aboutsummaryrefslogtreecommitdiff
path: root/build2/cc/lexer.cxx
blob: 779519216c74e9b731eb91432c4e2428457099c2 (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
// file      : build2/cc/lexer.cxx -*- C++ -*-
// copyright : Copyright (c) 2014-2019 Code Synthesis Ltd
// license   : MIT; see accompanying LICENSE file

#include <build2/cc/lexer.hxx>

using namespace std;
using namespace butl;

// bit 0 - identifier character (_0-9A-Ba-b).
//
static const uint8_t char_flags[256] =
//0    1    2    3    4    5    6    7      8    9    A    B    C    D    E    F
{
  0,   0,   0,   0,   0,   0,   0,   0,     0,   0,   0,   0,   0,   0,   0,   0, // 0
  0,   0,   0,   0,   0,   0,   0,   0,     0,   0,   0,   0,   0,   0,   0,   0, // 1
  0,   0,   0,   0,   0,   0,   0,   0,     0,   0,   0,   0,   0,   0,   0,   0, // 2
  1,   1,   1,   1,   1,   1,   1,   1,     1,   1,   0,   0,   0,   0,   0,   0, // 3
  0,   1,   1,   1,   1,   1,   1,   1,     1,   1,   1,   1,   1,   1,   1,   1, // 4
  1,   1,   1,   1,   1,   1,   1,   1,     1,   1,   1,   0,   0,   0,   0,   1, // 5
  0,   1,   1,   1,   1,   1,   1,   1,     1,   1,   1,   1,   1,   1,   1,   1, // 6
  1,   1,   1,   1,   1,   1,   1,   1,     1,   1,   1,   0,   0,   0,   0,   0, // 7

  // 128-255
  0,0,0,0,0,0,0,0,  0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,  0,0,0,0,0,0,0,0,
  0,0,0,0,0,0,0,0,  0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,  0,0,0,0,0,0,0,0,
  0,0,0,0,0,0,0,0,  0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,  0,0,0,0,0,0,0,0,
  0,0,0,0,0,0,0,0,  0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,  0,0,0,0,0,0,0,0
};

// Diagnostics plumbing.
//
namespace butl // ADL
{
  inline build2::location
  get_location (const butl::char_scanner::xchar& c, const void* data)
  {
    using namespace build2;

    assert (data != nullptr); // E.g., must be &lexer::name_.
    return location (static_cast<const path*> (data), c.line, c.column);
  }
}

namespace build2
{
  namespace cc
  {
    auto lexer::
    peek (bool e) -> xchar
    {
      if (unget_)
        return ungetc_;

      if (unpeek_)
        return unpeekc_;

      xchar c (base::peek ());

      if (e && c == '\\')
      {
        get (c);
        xchar p (base::peek ());

        // Handle Windows CRLF sequence. Similar to char_scanner, we treat a
        // single CR as if it was followed by LF and also collapse multiple
        // CRs.
        //
        while (p == '\r')
        {
          get (p);
          p = base::peek ();

          if (p == '\n')
            break;

          // Pretend '\n' was there and recurse.
          //
          if (p != '\r')
            return peek (e);
        }

        if (p == '\n')
        {
          get (p);
          return peek (e); // Recurse.
        }

        // Save in the unpeek buffer so that it is returned on the subsequent
        // calls to peek() (until get()).
        //
        unpeek_ = true;
        unpeekc_ = c;
      }

      return c;
    }

    inline auto lexer::
    get (bool e) -> xchar
    {
      if (unget_)
      {
        unget_ = false;
        return ungetc_;
      }
      else
      {
        xchar c (peek (e));
        get (c);
        return c;
      }
    }

    inline void lexer::
    get (const xchar& c)
    {
      // Increment the logical line similar to how base will increment the
      // physical (the column counts are the same).
      //
      if (log_line_ && c == '\n' && !unget_)
        ++*log_line_;

      base::get (c);
    }

    inline auto lexer::
    geth (bool e) -> xchar
    {
      xchar c (get (e));
      cs_.append (c);
      return c;
    }

    inline void lexer::
    geth (const xchar& c)
    {
      get (c);
      cs_.append (c);
    }

    using type = token_type;

    void lexer::
    next (token& t, xchar c, bool ignore_pp)
    {
      for (;; c = skip_spaces ())
      {
        t.file = log_file_;
        t.line = log_line_ ? *log_line_ : c.line;
        t.column = c.column;

        if (eos (c))
        {
          t.type = type::eos;
          return;
        }

        const location l (&name_, c.line, c.column);

