 */
static void
update_array_sizes(struct gl_shader_program *prog)
{
   for (unsigned i = 0; i < MESA_SHADER_TYPES; i++) {
	 if (prog->_LinkedShaders[i] == NULL)
	    continue;

      foreach_list(node, prog->_LinkedShaders[i]->ir) {
	 ir_variable *const var = ((ir_instruction *) node)->as_variable();

	 if ((var == NULL) || (var->mode != ir_var_uniform &&
			       var->mode != ir_var_in &&
			       var->mode != ir_var_out) ||
	     !var->type->is_array())
	    continue;

	 /* GL_ARB_uniform_buffer_object says that std140 uniforms
	  * will not be eliminated.  Since we always do std140, just
	  * don't resize arrays in UBOs.
	  */
	 if (var->uniform_block != -1)
	    continue;

	 unsigned int size = var->max_array_access;
	 for (unsigned j = 0; j < MESA_SHADER_TYPES; j++) {
	       if (prog->_LinkedShaders[j] == NULL)
		  continue;

	    foreach_list(node2, prog->_LinkedShaders[j]->ir) {
	       ir_variable *other_var = ((ir_instruction *) node2)->as_variable();
	       if (!other_var)
		  continue;

	       if (strcmp(var->name, other_var->name) == 0 &&
		   other_var->max_array_access > size) {
		  size = other_var->max_array_access;
	       }
	    }
	 }

	 if (size + 1 != var->type->fields.array->length) {
	    /* If this is a built-in uniform (i.e., it's backed by some
	     * fixed-function state), adjust the number of state slots to
	     * match the new array size.  The number of slots per array entry
	     * is not known.  It seems safe to assume that the total number of
	     * slots is an integer multiple of the number of array elements.
	     * Determine the number of slots per array element by dividing by
	     * the old (total) size.
	     */
	    if (var->num_state_slots > 0) {
	       var->num_state_slots = (size + 1)
		  * (var->num_state_slots / var->type->length);
	    }

	    var->type = glsl_type::get_array_instance(var->type->fields.array,
						      size + 1);
	    /* FINISHME: We should update the types of array
	     * dereferences of this variable now.
	     */
	 }
      }
   }
}

/**
 * Find a contiguous set of available bits in a bitmask.
 *
 * \param used_mask     Bits representing used (1) and unused (0) locations
 * \param needed_count  Number of contiguous bits needed.
 *
 * \return
 * Base location of the available bits on success or -1 on failure.
 */
int
find_available_slots(unsigned used_mask, unsigned needed_count)
{
   unsigned needed_mask = (1 << needed_count) - 1;
   const int max_bit_to_test = (8 * sizeof(used_mask)) - needed_count;

   /* The comparison to 32 is redundant, but without it GCC emits "warning:
    * cannot optimize possibly infinite loops" for the loop below.
    */
   if ((needed_count == 0) || (max_bit_to_test < 0) || (max_bit_to_test > 32))
      return -1;

   for (int i = 0; i <= max_bit_to_test; i++) {
      if ((needed_mask & ~used_mask) == needed_mask)
	 return i;

      needed_mask <<= 1;
   }

   return -1;
}


/**
 * Assign locations for either VS inputs for FS outputs
 *
 * \param prog          Shader program whose variables need locations assigned
 * \param target_index  Selector for the program target to receive location
 *                      assignmnets.  Must be either \c MESA_SHADER_VERTEX or
 *                      \c MESA_SHADER_FRAGMENT.
 * \param max_index     Maximum number of generic locations.  This corresponds
 *                      to either the maximum number of draw buffers or the
 *                      maximum number of generic attributes.
 *
 * \return
 * If locations are successfully assigned, true is returned.  Otherwise an
 * error is emitted to the shader link log and false is returned.
 */
bool
assign_attribute_or_color_locations(gl_shader_program *prog,
				    unsigned target_index,
				    unsigned max_index)
{
   /* Mark invalid locations as being used.
    */
   unsigned used_locations = (max_index >= 32)
      ? ~0 : ~((1 << max_index) - 1);

   assert((target_index == MESA_SHADER_VERTEX)
	  || (target_index == MESA_SHADER_FRAGMENT));

   gl_shader *const sh = prog->_LinkedShaders[target_index];
   if (sh == NULL)
      return true;

   /* Operate in a total of four passes.
    *
    * 1. Invalidate the location assignments for all vertex shader inputs.
    *
    * 2. Assign locations for inputs that have user-defined (via
    *    glBindVertexAttribLocation) locations and outputs that have
    *    user-defined locations (via glBindFragDataLocation).
    *
    * 3. Sort the attributes without assigned locations by number of slots
    *    required in decreasing order.  Fragmentation caused by attribute
    *    locations assigned by the application may prevent large attributes
    *    from having enough contiguous space.
    *
    * 4. Assign locations to any inputs without assigned locations.
    */

   const int generic_base = (target_index == MESA_SHADER_VERTEX)
      ? (int) VERT_ATTRIB_GENERIC0 : (int) FRAG_RESULT_DATA0;

   const enum ir_variable_mode direction =
      (target_index == MESA_SHADER_VERTEX) ? ir_var_in : ir_var_out;


   /* Temporary storage for the set of attributes that need locations assigned.
    */
   struct temp_attr {
      unsigned slots;
      ir_variable *var;

      /* Used below in the call to qsort. */
      static int compare(const void *a, const void *b)
      {
	 const temp_attr *const l = (const temp_attr *) a;
	 const temp_attr *const r = (const temp_attr *) b;

	 /* Reversed because we want a descending order sort below. */
	 return r->slots - l->slots;
      }
   } to_assign[16];

   unsigned num_attr = 0;

   foreach_list(node, sh->ir) {
      ir_variable *const var = ((ir_instruction *) node)->as_variable();

      if ((var == NULL) || (var->mode != (unsigned) direction))
	 continue;

      if (var->explicit_location) {
	 if ((var->location >= (int)(max_index + generic_base))
	     || (var->location < 0)) {
	    linker_error(prog,
			 "invalid explicit location %d specified for `%s'\n",
			 (var->location < 0)
			 ? var->location : var->location - generic_base,
			 var->name);
	    return false;
	 }
      } else if (target_index == MESA_SHADER_VERTEX) {
	 unsigned binding;

	 if (prog->AttributeBindings->get(binding, var->name)) {
	    assert(binding >= VERT_ATTRIB_GENERIC0);
	    var->location = binding;
            var->is_unmatched_generic_inout = 0;
	 }
      } else if (target_index == MESA_SHADER_FRAGMENT) {
	 unsigned binding;
	 unsigned index;

	 if (prog->FragDataBindings->get(binding, var->name)) {
	    assert(binding >= FRAG_RESULT_DATA0);
	    var->location = binding;
            var->is_unmatched_generic_inout = 0;

	    if (prog->FragDataIndexBindings->get(index, var->name)) {
	       var->index = index;
	    }
	 }
      }

