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
|
\input texinfo
@setfilename cppinternals.info
@settitle The GNU C Preprocessor Internals
@ifinfo
@dircategory Programming
@direntry
* Cpplib: (cppinternals). Cpplib internals.
@end direntry
@end ifinfo
@c @smallbook
@c @cropmarks
@c @finalout
@setchapternewpage odd
@ifinfo
This file documents the internals of the GNU C Preprocessor.
Copyright 2000, 2001, 2002, 2004 Free Software Foundation, Inc.
Permission is granted to make and distribute verbatim copies of
this manual provided the copyright notice and this permission notice
are preserved on all copies.
@ignore
Permission is granted to process this file through Tex and print the
results, provided the printed document carries copying permission
notice identical to this one except for the removal of this paragraph
(this paragraph not being relevant to the printed manual).
@end ignore
Permission is granted to copy and distribute modified versions of this
manual under the conditions for verbatim copying, provided also that
the entire resulting derived work is distributed under the terms of a
permission notice identical to this one.
Permission is granted to copy and distribute translations of this manual
into another language, under the above conditions for modified versions.
@end ifinfo
@titlepage
@c @finalout
@title Cpplib Internals
@subtitle Last revised January 2002
@subtitle for GCC version 3.1
@author Neil Booth
@page
@vskip 0pt plus 1filll
@c man begin COPYRIGHT
Copyright @copyright{} 2000, 2001, 2002, 2004
Free Software Foundation, Inc.
Permission is granted to make and distribute verbatim copies of
this manual provided the copyright notice and this permission notice
are preserved on all copies.
Permission is granted to copy and distribute modified versions of this
manual under the conditions for verbatim copying, provided also that
the entire resulting derived work is distributed under the terms of a
permission notice identical to this one.
Permission is granted to copy and distribute translations of this manual
into another language, under the above conditions for modified versions.
@c man end
@end titlepage
@contents
@page
@node Top
@top
@chapter Cpplib---the GNU C Preprocessor
The GNU C preprocessor in GCC 3.x has been completely rewritten. It is
now implemented as a library, @dfn{cpplib}, so it can be easily shared between
a stand-alone preprocessor, and a preprocessor integrated with the C,
C++ and Objective-C front ends. It is also available for use by other
programs, though this is not recommended as its exposed interface has
not yet reached a point of reasonable stability.
The library has been written to be re-entrant, so that it can be used
to preprocess many files simultaneously if necessary. It has also been
written with the preprocessing token as the fundamental unit; the
preprocessor in previous versions of GCC would operate on text strings
as the fundamental unit.
This brief manual documents the internals of cpplib, and explains some
of the tricky issues. It is intended that, along with the comments in
the source code, a reasonably competent C programmer should be able to
figure out what the code is doing, and why things have been implemented
the way they have.
@menu
* Conventions:: Conventions used in the code.
* Lexer:: The combined C, C++ and Objective-C Lexer.
* Hash Nodes:: All identifiers are entered into a hash table.
* Macro Expansion:: Macro expansion algorithm.
* Token Spacing:: Spacing and paste avoidance issues.
* Line Numbering:: Tracking location within files.
* Guard Macros:: Optimizing header files with guard macros.
* Files:: File handling.
* Concept Index:: Index.
@end menu
@node Conventions
@unnumbered Conventions
@cindex interface
@cindex header files
cpplib has two interfaces---one is exposed internally only, and the
other is for both internal and external use.
The convention is that functions and types that are exposed to multiple
files internally are prefixed with @samp{_cpp_}, and are to be found in
the file @file{cpphash.h}. Functions and types exposed to external
clients are in @file{cpplib.h}, and prefixed with @samp{cpp_}. For
historical reasons this is no longer quite true, but we should strive to
stick to it.
We are striving to reduce the information exposed in @file{cpplib.h} to the
bare minimum necessary, and then to keep it there. This makes clear
exactly what external clients are entitled to assume, and allows us to
change internals in the future without worrying whether library clients
are perhaps relying on some kind of undocumented implementation-specific
behavior.
@node Lexer
@unnumbered The Lexer
@cindex lexer
@cindex newlines
@cindex escaped newlines
@section Overview
The lexer is contained in the file @file{cpplex.c}. It is a hand-coded
lexer, and not implemented as a state machine. It can understand C, C++
and Objective-C source code, and has been extended to allow reasonably
successful preprocessing of assembly language. The lexer does not make
an initial pass to strip out trigraphs and escaped newlines, but handles
them as they are encountered in a single pass of the input file. It
returns preprocessing tokens individually, not a line at a time.