        // Hash the token's line. The reason is debug info. In fact, doing
        // this will make quite a few "noop" changes (like adding a newline
        // anywhere in the source) cause the checksum change. But there
        // doesn't seem to be any way around it: the case where we benefit
        // from the precise change detection the most (development) is also
        // where we will most likely have debug info enable.
        //
        // Note that in order not to make this completely useless we don't
        // hash the column. Even if it is part of the debug info, having it a
        // bit off shouldn't cause any significant mis-positioning. We also
        // don't hash the file path for each token instead only hashing it
        // when changed with the #line directive (as well as in the
        // constructor for the initial path).
        //
        cs_.append (t.line);
        cs_.append (c);

        switch (c)
        {
          // Preprocessor lines.
          //
        case '#':
          {
            // It is tempting to simply scan until the newline ignoring
            // anything in between. However, these lines can start a
            // multi-line C-style comment. So we have to tokenize them (and
            // hash the data for each token).
            //
            // Note that this may not work for things like #error that can
            // contain pretty much anything. Also note that lines that start
            // with '#' can contain '#' further down. In this case we need to
            // be careful not to recurse (and consume multiple newlines). Thus
            // the ignore_pp flag.
            //
            // Finally, to support diagnostics properly we need to recognize
            // #line directives.
            //
            if (ignore_pp)
            {
              for (bool first (true);;)
              {
                // Note that we keep using the passed token for buffers.
                //
                c = skip_spaces (false); // Stop at newline.

                if (eos (c) || c == '\n')
                  break;

                if (first)
                {
                  first = false;

                  // Recognize #line and its shorthand version:
                  //
                  // #line <integer> [<string literal>] ...
                  // #     <integer> [<string literal>] ...
                  //
                  // Also diagnose #include while at it.
                  //
                  if (!(c >= '0' && c <= '9'))
                  {
                    next (t, c, false);

                    if (t.type == type::identifier)
                    {
                      if (t.value == "include")
                        fail (l) << "unexpected #include directive";
                      else if (t.value != "line")
                        continue;
                    }
                    else
                      continue;

                    if (t.type != type::identifier || t.value != "line")
                      continue;

                    c = skip_spaces (false);

                    if (!(c >= '0' && c <= '9'))
                      fail (c) << "line number expected after #line directive";
                  }

                  // Ok, this is #line and next comes the line number.
                  //
                  line_directive (t, c);
                  continue; // Parse the tail, if any.
                }

                next (t, c, false);
              }
              break;
            }
            else
            {
              t.type = type::punctuation;
              return;
            }
          }
          // Single-letter punctuation.
          //
        case ';': t.type = type::semi;    return;
        case '{': t.type = type::lcbrace; return;
        case '}': t.type = type::rcbrace; return;
          // Other single-letter punctuation.
          //
        case '(':
        case ')':
        case '[':
        case ']':
        case ',':
        case '?':
        case '~':
        case '\\': t.type = type::punctuation; return;
          // Potentially multi-letter punctuation.
          //
        case '.': // . .* .<N> ...
          {
            xchar p (peek ());

            if (p == '*')
            {
              geth (p);
              t.type = type::punctuation;
              return;
            }
            else if (p >= '0' && p <= '9')
            {
              number_literal (t, c);
              return;
            }
            else if (p == '.')
            {
              get (p);

              xchar q (peek ());
              if (q == '.')
              {
                cs_.append (p);

                geth (q);
                t.type = type::punctuation;
                return;
              }
              unget (p);
              // Fall through.
            }

            t.type = type::dot;
            return;
          }
        case '=': // = ==
        case '!': // ! !=
        case '*': // * *=
        case '/': // / /=   (/* and // handled by skip_spaced() above)
        case '%': // % %=
        case '^': // ^ ^=
          {
            xchar p (peek ());

            if (p == '=')
              geth (p);

            t.type = type::punctuation;
            return;
          }
        case '<': // < <= << <<=
        case '>': // > >= >> >>=
          {
            xchar p (peek ());

            if (p == c)
            {
              geth (p);
              if ((p = peek ()) == '=')
                geth (p);
              t.type = type::punctuation;
            }
            else if (p == '=')
            {
              geth (p);
              t.type = type::punctuation;
            }
            else
              t.type = (c == '<' ? type::less : type::greater);

            return;
          }
        case '+': // + ++ +=
        case '-': // - -- -= -> ->*
          {
            xchar p (peek ());