      /* If the variable is not a built-in and has a location statically
       * assigned in the shader (presumably via a layout qualifier), make sure
       * that it doesn't collide with other assigned locations.  Otherwise,
       * add it to the list of variables that need linker-assigned locations.
       */
      const unsigned slots = count_attribute_slots(var->type);
      if (var->location != -1) {
	 if (var->location >= generic_base && var->index < 1) {
	    /* From page 61 of the OpenGL 4.0 spec:
	     *
	     *     "LinkProgram will fail if the attribute bindings assigned
	     *     by BindAttribLocation do not leave not enough space to
	     *     assign a location for an active matrix attribute or an
	     *     active attribute array, both of which require multiple
	     *     contiguous generic attributes."
	     *
	     * Previous versions of the spec contain similar language but omit
	     * the bit about attribute arrays.
	     *
	     * Page 61 of the OpenGL 4.0 spec also says:
	     *
	     *     "It is possible for an application to bind more than one
	     *     attribute name to the same location. This is referred to as
	     *     aliasing. This will only work if only one of the aliased
	     *     attributes is active in the executable program, or if no
	     *     path through the shader consumes more than one attribute of
	     *     a set of attributes aliased to the same location. A link
	     *     error can occur if the linker determines that every path
	     *     through the shader consumes multiple aliased attributes,
	     *     but implementations are not required to generate an error
	     *     in this case."
	     *
	     * These two paragraphs are either somewhat contradictory, or I
	     * don't fully understand one or both of them.
	     */
	    /* FINISHME: The code as currently written does not support
	     * FINISHME: attribute location aliasing (see comment above).
	     */
	    /* Mask representing the contiguous slots that will be used by
	     * this attribute.
	     */
	    const unsigned attr = var->location - generic_base;
	    const unsigned use_mask = (1 << slots) - 1;

	    /* Generate a link error if the set of bits requested for this
	     * attribute overlaps any previously allocated bits.
	     */
	    if ((~(use_mask << attr) & used_locations) != used_locations) {
	       const char *const string = (target_index == MESA_SHADER_VERTEX)
		  ? "vertex shader input" : "fragment shader output";
	       linker_error(prog,
			    "insufficient contiguous locations "
			    "available for %s `%s' %d %d %d", string,
			    var->name, used_locations, use_mask, attr);
	       return false;
	    }

	    used_locations |= (use_mask << attr);
	 }

	 continue;
      }

      to_assign[num_attr].slots = slots;
      to_assign[num_attr].var = var;
      num_attr++;
   }

   /* If all of the attributes were assigned locations by the application (or
    * are built-in attributes with fixed locations), return early.  This should
    * be the common case.
    */
   if (num_attr == 0)
      return true;

   qsort(to_assign, num_attr, sizeof(to_assign[0]), temp_attr::compare);

   if (target_index == MESA_SHADER_VERTEX) {
      /* VERT_ATTRIB_GENERIC0 is a pseudo-alias for VERT_ATTRIB_POS.  It can
       * only be explicitly assigned by via glBindAttribLocation.  Mark it as
       * reserved to prevent it from being automatically allocated below.
       */
      find_deref_visitor find("gl_Vertex");
      find.run(sh->ir);
      if (find.variable_found())
	 used_locations |= (1 << 0);
   }

   for (unsigned i = 0; i < num_attr; i++) {
      /* Mask representing the contiguous slots that will be used by this
       * attribute.
       */
      const unsigned use_mask = (1 << to_assign[i].slots) - 1;

      int location = find_available_slots(used_locations, to_assign[i].slots);

      if (location < 0) {
	 const char *const string = (target_index == MESA_SHADER_VERTEX)
	    ? "vertex shader input" : "fragment shader output";

	 linker_error(prog,
		      "insufficient contiguous locations "
		      "available for %s `%s'",
		      string, to_assign[i].var->name);
	 return false;
      }

      to_assign[i].var->location = generic_base + location;
      to_assign[i].var->is_unmatched_generic_inout = 0;
      used_locations |= (use_mask << location);
   }

   return true;
}


/**
 * Demote shader inputs and outputs that are not used in other stages
 */
void
demote_shader_inputs_and_outputs(gl_shader *sh, enum ir_variable_mode mode)
{
   foreach_list(node, sh->ir) {
      ir_variable *const var = ((ir_instruction *) node)->as_variable();

      if ((var == NULL) || (var->mode != int(mode)))
	 continue;

      /* A shader 'in' or 'out' variable is only really an input or output if
       * its value is used by other shader stages.  This will cause the variable
       * to have a location assigned.
       */
      if (var->is_unmatched_generic_inout) {
	 var->mode = ir_var_auto;
      }
   }
}


/**
 * Data structure tracking information about a transform feedback declaration
 * during linking.
 */
class tfeedback_decl
{
public:
   bool init(struct gl_context *ctx, struct gl_shader_program *prog,
             const void *mem_ctx, const char *input);
   static bool is_same(const tfeedback_decl &x, const tfeedback_decl &y);
   bool assign_location(struct gl_context *ctx, struct gl_shader_program *prog,
                        ir_variable *output_var);
   unsigned get_num_outputs() const;
   bool store(struct gl_context *ctx, struct gl_shader_program *prog,
              struct gl_transform_feedback_info *info, unsigned buffer,
              const unsigned max_outputs) const;
   ir_variable *find_output_var(gl_shader_program *prog,
                                gl_shader *producer) const;

   bool is_next_buffer_separator() const
   {
      return this->next_buffer_separator;
   }

   bool is_varying() const
   {
      return !this->next_buffer_separator && !this->skip_components;
   }

   /**
    * The total number of varying components taken up by this variable.  Only
    * valid if assign_location() has been called.
    */
   unsigned num_components() const
   {
      if (this->is_clip_distance_mesa)
         return this->size;
      else
         return this->vector_elements * this->matrix_columns * this->size;
   }

private:
   /**
    * The name that was supplied to glTransformFeedbackVaryings.  Used for
    * error reporting and glGetTransformFeedbackVarying().
    */
   const char *orig_name;

   /**
    * The name of the variable, parsed from orig_name.
    */
   const char *var_name;

   /**
    * True if the declaration in orig_name represents an array.
    */
   bool is_subscripted;

   /**
    * If is_subscripted is true, the subscript that was specified in orig_name.
    */
   unsigned array_subscript;

   /**
    * True if the variable is gl_ClipDistance and the driver lowers
    * gl_ClipDistance to gl_ClipDistanceMESA.
    */
   bool is_clip_distance_mesa;

   /**
    * The vertex shader output location that the linker assigned for this
    * variable.  -1 if a location hasn't been assigned yet.
    */
   int location;

   /**
    * If non-zero, then this variable may be packed along with other variables
    * into a single varying slot, so this offset should be applied when
    * accessing components.  For example, an offset of 1 means that the x
    * component of this variable is actually stored in component y of the
    * location specified by \c location.
    *
    * Only valid if location != -1.
    */
   unsigned location_frac;

   /**
    * If location != -1, the number of vector elements in this variable, or 1
    * if this variable is a scalar.
    */
   unsigned vector_elements;

   /**
    * If location != -1, the number of matrix columns in this variable, or 1
    * if this variable is not a matrix.
    */
   unsigned matrix_columns;

   /** Type of the varying returned by glGetTransformFeedbackVarying() */
   GLenum type;

   /**
    * If location != -1, the size that should be returned by
    * glGetTransformFeedbackVarying().
    */
   unsigned size;

   /**
    * How many components to skip. If non-zero, this is
    * gl_SkipComponents{1,2,3,4} from ARB_transform_feedback3.
    */
   unsigned skip_components;

   /**
    * Whether this is gl_NextBuffer from ARB_transform_feedback3.
    */
   bool next_buffer_separator;
};


/**
 * Initialize this object based on a string that was passed to
 * glTransformFeedbackVaryings.  If there is a parse error, the error is
 * reported using linker_error(), and false is returned.
 */
bool
tfeedback_decl::init(struct gl_context *ctx, struct gl_shader_program *prog,
                     const void *mem_ctx, const char *input)
{
   /* We don't have to be pedantic about what is a valid GLSL variable name,
    * because any variable with an invalid name can't exist in the IR anyway.
    */

   this->location = -1;
   this->orig_name = input;
   this->is_clip_distance_mesa = false;
   this->skip_components = 0;
   this->next_buffer_separator = false;

   if (ctx->Extensions.ARB_transform_feedback3) {
      /* Parse gl_NextBuffer. */
      if (strcmp(input, "gl_NextBuffer") == 0) {
         this->next_buffer_separator = true;
         return true;
      }