It is mostly transparent to users of the library, since the library's
interface for obtaining the next token, @code{cpp_get_token}, takes care
of lexing new tokens, handling directives, and expanding macros as
necessary. However, the lexer does expose some functionality so that
clients of the library can easily spell a given token, such as
@code{cpp_spell_token} and @code{cpp_token_len}. These functions are
useful when generating diagnostics, and for emitting the preprocessed
output.
@section Lexing a token
Lexing of an individual token is handled by @code{_cpp_lex_direct} and
its subroutines. In its current form the code is quite complicated,
with read ahead characters and such-like, since it strives to not step
back in the character stream in preparation for handling non-ASCII file
encodings. The current plan is to convert any such files to UTF-8
before processing them. This complexity is therefore unnecessary and
will be removed, so I'll not discuss it further here.
The job of @code{_cpp_lex_direct} is simply to lex a token. It is not
responsible for issues like directive handling, returning lookahead
tokens directly, multiple-include optimization, or conditional block
skipping. It necessarily has a minor r@^ole to play in memory
management of lexed lines. I discuss these issues in a separate section
(@pxref{Lexing a line}).
The lexer places the token it lexes into storage pointed to by the
variable @code{cur_token}, and then increments it. This variable is
important for correct diagnostic positioning. Unless a specific line
and column are passed to the diagnostic routines, they will examine the
@code{line} and @code{col} values of the token just before the location
that @code{cur_token} points to, and use that location to report the
diagnostic.
The lexer does not consider whitespace to be a token in its own right.
If whitespace (other than a new line) precedes a token, it sets the
@code{PREV_WHITE} bit in the token's flags. Each token has its
@code{line} and @code{col} variables set to the line and column of the
first character of the token. This line number is the line number in
the translation unit, and can be converted to a source (file, line) pair
using the line map code.
The first token on a logical, i.e.@: unescaped, line has the flag
@code{BOL} set for beginning-of-line. This flag is intended for
internal use, both to distinguish a @samp{#} that begins a directive
from one that doesn't, and to generate a call-back to clients that want
to be notified about the start of every non-directive line with tokens
on it. Clients cannot reliably determine this for themselves: the first
token might be a macro, and the tokens of a macro expansion do not have
the @code{BOL} flag set. The macro expansion may even be empty, and the
next token on the line certainly won't have the @code{BOL} flag set.
New lines are treated specially; exactly how the lexer handles them is
context-dependent. The C standard mandates that directives are
terminated by the first unescaped newline character, even if it appears
in the middle of a macro expansion. Therefore, if the state variable
@code{in_directive} is set, the lexer returns a @code{CPP_EOF} token,
which is normally used to indicate end-of-file, to indicate
end-of-directive. In a directive a @code{CPP_EOF} token never means
end-of-file. Conveniently, if the caller was @code{collect_args}, it
already handles @code{CPP_EOF} as if it were end-of-file, and reports an
error about an unterminated macro argument list.
The C standard also specifies that a new line in the middle of the
arguments to a macro is treated as whitespace. This white space is
important in case the macro argument is stringified. The state variable
@code{parsing_args} is nonzero when the preprocessor is collecting the
arguments to a macro call. It is set to 1 when looking for the opening
parenthesis to a function-like macro, and 2 when collecting the actual
arguments up to the closing parenthesis, since these two cases need to
be distinguished sometimes. One such time is here: the lexer sets the
@code{PREV_WHITE} flag of a token if it meets a new line when
@code{parsing_args} is set to 2. It doesn't set it if it meets a new
line when @code{parsing_args} is 1, since then code like
@smallexample
#define foo() bar
foo
baz
@end smallexample
@noindent would be output with an erroneous space before @samp{baz}:
@smallexample
foo
baz
@end smallexample
This is a good example of the subtlety of getting token spacing correct
in the preprocessor; there are plenty of tests in the testsuite for
corner cases like this.
The lexer is written to treat each of @samp{\r}, @samp{\n}, @samp{\r\n}
and @samp{\n\r} as a single new line indicator. This allows it to
transparently preprocess MS-DOS, Macintosh and Unix files without their
needing to pass through a special filter beforehand.