            if (p == c || p == '=')
              geth (p);
            else if (c == '-' && p == '>')
            {
              geth (p);
              if ((p = peek ()) == '*')
                geth (p);
            }

            t.type = type::punctuation;
            return;
          }
        case '&': // & && &=
        case '|': // | || |=
          {
            xchar p (peek ());

            if (p == c || p == '=')
              geth (p);

            t.type = type::punctuation;
            return;
          }
        case ':': // : ::
          {
            xchar p (peek ());

            if (p == ':')
              geth (p);

            t.type = type::punctuation;
            return;
          }
          // Number (and also .<N> above).
          //
        case '0':
        case '1':
        case '2':
        case '3':
        case '4':
        case '5':
        case '6':
        case '7':
        case '8':
        case '9':
          {
            number_literal (t, c);
            return;
          }
          // Char/string literal, identifier, or other (\, $, @, `).
          //
        default:
          {
            bool raw (false); // Raw string literal.

            // Note: known not to be a digit (see above).
            //
            if (char_flags[static_cast<uint8_t> (c)] & 0x01)
            {
              // This smells a little: we know skip_spaces() did not peek at
              // the next character because this is not '/'. Which means the
              // position in the stream must be of this character + 1.
              //
              t.position = buf_->tellg () - 1;

              string& id (t.value);
              id = c;

              while (char_flags[static_cast<uint8_t> (c = peek ())] & 0x01)
              {
                geth (c);
                id += c;

                // Direct buffer scan. Note that we always follow up with the
                // normal peek() call which may load the next chunk, handle
                // line continuations, etc. In other words, the end of the
                // "raw" scan doesn't necessarily mean the end.
                //
                const char* b (gptr_);
                const char* p (b);

                for (const char* e (egptr_);
                     p != e && char_flags[static_cast<uint8_t> (*p)] & 0x01;
                     ++p) ;

                // Unrolling this loop doesn't make a difference.
                //
                // for (const char* e (egptr_ - 4); p < e; p += 4)
                // {
                //   uint8_t c;
                //
                //  c = static_cast<uint8_t> (p[0]);
                //  if (!(char_flags[c] & 0x01)) break;
                //
                //  c = static_cast<uint8_t> (p[1]);
                //  if (!(char_flags[c] & 0x01)) {p += 1; break;}
                //
                //  c = static_cast<uint8_t> (p[2]);
                //  if (!(char_flags[c] & 0x01)) {p += 2; break;}
                //
                //  c = static_cast<uint8_t> (p[3]);
                //  if (!(char_flags[c] & 0x01)) {p += 3; break;}
                // }

                size_t n (p - b);
                id.append (b, n); cs_.append (b, n);
                gptr_ = p; buf_->gbump (static_cast<int> (n)); column += n;
              }

              // If the following character is a quote, see if the identifier
              // is one of the literal prefixes.
              //
              if (c == '\'' || c == '\"')
              {
                size_t n (id.size ()), i (0);
                switch (id[0])
                {
                case 'u':
                  {
                    if (n > 1 && id[1] == '8')
                      ++i;
                  }
                  // Fall through.
                case 'L':
                case 'U':
                  {
                    ++i;

                    if (c == '\"' && n > i && id[i] == 'R')
                    {
                      ++i;
                      raw = true;
                    }
                    break;
                  }
                case 'R':
                  {
                    if (c == '\"')
                    {
                      ++i;
                      raw = true;
                    }
                    break;
                  }
                }

                if (i == n) // All characters "consumed".
                {
                  geth (c);
                  id.clear ();
                }
              }

              if (!id.empty ())
              {
                t.type = type::identifier;
                return;
              }
            }

            switch (c)
            {
            case '\'':
              {
                char_literal (t, c);
                return;
              }
            case '\"':
              {
                if (raw)
                  raw_string_literal (t, c);
                else
                  string_literal (t, c);
                return;
              }
            default:
              {
                t.type = type::other;
                return;
              }
            }
          }
        }
      }
    }

    void lexer::
    number_literal (token& t, xchar c)
    {
      // note: c is hashed