      /* Parse gl_SkipComponents. */
      if (strcmp(input, "gl_SkipComponents1") == 0)
         this->skip_components = 1;
      else if (strcmp(input, "gl_SkipComponents2") == 0)
         this->skip_components = 2;
      else if (strcmp(input, "gl_SkipComponents3") == 0)
         this->skip_components = 3;
      else if (strcmp(input, "gl_SkipComponents4") == 0)
         this->skip_components = 4;

      if (this->skip_components)
         return true;
   }

   /* Parse a declaration. */
   const char *bracket = strrchr(input, '[');

   if (bracket) {
      this->var_name = ralloc_strndup(mem_ctx, input, bracket - input);
      if (sscanf(bracket, "[%u]", &this->array_subscript) != 1) {
         linker_error(prog, "Cannot parse transform feedback varying %s", input);
         return false;
      }
      this->is_subscripted = true;
   } else {
      this->var_name = ralloc_strdup(mem_ctx, input);
      this->is_subscripted = false;
   }

   /* For drivers that lower gl_ClipDistance to gl_ClipDistanceMESA, this
    * class must behave specially to account for the fact that gl_ClipDistance
    * is converted from a float[8] to a vec4[2].
    */
   if (ctx->ShaderCompilerOptions[MESA_SHADER_VERTEX].LowerClipDistance &&
       strcmp(this->var_name, "gl_ClipDistance") == 0) {
      this->is_clip_distance_mesa = true;
   }

   return true;
}


/**
 * Determine whether two tfeedback_decl objects refer to the same variable and
 * array index (if applicable).
 */
bool
tfeedback_decl::is_same(const tfeedback_decl &x, const tfeedback_decl &y)
{
   assert(x.is_varying() && y.is_varying());

   if (strcmp(x.var_name, y.var_name) != 0)
      return false;
   if (x.is_subscripted != y.is_subscripted)
      return false;
   if (x.is_subscripted && x.array_subscript != y.array_subscript)
      return false;
   return true;
}


/**
 * Assign a location for this tfeedback_decl object based on the location
 * assignment in output_var.
 *
 * If an error occurs, the error is reported through linker_error() and false
 * is returned.
 */
bool
tfeedback_decl::assign_location(struct gl_context *ctx,
                                struct gl_shader_program *prog,
                                ir_variable *output_var)
{
   assert(this->is_varying());

   if (output_var->type->is_array()) {
      /* Array variable */
      const unsigned matrix_cols =
         output_var->type->fields.array->matrix_columns;
      const unsigned vector_elements =
         output_var->type->fields.array->vector_elements;
      unsigned actual_array_size = this->is_clip_distance_mesa ?
         prog->Vert.ClipDistanceArraySize : output_var->type->array_size();

      if (this->is_subscripted) {
         /* Check array bounds. */
         if (this->array_subscript >= actual_array_size) {
            linker_error(prog, "Transform feedback varying %s has index "
                         "%i, but the array size is %u.",
                         this->orig_name, this->array_subscript,
                         actual_array_size);
            return false;
         }
         if (this->is_clip_distance_mesa) {
            this->location =
               output_var->location + this->array_subscript / 4;
            this->location_frac = this->array_subscript % 4;
         } else {
            unsigned fine_location
               = output_var->location * 4 + output_var->location_frac;
            unsigned array_elem_size = vector_elements * matrix_cols;
            fine_location += array_elem_size * this->array_subscript;
            this->location = fine_location / 4;
            this->location_frac = fine_location % 4;
         }
         this->size = 1;
      } else {
         this->location = output_var->location;
         this->location_frac = output_var->location_frac;
         this->size = actual_array_size;
      }
      this->vector_elements = vector_elements;
      this->matrix_columns = matrix_cols;
      if (this->is_clip_distance_mesa)
         this->type = GL_FLOAT;
      else
         this->type = output_var->type->fields.array->gl_type;
   } else {
      /* Regular variable (scalar, vector, or matrix) */
      if (this->is_subscripted) {
         linker_error(prog, "Transform feedback varying %s requested, "
                      "but %s is not an array.",
                      this->orig_name, this->var_name);
         return false;
      }
      this->location = output_var->location;
      this->location_frac = output_var->location_frac;
      this->size = 1;
      this->vector_elements = output_var->type->vector_elements;
      this->matrix_columns = output_var->type->matrix_columns;
      this->type = output_var->type->gl_type;
   }

   /* From GL_EXT_transform_feedback:
    *   A program will fail to link if:
    *
    *   * the total number of components to capture in any varying
    *     variable in <varyings> is greater than the constant
    *     MAX_TRANSFORM_FEEDBACK_SEPARATE_COMPONENTS_EXT and the
    *     buffer mode is SEPARATE_ATTRIBS_EXT;
    */
   if (prog->TransformFeedback.BufferMode == GL_SEPARATE_ATTRIBS &&
       this->num_components() >
       ctx->Const.MaxTransformFeedbackSeparateComponents) {
      linker_error(prog, "Transform feedback varying %s exceeds "
                   "MAX_TRANSFORM_FEEDBACK_SEPARATE_COMPONENTS.",
                   this->orig_name);
      return false;
   }

   return true;
}


unsigned
tfeedback_decl::get_num_outputs() const
{
   if (!this->is_varying()) {
      return 0;
   }

   return (this->num_components() + this->location_frac + 3)/4;
}


/**
 * Update gl_transform_feedback_info to reflect this tfeedback_decl.
 *
 * If an error occurs, the error is reported through linker_error() and false
 * is returned.
 */
bool
tfeedback_decl::store(struct gl_context *ctx, struct gl_shader_program *prog,
                      struct gl_transform_feedback_info *info,
                      unsigned buffer, const unsigned max_outputs) const
{
   assert(!this->next_buffer_separator);

   /* Handle gl_SkipComponents. */
   if (this->skip_components) {
      info->BufferStride[buffer] += this->skip_components;
      return true;
   }

   /* From GL_EXT_transform_feedback:
    *   A program will fail to link if:
    *
    *     * the total number of components to capture is greater than
    *       the constant MAX_TRANSFORM_FEEDBACK_INTERLEAVED_COMPONENTS_EXT
    *       and the buffer mode is INTERLEAVED_ATTRIBS_EXT.
    */
   if (prog->TransformFeedback.BufferMode == GL_INTERLEAVED_ATTRIBS &&
       info->BufferStride[buffer] + this->num_components() >
       ctx->Const.MaxTransformFeedbackInterleavedComponents) {
      linker_error(prog, "The MAX_TRANSFORM_FEEDBACK_INTERLEAVED_COMPONENTS "
                   "limit has been exceeded.");
      return false;
   }

   unsigned location = this->location;
   unsigned location_frac = this->location_frac;
   unsigned num_components = this->num_components();
   while (num_components > 0) {
      unsigned output_size = MIN2(num_components, 4 - location_frac);
      assert(info->NumOutputs < max_outputs);
      info->Outputs[info->NumOutputs].ComponentOffset = location_frac;
      info->Outputs[info->NumOutputs].OutputRegister = location;
      info->Outputs[info->NumOutputs].NumComponents = output_size;
      info->Outputs[info->NumOutputs].OutputBuffer = buffer;
      info->Outputs[info->NumOutputs].DstOffset = info->BufferStride[buffer];
      ++info->NumOutputs;
      info->BufferStride[buffer] += output_size;
      num_components -= output_size;
      location++;
      location_frac = 0;
   }

   info->Varyings[info->NumVarying].Name = ralloc_strdup(prog, this->orig_name);
   info->Varyings[info->NumVarying].Type = this->type;
   info->Varyings[info->NumVarying].Size = this->size;
   info->NumVarying++;

   return true;
}


ir_variable *
tfeedback_decl::find_output_var(gl_shader_program *prog,
                                gl_shader *producer) const
{
   const char *name = this->is_clip_distance_mesa
      ? "gl_ClipDistanceMESA" : this->var_name;
   ir_variable *var = producer->symbols->get_variable(name);
   if (var && var->mode == ir_var_out)
      return var;