We also decided to treat a backslash, either @samp{\} or the trigraph
@samp{??/}, separated from one of the above newline indicators by
non-comment whitespace only, as intending to escape the newline. It
tends to be a typing mistake, and cannot reasonably be mistaken for
anything else in any of the C-family grammars. Since handling it this
way is not strictly conforming to the ISO standard, the library issues a
warning wherever it encounters it.
Handling newlines like this is made simpler by doing it in one place
only. The function @code{handle_newline} takes care of all newline
characters, and @code{skip_escaped_newlines} takes care of arbitrarily
long sequences of escaped newlines, deferring to @code{handle_newline}
to handle the newlines themselves.
The most painful aspect of lexing ISO-standard C and C++ is handling
trigraphs and backlash-escaped newlines. Trigraphs are processed before
any interpretation of the meaning of a character is made, and unfortunately
there is a trigraph representation for a backslash, so it is possible for
the trigraph @samp{??/} to introduce an escaped newline.
Escaped newlines are tedious because theoretically they can occur
anywhere---between the @samp{+} and @samp{=} of the @samp{+=} token,
within the characters of an identifier, and even between the @samp{*}
and @samp{/} that terminates a comment. Moreover, you cannot be sure
there is just one---there might be an arbitrarily long sequence of them.
So, for example, the routine that lexes a number, @code{parse_number},
cannot assume that it can scan forwards until the first non-number
character and be done with it, because this could be the @samp{\}
introducing an escaped newline, or the @samp{?} introducing the trigraph
sequence that represents the @samp{\} of an escaped newline. If it
encounters a @samp{?} or @samp{\}, it calls @code{skip_escaped_newlines}
to skip over any potential escaped newlines before checking whether the
number has been finished.
Similarly code in the main body of @code{_cpp_lex_direct} cannot simply
check for a @samp{=} after a @samp{+} character to determine whether it
has a @samp{+=} token; it needs to be prepared for an escaped newline of
some sort. Such cases use the function @code{get_effective_char}, which
returns the first character after any intervening escaped newlines.
The lexer needs to keep track of the correct column position, including
counting tabs as specified by the @option{-ftabstop=} option. This
should be done even within C-style comments; they can appear in the
middle of a line, and we want to report diagnostics in the correct
position for text appearing after the end of the comment.
@anchor{Invalid identifiers}
Some identifiers, such as @code{__VA_ARGS__} and poisoned identifiers,
may be invalid and require a diagnostic. However, if they appear in a
macro expansion we don't want to complain with each use of the macro.
It is therefore best to catch them during the lexing stage, in
@code{parse_identifier}. In both cases, whether a diagnostic is needed
or not is dependent upon the lexer's state. For example, we don't want
to issue a diagnostic for re-poisoning a poisoned identifier, or for
using @code{__VA_ARGS__} in the expansion of a variable-argument macro.
Therefore @code{parse_identifier} makes use of state flags to determine
whether a diagnostic is appropriate. Since we change state on a
per-token basis, and don't lex whole lines at a time, this is not a
problem.
Another place where state flags are used to change behavior is whilst
lexing header names. Normally, a @samp{<} would be lexed as a single
token. After a @code{#include} directive, though, it should be lexed as
a single token as far as the nearest @samp{>} character. Note that we
don't allow the terminators of header names to be escaped; the first
@samp{"} or @samp{>} terminates the header name.
Interpretation of some character sequences depends upon whether we are
lexing C, C++ or Objective-C, and on the revision of the standard in
force. For example, @samp{::} is a single token in C++, but in C it is
two separate @samp{:} tokens and almost certainly a syntax error. Such
cases are handled by @code{_cpp_lex_direct} based upon command-line
flags stored in the @code{cpp_options} structure.
Once a token has been lexed, it leads an independent existence. The
spelling of numbers, identifiers and strings is copied to permanent
storage from the original input buffer, so a token remains valid and
correct even if its source buffer is freed with @code{_cpp_pop_buffer}.
The storage holding the spellings of such tokens remains until the
client program calls cpp_destroy, probably at the end of the translation
unit.
@anchor{Lexing a line}
@section Lexing a line
@cindex token run
When the preprocessor was changed to return pointers to tokens, one
feature I wanted was some sort of guarantee regarding how long a
returned pointer remains valid. This is important to the stand-alone
preprocessor, the future direction of the C family front ends, and even
to cpplib itself internally.
Occasionally the preprocessor wants to be able to peek ahead in the
token stream. For example, after the name of a function-like macro, it
wants to check the next token to see if it is an opening parenthesis.