      // A number (integer or floating point literal) can:
      //
      // 1. Start with a dot (which must be followed by a digit, e.g., .123).
      //
      // 2. Can have a radix prefix (0b101, 0123, 0X12AB).
      //
      // 3. Can have an exponent (1e10, 0x1.p-10, 1.).
      //
      // 4. Digits can be separated with ' (123'456, 0xff00'00ff).
      //
      // 5. End with a built-in or user defined literal (123f, 123UL, 123_X)
      //
      // Quoting from GCC's preprocessor documentation:
      //
      // "Formally preprocessing numbers begin with an optional period, a
      // required decimal digit, and then continue with any sequence of
      // letters, digits, underscores, periods, and exponents. Exponents are
      // the two-character sequences 'e+', 'e-', 'E+', 'E-', 'p+', 'p-', 'P+',
      // and 'P-'."
      //
      // So it looks like a "C++ number" is then any unseparated (with
      // whitespace or punctuation) sequence of those plus '. The only mildly
      // tricky part is then to recognize +/- as being part of the exponent.
      //
      while (!eos ((c = peek ())))
      {
        switch (c)
        {
          // All the whitespace, punctuation, and other characters that end
          // the number.
          //
        case ' ':
        case '\n':
        case '\t':
        case '\r':
        case '\f':
        case '\v':

        case '#':
        case ';':
        case '{':
        case '}':
        case '(':
        case ')':
        case '[':
        case ']':
        case ',':
        case '?':
        case '~':
        case '=':
        case '!':
        case '*':
        case '/':
        case '%':
        case '^':
        case '>':
        case '<':
        case '&':
        case '|':
        case ':':
        case '+': // The exponent case is handled below.
        case '-': // The exponent case is handled below.
        case '"':
        case '\\':

        case '@':
        case '$':
        case '`':
          break;

          // Recognize +/- after the exponent.
          //
        case 'e':
        case 'E':
        case 'p':
        case 'P':
          {
            geth (c);
            c = peek ();
            if (c == '+' || c == '-')
              geth (c);
            continue;
          }

        case '_':
        case '.':
        case '\'':
        default: // Digits and letters.
          {
            geth (c);
            continue;
          }
        }

        break;
      }

      t.type = type::number;
    }

    void lexer::
    char_literal (token& t, xchar c)
    {
      // note: c is hashed

      const location l (&name_, c.line, c.column);

      for (char p (c);;) // Previous character (see below).
      {
        c = geth ();

        if (eos (c) || c == '\n')
          fail (l) << "unterminated character literal";

        if (c == '\'' && p != '\\')
          break;

        // Keep track of \\-escapings so we don't confuse them with \', as in
        // '\\'.
        //
        p = (c == '\\' && p == '\\') ? '\0' : static_cast<char> (c);
      }

      // See if we have a user-defined suffix (which is an identifier).
      //
      if ((c = peek ()) == '_' || alpha (c))
        literal_suffix (c);

      t.type = type::character;
    }

    void lexer::
    string_literal (token& t, xchar c)
    {
      // note: c is hashed

      const location l (&name_, c.line, c.column);

      for (char p (c);;) // Previous character (see below).
      {
        c = geth ();

        if (eos (c) || c == '\n')
          fail (l) << "unterminated string literal";

        if (c == '\"' && p != '\\')
          break;

        // Keep track of \\-escapings so we don't confuse them with \", as in
        // "\\".
        //
        p = (c == '\\' && p == '\\') ? '\0' : static_cast<char> (c);

        // Direct buffer scan.
        //
        if (p != '\\')
        {
          const char* b (gptr_);
          const char* e (egptr_);
          const char* p (b);

          for (char c;
               p != e && (c = *p) != '\"' && c != '\\' && c != '\n';
               ++p) ;

          size_t n (p - b);
          cs_.append (b, n);
          gptr_ = p; buf_->gbump (static_cast<int> (n)); column += n;
        }
      }

      // See if we have a user-defined suffix (which is an identifier).
      //
      if ((c = peek ()) == '_' || alpha (c))
        literal_suffix (c);

      t.type = type::string;
    }

    void lexer::
    raw_string_literal (token& t, xchar c)
    {
      // note: c is hashed

      // The overall form is:
      //
      // R"<delimiter>(<raw_characters>)<delimiter>"
      //
      // Where <delimiter> is a potentially-empty character sequence made of
      // any source character but parentheses, backslash and spaces. It can be
      // at most 16 characters long.
      //
      // Note that the <raw_characters> are not processed in any way, not even
      // for line continuations.
      //
      const location l (&name_, c.line, c.column);

      // As a first step, parse the delimiter (including the openning paren).
      //
      string d (1, ')');

      for (;;)
      {
        c = geth ();

        if (eos (c) || c == '\"' || c == ')' || c == '\\' || c == ' ')
          fail (l) << "invalid raw string literal";

        if (c == '(')
          break;

        d += c;
      }

      d += '"';