   /* From GL_EXT_transform_feedback:
    *   A program will fail to link if:
    *
    *   * any variable name specified in the <varyings> array is not
    *     declared as an output in the geometry shader (if present) or
    *     the vertex shader (if no geometry shader is present);
    */
   linker_error(prog, "Transform feedback varying %s undeclared.",
                this->orig_name);
   return NULL;
}


/**
 * Parse all the transform feedback declarations that were passed to
 * glTransformFeedbackVaryings() and store them in tfeedback_decl objects.
 *
 * If an error occurs, the error is reported through linker_error() and false
 * is returned.
 */
static bool
parse_tfeedback_decls(struct gl_context *ctx, struct gl_shader_program *prog,
                      const void *mem_ctx, unsigned num_names,
                      char **varying_names, tfeedback_decl *decls)
{
   for (unsigned i = 0; i < num_names; ++i) {
      if (!decls[i].init(ctx, prog, mem_ctx, varying_names[i]))
         return false;

      if (!decls[i].is_varying())
         continue;

      /* From GL_EXT_transform_feedback:
       *   A program will fail to link if:
       *
       *   * any two entries in the <varyings> array specify the same varying
       *     variable;
       *
       * We interpret this to mean "any two entries in the <varyings> array
       * specify the same varying variable and array index", since transform
       * feedback of arrays would be useless otherwise.
       */
      for (unsigned j = 0; j < i; ++j) {
         if (!decls[j].is_varying())
            continue;

         if (tfeedback_decl::is_same(decls[i], decls[j])) {
            linker_error(prog, "Transform feedback varying %s specified "
                         "more than once.", varying_names[i]);
            return false;
         }
      }
   }
   return true;
}


/**
 * Data structure recording the relationship between outputs of one shader
 * stage (the "producer") and inputs of another (the "consumer").
 */
class varying_matches
{
public:
   varying_matches(bool disable_varying_packing);
   ~varying_matches();
   void record(ir_variable *producer_var, ir_variable *consumer_var);
   unsigned assign_locations();
   void store_locations(unsigned producer_base, unsigned consumer_base) const;

private:
   /**
    * If true, this driver disables varying packing, so all varyings need to
    * be aligned on slot boundaries, and take up a number of slots equal to
    * their number of matrix columns times their array size.
    */
   const bool disable_varying_packing;

   /**
    * Enum representing the order in which varyings are packed within a
    * packing class.
    *
    * Currently we pack vec4's first, then vec2's, then scalar values, then
    * vec3's.  This order ensures that the only vectors that are at risk of
    * having to be "double parked" (split between two adjacent varying slots)
    * are the vec3's.
    */
   enum packing_order_enum {
      PACKING_ORDER_VEC4,
      PACKING_ORDER_VEC2,
      PACKING_ORDER_SCALAR,
      PACKING_ORDER_VEC3,
   };

   static unsigned compute_packing_class(ir_variable *var);
   static packing_order_enum compute_packing_order(ir_variable *var);
   static int match_comparator(const void *x_generic, const void *y_generic);

   /**
    * Structure recording the relationship between a single producer output
    * and a single consumer input.
    */
   struct match {
      /**
       * Packing class for this varying, computed by compute_packing_class().
       */
      unsigned packing_class;

      /**
       * Packing order for this varying, computed by compute_packing_order().
       */
      packing_order_enum packing_order;
      unsigned num_components;

      /**
       * The output variable in the producer stage.
       */
      ir_variable *producer_var;

      /**
       * The input variable in the consumer stage.
       */
      ir_variable *consumer_var;

      /**
       * The location which has been assigned for this varying.  This is
       * expressed in multiples of a float, with the first generic varying
       * (i.e. the one referred to by VERT_RESULT_VAR0 or FRAG_ATTRIB_VAR0)
       * represented by the value 0.
       */
      unsigned generic_location;
   } *matches;

   /**
    * The number of elements in the \c matches array that are currently in
    * use.
    */
   unsigned num_matches;

   /**
    * The number of elements that were set aside for the \c matches array when
    * it was allocated.
    */
   unsigned matches_capacity;
};


varying_matches::varying_matches(bool disable_varying_packing)
   : disable_varying_packing(disable_varying_packing)
{
   /* Note: this initial capacity is rather arbitrarily chosen to be large
    * enough for many cases without wasting an unreasonable amount of space.
    * varying_matches::record() will resize the array if there are more than
    * this number of varyings.
    */
   this->matches_capacity = 8;
   this->matches = (match *)
      malloc(sizeof(*this->matches) * this->matches_capacity);
   this->num_matches = 0;
}


varying_matches::~varying_matches()
{
   free(this->matches);
}


/**
 * Record the given producer/consumer variable pair in the list of variables
 * that should later be assigned locations.
 *
 * It is permissible for \c consumer_var to be NULL (this happens if a
 * variable is output by the producer and consumed by transform feedback, but
 * not consumed by the consumer).
 *
 * If \c producer_var has already been paired up with a consumer_var, or
 * producer_var is part of fixed pipeline functionality (and hence already has
 * a location assigned), this function has no effect.
 */
void
varying_matches::record(ir_variable *producer_var, ir_variable *consumer_var)
{
   if (!producer_var->is_unmatched_generic_inout) {
      /* Either a location already exists for this variable (since it is part
       * of fixed functionality), or it has already been recorded as part of a
       * previous match.
       */
      return;
   }

   if (this->num_matches == this->matches_capacity) {
      this->matches_capacity *= 2;
      this->matches = (match *)
         realloc(this->matches,
                 sizeof(*this->matches) * this->matches_capacity);
   }
   this->matches[this->num_matches].packing_class
      = this->compute_packing_class(producer_var);
   this->matches[this->num_matches].packing_order
      = this->compute_packing_order(producer_var);
   if (this->disable_varying_packing) {
      unsigned slots = producer_var->type->is_array()
         ? (producer_var->type->length
            * producer_var->type->fields.array->matrix_columns)
         : producer_var->type->matrix_columns;
      this->matches[this->num_matches].num_components = 4 * slots;
   } else {
      this->matches[this->num_matches].num_components
         = producer_var->type->component_slots();
   }
   this->matches[this->num_matches].producer_var = producer_var;
   this->matches[this->num_matches].consumer_var = consumer_var;
   this->num_matches++;
   producer_var->is_unmatched_generic_inout = 0;
   if (consumer_var)
      consumer_var->is_unmatched_generic_inout = 0;
}