Another example is that, after reading the first few tokens of a
@code{#pragma} directive and not recognizing it as a registered pragma,
it wants to backtrack and allow the user-defined handler for unknown
pragmas to access the full @code{#pragma} token stream. The stand-alone
preprocessor wants to be able to test the current token with the
previous one to see if a space needs to be inserted to preserve their
separate tokenization upon re-lexing (paste avoidance), so it needs to
be sure the pointer to the previous token is still valid. The
recursive-descent C++ parser wants to be able to perform tentative
parsing arbitrarily far ahead in the token stream, and then to be able
to jump back to a prior position in that stream if necessary.
The rule I chose, which is fairly natural, is to arrange that the
preprocessor lex all tokens on a line consecutively into a token buffer,
which I call a @dfn{token run}, and when meeting an unescaped new line
(newlines within comments do not count either), to start lexing back at
the beginning of the run. Note that we do @emph{not} lex a line of
tokens at once; if we did that @code{parse_identifier} would not have
state flags available to warn about invalid identifiers (@pxref{Invalid
identifiers}).
In other words, accessing tokens that appeared earlier in the current
line is valid, but since each logical line overwrites the tokens of the
previous line, tokens from prior lines are unavailable. In particular,
since a directive only occupies a single logical line, this means that
the directive handlers like the @code{#pragma} handler can jump around
in the directive's tokens if necessary.
Two issues remain: what about tokens that arise from macro expansions,
and what happens when we have a long line that overflows the token run?
Since we promise clients that we preserve the validity of pointers that
we have already returned for tokens that appeared earlier in the line,
we cannot reallocate the run. Instead, on overflow it is expanded by
chaining a new token run on to the end of the existing one.
The tokens forming a macro's replacement list are collected by the
@code{#define} handler, and placed in storage that is only freed by
@code{cpp_destroy}. So if a macro is expanded in the line of tokens,
the pointers to the tokens of its expansion that are returned will always
remain valid. However, macros are a little trickier than that, since
they give rise to three sources of fresh tokens. They are the built-in
macros like @code{__LINE__}, and the @samp{#} and @samp{##} operators
for stringification and token pasting. I handled this by allocating
space for these tokens from the lexer's token run chain. This means
they automatically receive the same lifetime guarantees as lexed tokens,
and we don't need to concern ourselves with freeing them.
Lexing into a line of tokens solves some of the token memory management
issues, but not all. The opening parenthesis after a function-like
macro name might lie on a different line, and the front ends definitely
want the ability to look ahead past the end of the current line. So
cpplib only moves back to the start of the token run at the end of a
line if the variable @code{keep_tokens} is zero. Line-buffering is
quite natural for the preprocessor, and as a result the only time cpplib
needs to increment this variable is whilst looking for the opening
parenthesis to, and reading the arguments of, a function-like macro. In
the near future cpplib will export an interface to increment and
decrement this variable, so that clients can share full control over the
lifetime of token pointers too.
The routine @code{_cpp_lex_token} handles moving to new token runs,
calling @code{_cpp_lex_direct} to lex new tokens, or returning
previously-lexed tokens if we stepped back in the token stream. It also
checks each token for the @code{BOL} flag, which might indicate a
directive that needs to be handled, or require a start-of-line call-back
to be made. @code{_cpp_lex_token} also handles skipping over tokens in
failed conditional blocks, and invalidates the control macro of the
multiple-include optimization if a token was successfully lexed outside
a directive. In other words, its callers do not need to concern
themselves with such issues.
@node Hash Nodes
@unnumbered Hash Nodes
@cindex hash table
@cindex identifiers
@cindex macros
@cindex assertions
@cindex named operators
When cpplib encounters an ``identifier'', it generates a hash code for
it and stores it in the hash table. By ``identifier'' we mean tokens
with type @code{CPP_NAME}; this includes identifiers in the usual C
sense, as well as keywords, directive names, macro names and so on. For
example, all of @code{pragma}, @code{int}, @code{foo} and
@code{__GNUC__} are identifiers and hashed when lexed.
Each node in the hash table contain various information about the
identifier it represents. For example, its length and type. At any one
time, each identifier falls into exactly one of three categories:
@itemize @bullet
@item Macros
These have been declared to be macros, either on the command line or
with @code{#define}. A few, such as @code{__TIME__} are built-ins
entered in the hash table during initialization. The hash node for a
normal macro points to a structure with more information about the
macro, such as whether it is function-like, how many arguments it takes,
and its expansion. Built-in macros are flagged as special, and instead
contain an enum indicating which of the various built-in macros it is.