      // Now parse the raw characters while trying to match the closing
      // delimiter.
      //
      for (size_t i (0);;) // Position to match in d.
      {
        c = geth (false); // No newline escaping.

        if (eos (c)) // Note: newline is ok.
          fail (l) << "invalid raw string literal";

        if (c != d[i] && i != 0) // Restart from the beginning.
          i = 0;

        if (c == d[i])
        {
          if (++i == d.size ())
            break;
        }
      }

      // See if we have a user-defined suffix (which is an identifier).
      //
      if ((c = peek ()) == '_' || alpha (c))
        literal_suffix (c);

      t.type = type::string;
    }

    void lexer::
    literal_suffix (xchar c)
    {
      // note: c is unhashed

      // Parse a user-defined literal suffix identifier.
      //
      for (geth (c); (c = peek ()) == '_' || alnum (c); geth (c)) ;
    }

    void lexer::
    line_directive (token& t, xchar c)
    {
      // enter: first digit of the line number
      // leave: last character of the line number or file string
      // note:  c is unhashed

      // If our number and string tokens contained the literal values, then we
      // could have used that. However, we ignore the value (along with escape
      // processing, etc), for performance. Let's keep it that way and instead
      // handle it ourselves.
      //
      // Note also that we are not hashing these at the character level
      // instead hashing the switch to a new file path below and leaving the
      // line number to the token line hashing.
      //
      {
        string& s (t.value);

        for (s = c; (c = peek ()) >= '0' && c <= '9'; get (c))
          s += c;

        // The newline that ends the directive will increment the logical line
        // so subtract one to compensate. Note: can't be 0 and shouldn't throw
        // for valid lines.
        //
        log_line_ = stoull (s.c_str ()) - 1;
      }

      // See if we have the file.
      //
      c = skip_spaces (false);

      if (c == '\"')
      {
        const location l (&name_, c.line, c.column);

        // It is common to have a large number of #line directives that don't
        // change the file (they seem to be used to track macro locations or
        // some such). So we are going to optimize for this by comparing the
        // current path to what's in #line.
        //
        string& s (tmp_file_);
        s.clear ();

        for (char p ('\0'); p != '\"'; ) // Previous character.
        {
          c = get ();

          if (eos (c) || c == '\n')
            fail (l) << "unterminated string literal";

          // Handle escapes.
          //
          if (p == '\\')
          {
            p = '\0'; // Clear so we don't confuse \" and \\".

            // We only handle what can reasonably be expected in a file name.
            //
            switch (c)
            {
            case '\\':
            case '\'':
            case '\"': break; // Add as is.
            default:
              fail (c) << "unsupported escape sequence in #line directive";
            }
          }
          else
          {
            p = c;

            switch (c)
            {
            case '\\':
            case '\"': continue;
            }
          }

          s += c;

          // Direct buffer scan.
          //
          if (p != '\\')
          {
            const char* b (gptr_);
            const char* e (egptr_);
            const char* p (b);

            for (char c;
                 p != e && (c = *p) != '\"' && c != '\\' && c != '\n';
                 ++p) ;

            size_t n (p - b);
            s.append (b, n);
            gptr_ = p; buf_->gbump (static_cast<int> (n)); column += n;
          }
        }

        if (log_file_.string () == s)
          return;

        // Swap the two string buffers.
        //
        {
          string r (move (log_file_).string ()); // Move string rep out.
          r.swap (s);
          log_file_ = path (move (r)); // Move back in.
        }

        // If the path is relative, then prefix it with the current working
        // directory. Failed that, we will end up with different checksums for
        // invocations from different directories.
        //
        // While this should work fine for normal cross-compilation, it's an
        // entirely different story for the emulated case (e.g., msvc-linux
        // where the preprocessed output contains absolute Windows paths). So
        // we try to sense if things look fishy and leave the path alone.
        //
        // Also detect special names like <built-in> and <command-line>. Plus
        // GCC sometimes adds what looks like working directory (has trailing
        // slash). So ignore that as well.
        //
        // We now switched to using absolute translation unit paths (because
        // of __FILE__/assert(); see compile.cxx for details). But we might
        // still need this logic when we try to calculate location-independent
        // hash for distributed compilation/caching. The idea is to only hash
        // the part starting from the project root which is immutable. Plus
        // we will need -ffile-prefix-map to deal with __FILE__.
        //
        if (!log_file_.to_directory ())
          cs_.append (log_file_.string ());
#if 0
        {
          using tr = path::traits;
          const string& f (log_file_.string ());