/**
 * Choose locations for all of the variable matches that were previously
 * passed to varying_matches::record().
 */
unsigned
varying_matches::assign_locations()
{
   /* Sort varying matches into an order that makes them easy to pack. */
   qsort(this->matches, this->num_matches, sizeof(*this->matches),
         &varying_matches::match_comparator);

   unsigned generic_location = 0;

   for (unsigned i = 0; i < this->num_matches; i++) {
      /* Advance to the next slot if this varying has a different packing
       * class than the previous one, and we're not already on a slot
       * boundary.
       */
      if (i > 0 &&
          this->matches[i - 1].packing_class
          != this->matches[i].packing_class) {
         generic_location = ALIGN(generic_location, 4);
      }

      this->matches[i].generic_location = generic_location;

      generic_location += this->matches[i].num_components;
   }

   return (generic_location + 3) / 4;
}


/**
 * Update the producer and consumer shaders to reflect the locations
 * assignments that were made by varying_matches::assign_locations().
 */
void
varying_matches::store_locations(unsigned producer_base,
                                 unsigned consumer_base) const
{
   for (unsigned i = 0; i < this->num_matches; i++) {
      ir_variable *producer_var = this->matches[i].producer_var;
      ir_variable *consumer_var = this->matches[i].consumer_var;
      unsigned generic_location = this->matches[i].generic_location;
      unsigned slot = generic_location / 4;
      unsigned offset = generic_location % 4;

      producer_var->location = producer_base + slot;
      producer_var->location_frac = offset;
      if (consumer_var) {
         assert(consumer_var->location == -1);
         consumer_var->location = consumer_base + slot;
         consumer_var->location_frac = offset;
      }
   }
}


/**
 * Compute the "packing class" of the given varying.  This is an unsigned
 * integer with the property that two variables in the same packing class can
 * be safely backed into the same vec4.
 */
unsigned
varying_matches::compute_packing_class(ir_variable *var)
{
   /* In this initial implementation we conservatively assume that variables
    * can only be packed if their base type (float/int/uint/bool) matches and
    * their interpolation and centroid qualifiers match.
    *
    * TODO: relax these restrictions when the driver back-end permits.
    */
   unsigned packing_class = var->centroid ? 1 : 0;
   packing_class *= 4;
   packing_class += var->interpolation;
   packing_class *= GLSL_TYPE_ERROR;
   packing_class += var->type->get_scalar_type()->base_type;
   return packing_class;
}


/**
 * Compute the "packing order" of the given varying.  This is a sort key we
 * use to determine when to attempt to pack the given varying relative to
 * other varyings in the same packing class.
 */
varying_matches::packing_order_enum
varying_matches::compute_packing_order(ir_variable *var)
{
   const glsl_type *element_type = var->type;

   /* FINISHME: Support for "varying" records in GLSL 1.50. */
   while (element_type->base_type == GLSL_TYPE_ARRAY) {
      element_type = element_type->fields.array;
   }

   switch (element_type->vector_elements) {
   case 1: return PACKING_ORDER_SCALAR;
   case 2: return PACKING_ORDER_VEC2;
   case 3: return PACKING_ORDER_VEC3;
   case 4: return PACKING_ORDER_VEC4;
   default:
      assert(!"Unexpected value of vector_elements");
      return PACKING_ORDER_VEC4;
   }
}


/**
 * Comparison function passed to qsort() to sort varyings by packing_class and
 * then by packing_order.
 */
int
varying_matches::match_comparator(const void *x_generic, const void *y_generic)
{
   const match *x = (const match *) x_generic;
   const match *y = (const match *) y_generic;

   if (x->packing_class != y->packing_class)
      return x->packing_class - y->packing_class;
   return x->packing_order - y->packing_order;
}


/**
 * Is the given variable a varying variable to be counted against the
 * limit in ctx->Const.MaxVarying?
 * This includes variables such as texcoords, colors and generic
 * varyings, but excludes variables such as gl_FrontFacing and gl_FragCoord.
 */
static bool
is_varying_var(GLenum shaderType, const ir_variable *var)
{
   /* Only fragment shaders will take a varying variable as an input */
   if (shaderType == GL_FRAGMENT_SHADER &&
       var->mode == ir_var_in) {
      switch (var->location) {
      case FRAG_ATTRIB_WPOS:
      case FRAG_ATTRIB_FACE:
      case FRAG_ATTRIB_PNTC:
         return false;
      default:
         return true;
      }
   }
   return false;
}


/**
 * Assign locations for all variables that are produced in one pipeline stage
 * (the "producer") and consumed in the next stage (the "consumer").
 *
 * Variables produced by the producer may also be consumed by transform
 * feedback.
 *
 * \param num_tfeedback_decls is the number of declarations indicating
 *        variables that may be consumed by transform feedback.
 *
 * \param tfeedback_decls is a pointer to an array of tfeedback_decl objects
 *        representing the result of parsing the strings passed to
 *        glTransformFeedbackVaryings().  assign_location() will be called for
 *        each of these objects that matches one of the outputs of the
 *        producer.
 *
 * When num_tfeedback_decls is nonzero, it is permissible for the consumer to
 * be NULL.  In this case, varying locations are assigned solely based on the
 * requirements of transform feedback.
 */
bool
assign_varying_locations(struct gl_context *ctx,
			 void *mem_ctx,
			 struct gl_shader_program *prog,
			 gl_shader *producer, gl_shader *consumer,
                         unsigned num_tfeedback_decls,
                         tfeedback_decl *tfeedback_decls)
{
   /* FINISHME: Set dynamically when geometry shader support is added. */
   const unsigned producer_base = VERT_RESULT_VAR0;
   const unsigned consumer_base = FRAG_ATTRIB_VAR0;
   varying_matches matches(ctx->Const.DisableVaryingPacking);

   /* Operate in a total of three passes.
    *
    * 1. Assign locations for any matching inputs and outputs.
    *
    * 2. Mark output variables in the producer that do not have locations as
    *    not being outputs.  This lets the optimizer eliminate them.
    *
    * 3. Mark input variables in the consumer that do not have locations as
    *    not being inputs.  This lets the optimizer eliminate them.
    */

   foreach_list(node, producer->ir) {
      ir_variable *const output_var = ((ir_instruction *) node)->as_variable();

      if ((output_var == NULL) || (output_var->mode != ir_var_out))
	 continue;

      ir_variable *input_var =
	 consumer ? consumer->symbols->get_variable(output_var->name) : NULL;

      if (input_var && input_var->mode != ir_var_in)
         input_var = NULL;

      if (input_var) {
         matches.record(output_var, input_var);
      }
   }

   for (unsigned i = 0; i < num_tfeedback_decls; ++i) {
      if (!tfeedback_decls[i].is_varying())
         continue;

      ir_variable *output_var
         = tfeedback_decls[i].find_output_var(prog, producer);

      if (output_var == NULL)
         return false;

      if (output_var->is_unmatched_generic_inout) {
         matches.record(output_var, NULL);
      }
   }

   const unsigned slots_used = matches.assign_locations();
   matches.store_locations(producer_base, consumer_base);

   for (unsigned i = 0; i < num_tfeedback_decls; ++i) {
      if (!tfeedback_decls[i].is_varying())
         continue;

      ir_variable *output_var
         = tfeedback_decls[i].find_output_var(prog, producer);

      if (!tfeedback_decls[i].assign_location(ctx, prog, output_var))
         return false;
   }

   if (ctx->Const.DisableVaryingPacking) {
      /* Transform feedback code assumes varyings are packed, so if the driver
       * has disabled varying packing, make sure it does not support transform
       * feedback.
       */
      assert(!ctx->Extensions.EXT_transform_feedback);
   } else {
      lower_packed_varyings(mem_ctx, producer_base, slots_used, ir_var_out,
                            producer);
      if (consumer) {
         lower_packed_varyings(mem_ctx, consumer_base, slots_used, ir_var_in,
                               consumer);
      }
   }

   unsigned varying_vectors = 0;

   if (consumer) {
      foreach_list(node, consumer->ir) {
         ir_variable *const var = ((ir_instruction *) node)->as_variable();

         if ((var == NULL) || (var->mode != ir_var_in))
            continue;

         if (var->is_unmatched_generic_inout) {
            if (prog->Version <= 120) {
               /* On page 25 (page 31 of the PDF) of the GLSL 1.20 spec:
                *
                *     Only those varying variables used (i.e. read) in
                *     the fragment shader executable must be written to
                *     by the vertex shader executable; declaring
                *     superfluous varying variables in a vertex shader is
                *     permissible.
                *
                * We interpret this text as meaning that the VS must
                * write the variable for the FS to read it.  See
                * "glsl1-varying read but not written" in piglit.
                */

               linker_error(prog, "fragment shader varying %s not written "
                            "by vertex shader\n.", var->name);
            }