          if (f.find (':') != string::npos            ||
              (f.front () == '<' && f.back () == '>') ||
              log_file_.absolute ())
            cs_.append (f);
          else
          {
            // This gets complicated and slow: the path may contain '..' and
            // '.'  so strictly speaking we would need to normalize it.
            // Instead, we are going to handle leading '..'s ourselves (the
            // sane case) and ignore everything else (so if you have '..'  or
            // '.' somewhere in the middle, then things might not work
            // optimally for you).
            //
            const string& d (work.string ());

            // Iterate over leading '..' in f "popping" the corresponding
            // number of trailing components from d.
            //
            size_t fp (0);
            size_t dp (d.size () - 1);

            for (size_t p;; )
            {
              // Note that in file we recognize any directory separator, not
              // just of this platform (see note about emulation above).
              //
              if (f.compare (fp, 2, "..") != 0  ||
                  (f[fp + 2] != '/' && f[fp + 2] != '\\') || // Could be '\0'.
                  (p = tr::rfind_separator (d, dp)) == string::npos)
                break;

              fp += 3;
              dp = p - 1;
            }

            cs_.append (d.c_str (), dp + 1);
            cs_.append (tr::directory_separator); // Canonical in work.
            cs_.append (f.c_str () + fp);
          }
        }
#endif
      }
      else
        unget (c);
    }

    auto lexer::
    skip_spaces (bool nl) -> xchar
    {
      xchar c (get ());

      for (; !eos (c); c = get ())
      {
        switch (c)
        {
        case '\n':
          if (!nl) break;
          // Fall through.
        case ' ':
        case '\t':
        case '\r':
        case '\f':
        case '\v':
          {
            // Direct buffer scan.
            //
            const char* b (gptr_);
            const char* e (egptr_);
            const char* p (b);

            for (char c;
                 p != e && ((c = *p) == ' ' || c == '\t');
                 ++p) ;

            size_t n (p - b);
            gptr_ = p; buf_->gbump (static_cast<int> (n)); column += n;

            continue;
          }
        case '/':
          {
            xchar p (peek ());

            // C++ comment.
            //
            if (p == '/')
            {
              get (p);

              for (;;)
              {
                c = get ();
                if (c == '\n' || eos (c))
                  break;

                // Direct buffer scan.
                //
                const char* b (gptr_);
                const char* e (egptr_);
                const char* p (b);

                for (char c;
                     p != e && (c = *p) != '\n' && c != '\\';
                     ++p) ;

                size_t n (p - b);
                gptr_ = p; buf_->gbump (static_cast<int> (n)); column += n;
              }

              if (!nl)
                break;

              continue;
            }

            // C comment.
            //
            if (p == '*')
            {
              get (p);

              for (;;)
              {
                c = get ();

                if (eos (c))
                  fail (p) << "unterminated comment";

                if (c == '*' && (c = peek ()) == '/')
                {
                  get (c);
                  break;
                }

                // Direct buffer scan.
                //
                const char* b (gptr_);
                const char* e (egptr_);
                const char* p (b);

                for (char c;
                     p != e && (c = *p) != '*' && c != '\\';
                     ++p)
                {
                  if (c == '\n')
                  {
                    if (log_line_) ++*log_line_;
                    ++line;
                    column = 1;
                  }
                  else
                    ++column;
                }

                gptr_ = p; buf_->gbump (static_cast<int> (p - b));
              }
              continue;
            }
            break;
          }
        }
        break;
      }

      return c;
    }

    ostream&
    operator<< (ostream& o, const token& t)
    {
      switch (t.type)
      {
      case type::dot:         o << "'.'";                   break;
      case type::semi:        o << "';'";                   break;
      case type::less:        o << "'<'";                   break;
      case type::greater:     o << "'>'";                   break;
      case type::lcbrace:     o << "'{'";                   break;
      case type::rcbrace:     o << "'}'";                   break;
      case type::punctuation: o << "<punctuation>";         break;

      case type::identifier:  o << '\'' << t.value << '\''; break;

      case type::number:      o << "<number literal>";      break;
      case type::character:   o << "<char literal>";        break;
      case type::string:      o << "<string literal>";      break;

      case type::other:       o << "<other>";               break;
      case type::eos:         o << "<end of file>";         break;
      }

      return o;
    }
  }
}