            /* An 'in' variable is only really a shader input if its
             * value is written by the previous stage.
             */
            var->mode = ir_var_auto;
         } else if (is_varying_var(consumer->Type, var)) {
            /* The packing rules are used for vertex shader inputs are also
             * used for fragment shader inputs.
             */
            varying_vectors += count_attribute_slots(var->type);
         }
      }
   }

   if (ctx->API == API_OPENGLES2 || prog->IsES) {
      if (varying_vectors > ctx->Const.MaxVarying) {
         if (ctx->Const.GLSLSkipStrictMaxVaryingLimitCheck) {
            linker_warning(prog, "shader uses too many varying vectors "
                           "(%u > %u), but the driver will try to optimize "
                           "them out; this is non-portable out-of-spec "
                           "behavior\n",
                           varying_vectors, ctx->Const.MaxVarying);
         } else {
            linker_error(prog, "shader uses too many varying vectors "
                         "(%u > %u)\n",
                         varying_vectors, ctx->Const.MaxVarying);
            return false;
         }
      }
   } else {
      const unsigned float_components = varying_vectors * 4;
      if (float_components > ctx->Const.MaxVarying * 4) {
         if (ctx->Const.GLSLSkipStrictMaxVaryingLimitCheck) {
            linker_warning(prog, "shader uses too many varying components "
                           "(%u > %u), but the driver will try to optimize "
                           "them out; this is non-portable out-of-spec "
                           "behavior\n",
                           float_components, ctx->Const.MaxVarying * 4);
         } else {
            linker_error(prog, "shader uses too many varying components "
                         "(%u > %u)\n",
                         float_components, ctx->Const.MaxVarying * 4);
            return false;
         }
      }
   }

   return true;
}


/**
 * Store transform feedback location assignments into
 * prog->LinkedTransformFeedback based on the data stored in tfeedback_decls.
 *
 * If an error occurs, the error is reported through linker_error() and false
 * is returned.
 */
static bool
store_tfeedback_info(struct gl_context *ctx, struct gl_shader_program *prog,
                     unsigned num_tfeedback_decls,
                     tfeedback_decl *tfeedback_decls)
{
   bool separate_attribs_mode =
      prog->TransformFeedback.BufferMode == GL_SEPARATE_ATTRIBS;

   ralloc_free(prog->LinkedTransformFeedback.Varyings);
   ralloc_free(prog->LinkedTransformFeedback.Outputs);

   memset(&prog->LinkedTransformFeedback, 0,
          sizeof(prog->LinkedTransformFeedback));

   prog->LinkedTransformFeedback.Varyings =
      rzalloc_array(prog,
		    struct gl_transform_feedback_varying_info,
		    num_tfeedback_decls);

   unsigned num_outputs = 0;
   for (unsigned i = 0; i < num_tfeedback_decls; ++i)
      num_outputs += tfeedback_decls[i].get_num_outputs();

   prog->LinkedTransformFeedback.Outputs =
      rzalloc_array(prog,
                    struct gl_transform_feedback_output,
                    num_outputs);

   unsigned num_buffers = 0;

   if (separate_attribs_mode) {
      /* GL_SEPARATE_ATTRIBS */
      for (unsigned i = 0; i < num_tfeedback_decls; ++i) {
         if (!tfeedback_decls[i].store(ctx, prog, &prog->LinkedTransformFeedback,
                                       num_buffers, num_outputs))
            return false;

         num_buffers++;
      }
   }
   else {
      /* GL_INVERLEAVED_ATTRIBS */
      for (unsigned i = 0; i < num_tfeedback_decls; ++i) {
         if (tfeedback_decls[i].is_next_buffer_separator()) {
            num_buffers++;
            continue;
         }

         if (!tfeedback_decls[i].store(ctx, prog,
                                       &prog->LinkedTransformFeedback,
                                       num_buffers, num_outputs))
            return false;
      }
      num_buffers++;
   }

   assert(prog->LinkedTransformFeedback.NumOutputs == num_outputs);

   prog->LinkedTransformFeedback.NumBuffers = num_buffers;
   return true;
}

/**
 * Store the gl_FragDepth layout in the gl_shader_program struct.
 */
static void
store_fragdepth_layout(struct gl_shader_program *prog)
{
   if (prog->_LinkedShaders[MESA_SHADER_FRAGMENT] == NULL) {
      return;
   }

   struct exec_list *ir = prog->_LinkedShaders[MESA_SHADER_FRAGMENT]->ir;

   /* We don't look up the gl_FragDepth symbol directly because if
    * gl_FragDepth is not used in the shader, it's removed from the IR.
    * However, the symbol won't be removed from the symbol table.
    *
    * We're only interested in the cases where the variable is NOT removed
    * from the IR.
    */
   foreach_list(node, ir) {
      ir_variable *const var = ((ir_instruction *) node)->as_variable();

      if (var == NULL || var->mode != ir_var_out) {
         continue;
      }

      if (strcmp(var->name, "gl_FragDepth") == 0) {
         switch (var->depth_layout) {
         case ir_depth_layout_none:
            prog->FragDepthLayout = FRAG_DEPTH_LAYOUT_NONE;
            return;
         case ir_depth_layout_any:
            prog->FragDepthLayout = FRAG_DEPTH_LAYOUT_ANY;
            return;
         case ir_depth_layout_greater:
            prog->FragDepthLayout = FRAG_DEPTH_LAYOUT_GREATER;
            return;
         case ir_depth_layout_less:
            prog->FragDepthLayout = FRAG_DEPTH_LAYOUT_LESS;
            return;
         case ir_depth_layout_unchanged:
            prog->FragDepthLayout = FRAG_DEPTH_LAYOUT_UNCHANGED;
            return;
         default:
            assert(0);
            return;
         }
      }
   }
}

/**
 * Validate the resources used by a program versus the implementation limits
 */
static bool
check_resources(struct gl_context *ctx, struct gl_shader_program *prog)
{
   static const char *const shader_names[MESA_SHADER_TYPES] = {
      "vertex", "fragment", "geometry"
   };

   const unsigned max_samplers[MESA_SHADER_TYPES] = {
      ctx->Const.MaxVertexTextureImageUnits,
      ctx->Const.MaxTextureImageUnits,
      ctx->Const.MaxGeometryTextureImageUnits
   };

   const unsigned max_uniform_components[MESA_SHADER_TYPES] = {
      ctx->Const.VertexProgram.MaxUniformComponents,
      ctx->Const.FragmentProgram.MaxUniformComponents,
      0          /* FINISHME: Geometry shaders. */
   };

   const unsigned max_uniform_blocks[MESA_SHADER_TYPES] = {
      ctx->Const.VertexProgram.MaxUniformBlocks,
      ctx->Const.FragmentProgram.MaxUniformBlocks,
      ctx->Const.GeometryProgram.MaxUniformBlocks,
   };

   for (unsigned i = 0; i < MESA_SHADER_TYPES; i++) {
      struct gl_shader *sh = prog->_LinkedShaders[i];

      if (sh == NULL)
	 continue;

      if (sh->num_samplers > max_samplers[i]) {
	 linker_error(prog, "Too many %s shader texture samplers",
		      shader_names[i]);
      }

      if (sh->num_uniform_components > max_uniform_components[i]) {
         if (ctx->Const.GLSLSkipStrictMaxUniformLimitCheck) {
            linker_warning(prog, "Too many %s shader uniform components, "
                           "but the driver will try to optimize them out; "
                           "this is non-portable out-of-spec behavior\n",
                           shader_names[i]);
         } else {
            linker_error(prog, "Too many %s shader uniform components",
                         shader_names[i]);
         }
      }
   }

   unsigned blocks[MESA_SHADER_TYPES] = {0};
   unsigned total_uniform_blocks = 0;

   for (unsigned i = 0; i < prog->NumUniformBlocks; i++) {
      for (unsigned j = 0; j < MESA_SHADER_TYPES; j++) {
	 if (prog->UniformBlockStageIndex[j][i] != -1) {
	    blocks[j]++;
	    total_uniform_blocks++;
	 }
      }

      if (total_uniform_blocks > ctx->Const.MaxCombinedUniformBlocks) {
	 linker_error(prog, "Too many combined uniform blocks (%d/%d)",
		      prog->NumUniformBlocks,
		      ctx->Const.MaxCombinedUniformBlocks);
      } else {
	 for (unsigned i = 0; i < MESA_SHADER_TYPES; i++) {
	    if (blocks[i] > max_uniform_blocks[i]) {
	       linker_error(prog, "Too many %s uniform blocks (%d/%d)",
			    shader_names[i],
			    blocks[i],
			    max_uniform_blocks[i]);
	       break;
	    }
	 }
      }
   }

   return prog->LinkStatus;
}

void
link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
{
   tfeedback_decl *tfeedback_decls = NULL;
   unsigned num_tfeedback_decls = prog->TransformFeedback.NumVarying;

   void *mem_ctx = ralloc_context(NULL); // temporary linker context

   prog->LinkStatus = false;
   prog->Validated = false;
   prog->_Used = false;

   ralloc_free(prog->InfoLog);
   prog->InfoLog = ralloc_strdup(NULL, "");

   ralloc_free(prog->UniformBlocks);
   prog->UniformBlocks = NULL;
   prog->NumUniformBlocks = 0;
   for (int i = 0; i < MESA_SHADER_TYPES; i++) {
      ralloc_free(prog->UniformBlockStageIndex[i]);
      prog->UniformBlockStageIndex[i] = NULL;
   }

   /* Separate the shaders into groups based on their type.
    */
   struct gl_shader **vert_shader_list;
   unsigned num_vert_shaders = 0;
   struct gl_shader **frag_shader_list;
   unsigned num_frag_shaders = 0;

   vert_shader_list = (struct gl_shader **)
      calloc(2 * prog->NumShaders, sizeof(struct gl_shader *));
   frag_shader_list =  &vert_shader_list[prog->NumShaders];

   unsigned min_version = UINT_MAX;
   unsigned max_version = 0;
   const bool is_es_prog =
      (prog->NumShaders > 0 && prog->Shaders[0]->IsES) ? true : false;
   for (unsigned i = 0; i < prog->NumShaders; i++) {
      min_version = MIN2(min_version, prog->Shaders[i]->Version);
      max_version = MAX2(max_version, prog->Shaders[i]->Version);

      if (prog->Shaders[i]->IsES != is_es_prog) {
	 linker_error(prog, "all shaders must use same shading "
		      "language version\n");
	 goto done;
      }

      switch (prog->Shaders[i]->Type) {
      case GL_VERTEX_SHADER:
	 vert_shader_list[num_vert_shaders] = prog->Shaders[i];
	 num_vert_shaders++;
	 break;
      case GL_FRAGMENT_SHADER:
	 frag_shader_list[num_frag_shaders] = prog->Shaders[i];
	 num_frag_shaders++;
	 break;
      case GL_GEOMETRY_SHADER:
	 /* FINISHME: Support geometry shaders. */
	 assert(prog->Shaders[i]->Type != GL_GEOMETRY_SHADER);
	 break;
      }
   }

   /* Previous to GLSL version 1.30, different compilation units could mix and
    * match shading language versions.  With GLSL 1.30 and later, the versions
    * of all shaders must match.
    *
    * GLSL ES has never allowed mixing of shading language versions.
    */
   if ((is_es_prog || max_version >= 130)
       && min_version != max_version) {
      linker_error(prog, "all shaders must use same shading "
		   "language version\n");
      goto done;
   }

   prog->Version = max_version;
   prog->IsES = is_es_prog;

   for (unsigned int i = 0; i < MESA_SHADER_TYPES; i++) {
      if (prog->_LinkedShaders[i] != NULL)
	 ctx->Driver.DeleteShader(ctx, prog->_LinkedShaders[i]);

      prog->_LinkedShaders[i] = NULL;
   }

   /* Link all shaders for a particular stage and validate the result.
    */
   if (num_vert_shaders > 0) {
      gl_shader *const sh =
	 link_intrastage_shaders(mem_ctx, ctx, prog, vert_shader_list,
				 num_vert_shaders);

      if (sh == NULL)
	 goto done;

      if (!validate_vertex_shader_executable(prog, sh))
	 goto done;

      _mesa_reference_shader(ctx, &prog->_LinkedShaders[MESA_SHADER_VERTEX],
			     sh);
   }

   if (num_frag_shaders > 0) {
      gl_shader *const sh =
	 link_intrastage_shaders(mem_ctx, ctx, prog, frag_shader_list,
				 num_frag_shaders);

      if (sh == NULL)
	 goto done;

      if (!validate_fragment_shader_executable(prog, sh))
	 goto done;

      _mesa_reference_shader(ctx, &prog->_LinkedShaders[MESA_SHADER_FRAGMENT],
			     sh);
   }

   /* Here begins the inter-stage linking phase.  Some initial validation is
    * performed, then locations are assigned for uniforms, attributes, and
    * varyings.
    */
   if (cross_validate_uniforms(prog)) {
      unsigned prev;

      for (prev = 0; prev < MESA_SHADER_TYPES; prev++) {
	 if (prog->_LinkedShaders[prev] != NULL)
	    break;
      }

      /* Validate the inputs of each stage with the output of the preceding
       * stage.
       */
      for (unsigned i = prev + 1; i < MESA_SHADER_TYPES; i++) {
	 if (prog->_LinkedShaders[i] == NULL)
	    continue;

	 if (!cross_validate_outputs_to_inputs(prog,
					       prog->_LinkedShaders[prev],
					       prog->_LinkedShaders[i]))
	    goto done;

	 prev = i;
      }

      prog->LinkStatus = true;
   }

   /* Implement the GLSL 1.30+ rule for discard vs infinite loops Do
    * it before optimization because we want most of the checks to get
    * dropped thanks to constant propagation.
    *
    * This rule also applies to GLSL ES 3.00.
    */
   if (max_version >= (is_es_prog ? 300 : 130)) {
      struct gl_shader *sh = prog->_LinkedShaders[MESA_SHADER_FRAGMENT];
      if (sh) {
	 lower_discard_flow(sh->ir);
      }
   }

   if (!interstage_cross_validate_uniform_blocks(prog))
      goto done;

   /* Do common optimization before assigning storage for attributes,
    * uniforms, and varyings.  Later optimization could possibly make
    * some of that unused.
    */
   for (unsigned i = 0; i < MESA_SHADER_TYPES; i++) {
      if (prog->_LinkedShaders[i] == NULL)
	 continue;

      detect_recursion_linked(prog, prog->_LinkedShaders[i]->ir);
      if (!prog->LinkStatus)
	 goto done;

      if (ctx->ShaderCompilerOptions[i].LowerClipDistance) {
         lower_clip_distance(prog->_LinkedShaders[i]);
      }

      unsigned max_unroll = ctx->ShaderCompilerOptions[i].MaxUnrollIterations;

      while (do_common_optimization(prog->_LinkedShaders[i]->ir, true, false, max_unroll))
	 ;
   }

   /* Mark all generic shader inputs and outputs as unpaired. */
   if (prog->_LinkedShaders[MESA_SHADER_VERTEX] != NULL) {
      link_invalidate_variable_locations(
            prog->_LinkedShaders[MESA_SHADER_VERTEX],
            VERT_ATTRIB_GENERIC0, VERT_RESULT_VAR0);
   }
   /* FINISHME: Geometry shaders not implemented yet */
   if (prog->_LinkedShaders[MESA_SHADER_FRAGMENT] != NULL) {
      link_invalidate_variable_locations(
            prog->_LinkedShaders[MESA_SHADER_FRAGMENT],
            FRAG_ATTRIB_VAR0, FRAG_RESULT_DATA0);
   }

   /* FINISHME: The value of the max_attribute_index parameter is
    * FINISHME: implementation dependent based on the value of
    * FINISHME: GL_MAX_VERTEX_ATTRIBS.  GL_MAX_VERTEX_ATTRIBS must be
    * FINISHME: at least 16, so hardcode 16 for now.
    */
   if (!assign_attribute_or_color_locations(prog, MESA_SHADER_VERTEX, 16)) {
      goto done;
   }

   if (!assign_attribute_or_color_locations(prog, MESA_SHADER_FRAGMENT, MAX2(ctx->Const.MaxDrawBuffers, ctx->Const.MaxDualSourceDrawBuffers))) {
      goto done;
   }

   unsigned prev;
   for (prev = 0; prev < MESA_SHADER_TYPES; prev++) {
      if (prog->_LinkedShaders[prev] != NULL)
	 break;
   }

   if (num_tfeedback_decls != 0) {
      /* From GL_EXT_transform_feedback:
       *   A program will fail to link if:
       *
       *   * the <count> specified by TransformFeedbackVaryingsEXT is
       *     non-zero, but the program object has no vertex or geometry
       *     shader;
       */
      if (prev >= MESA_SHADER_FRAGMENT) {
         linker_error(prog, "Transform feedback varyings specified, but "
                      "no vertex or geometry shader is present.");
         goto done;
      }

      tfeedback_decls = ralloc_array(mem_ctx, tfeedback_decl,
                                     prog->TransformFeedback.NumVarying);
      if (!parse_tfeedback_decls(ctx, prog, mem_ctx, num_tfeedback_decls,
                                 prog->TransformFeedback.VaryingNames,
                                 tfeedback_decls))
         goto done;
   }

   for (unsigned i = prev + 1; i < MESA_SHADER_TYPES; i++) {
      if (prog->_LinkedShaders[i] == NULL)
	 continue;

      if (!assign_varying_locations(
				    ctx, mem_ctx, prog, prog->_LinkedShaders[prev], prog->_LinkedShaders[i],
             i == MESA_SHADER_FRAGMENT ? num_tfeedback_decls : 0,
             tfeedback_decls))
	 goto done;

      prev = i;
   }

   if (prev != MESA_SHADER_FRAGMENT && num_tfeedback_decls != 0) {
      /* There was no fragment shader, but we still have to assign varying
       * locations for use by transform feedback.
       */
      if (!assign_varying_locations(
				    ctx, mem_ctx, prog, prog->_LinkedShaders[prev], NULL, num_tfeedback_decls,
             tfeedback_decls))
         goto done;
   }

   if (!store_tfeedback_info(ctx, prog, num_tfeedback_decls, tfeedback_decls))
      goto done;

   if (prog->_LinkedShaders[MESA_SHADER_VERTEX] != NULL) {
      demote_shader_inputs_and_outputs(prog->_LinkedShaders[MESA_SHADER_VERTEX],
				       ir_var_out);

      /* Eliminate code that is now dead due to unused vertex outputs being
       * demoted.
       */
      while (do_dead_code(prog->_LinkedShaders[MESA_SHADER_VERTEX]->ir, false))
	 ;
   }

   if (prog->_LinkedShaders[MESA_SHADER_GEOMETRY] != NULL) {
      gl_shader *const sh = prog->_LinkedShaders[MESA_SHADER_GEOMETRY];

      demote_shader_inputs_and_outputs(sh, ir_var_in);
      demote_shader_inputs_and_outputs(sh, ir_var_inout);
      demote_shader_inputs_and_outputs(sh, ir_var_out);

      /* Eliminate code that is now dead due to unused geometry outputs being
       * demoted.
       */
      while (do_dead_code(prog->_LinkedShaders[MESA_SHADER_GEOMETRY]->ir, false))
	 ;
   }

   if (prog->_LinkedShaders[MESA_SHADER_FRAGMENT] != NULL) {
      gl_shader *const sh = prog->_LinkedShaders[MESA_SHADER_FRAGMENT];

      demote_shader_inputs_and_outputs(sh, ir_var_in);

      /* Eliminate code that is now dead due to unused fragment inputs being
       * demoted.  This shouldn't actually do anything other than remove
       * declarations of the (now unused) global variables.
       */
      while (do_dead_code(prog->_LinkedShaders[MESA_SHADER_FRAGMENT]->ir, false))
	 ;
   }

   update_array_sizes(prog);
   link_assign_uniform_locations(prog);
   store_fragdepth_layout(prog);

   if (!check_resources(ctx, prog))
      goto done;

   /* OpenGL ES requires that a vertex shader and a fragment shader both be
    * present in a linked program.  By checking prog->IsES, we also
    * catch the GL_ARB_ES2_compatibility case.
    */
   if (!prog->InternalSeparateShader &&
       (ctx->API == API_OPENGLES2 || prog->IsES)) {
      if (prog->_LinkedShaders[MESA_SHADER_VERTEX] == NULL) {
	 linker_error(prog, "program lacks a vertex shader\n");
      } else if (prog->_LinkedShaders[MESA_SHADER_FRAGMENT] == NULL) {
	 linker_error(prog, "program lacks a fragment shader\n");
      }
   }

   /* FINISHME: Assign fragment shader output locations. */

done:
   free(vert_shader_list);

   for (unsigned i = 0; i < MESA_SHADER_TYPES; i++) {
      if (prog->_LinkedShaders[i] == NULL)
	 continue;

      /* Retain any live IR, but trash the rest. */
      reparent_ir(prog->_LinkedShaders[i]->ir, prog->_LinkedShaders[i]->ir);

      /* The symbol table in the linked shaders may contain references to
       * variables that were removed (e.g., unused uniforms).  Since it may
       * contain junk, there is no possible valid use.  Delete it and set the
       * pointer to NULL.
       */
      delete prog->_LinkedShaders[i]->symbols;
      prog->_LinkedShaders[i]->symbols = NULL;
   }

   ralloc_free(mem_ctx);
}