@item Assertions
Assertions are in a separate namespace to macros. To enforce this, cpp
actually prepends a @code{#} character before hashing and entering it in
the hash table. An assertion's node points to a chain of answers to
that assertion.
@item Void
Everything else falls into this category---an identifier that is not
currently a macro, or a macro that has since been undefined with
@code{#undef}.
When preprocessing C++, this category also includes the named operators,
such as @code{xor}. In expressions these behave like the operators they
represent, but in contexts where the spelling of a token matters they
are spelt differently. This spelling distinction is relevant when they
are operands of the stringizing and pasting macro operators @code{#} and
@code{##}. Named operator hash nodes are flagged, both to catch the
spelling distinction and to prevent them from being defined as macros.
@end itemize
The same identifiers share the same hash node. Since each identifier
token, after lexing, contains a pointer to its hash node, this is used
to provide rapid lookup of various information. For example, when
parsing a @code{#define} statement, CPP flags each argument's identifier
hash node with the index of that argument. This makes duplicated
argument checking an O(1) operation for each argument. Similarly, for
each identifier in the macro's expansion, lookup to see if it is an
argument, and which argument it is, is also an O(1) operation. Further,
each directive name, such as @code{endif}, has an associated directive
enum stored in its hash node, so that directive lookup is also O(1).
@node Macro Expansion
@unnumbered Macro Expansion Algorithm
@cindex macro expansion
Macro expansion is a tricky operation, fraught with nasty corner cases
and situations that render what you thought was a nifty way to
optimize the preprocessor's expansion algorithm wrong in quite subtle
ways.
I strongly recommend you have a good grasp of how the C and C++
standards require macros to be expanded before diving into this
section, let alone the code!. If you don't have a clear mental
picture of how things like nested macro expansion, stringification and
token pasting are supposed to work, damage to your sanity can quickly
result.
@section Internal representation of macros
@cindex macro representation (internal)
The preprocessor stores macro expansions in tokenized form. This
saves repeated lexing passes during expansion, at the cost of a small
increase in memory consumption on average. The tokens are stored
contiguously in memory, so a pointer to the first one and a token
count is all you need to get the replacement list of a macro.
If the macro is a function-like macro the preprocessor also stores its
parameters, in the form of an ordered list of pointers to the hash
table entry of each parameter's identifier. Further, in the macro's
stored expansion each occurrence of a parameter is replaced with a
special token of type @code{CPP_MACRO_ARG}. Each such token holds the
index of the parameter it represents in the parameter list, which
allows rapid replacement of parameters with their arguments during
expansion. Despite this optimization it is still necessary to store
the original parameters to the macro, both for dumping with e.g.,
@option{-dD}, and to warn about non-trivial macro redefinitions when
the parameter names have changed.
@section Macro expansion overview
The preprocessor maintains a @dfn{context stack}, implemented as a
linked list of @code{cpp_context} structures, which together represent
the macro expansion state at any one time. The @code{struct
cpp_reader} member variable @code{context} points to the current top
of this stack. The top normally holds the unexpanded replacement list
of the innermost macro under expansion, except when cpplib is about to
pre-expand an argument, in which case it holds that argument's
unexpanded tokens.
When there are no macros under expansion, cpplib is in @dfn{base
context}. All contexts other than the base context contain a
contiguous list of tokens delimited by a starting and ending token.
When not in base context, cpplib obtains the next token from the list
of the top context. If there are no tokens left in the list, it pops
that context off the stack, and subsequent ones if necessary, until an
unexhausted context is found or it returns to base context. In base
context, cpplib reads tokens directly from the lexer.
If it encounters an identifier that is both a macro and enabled for
expansion, cpplib prepares to push a new context for that macro on the
stack by calling the routine @code{enter_macro_context}. When this
routine returns, the new context will contain the unexpanded tokens of
the replacement list of that macro. In the case of function-like
macros, @code{enter_macro_context} also replaces any parameters in the
replacement list, stored as @code{CPP_MACRO_ARG} tokens, with the
appropriate macro argument. If the standard requires that the
parameter be replaced with its expanded argument, the argument will
have been fully macro expanded first.
@code{enter_macro_context} also handles special macros like
@code{__LINE__}. Although these macros expand to a single token which
cannot contain any further macros, for reasons of token spacing
(@pxref{Token Spacing}) and simplicity of implementation, cpplib
handles these special macros by pushing a context containing just that
one token.
The final thing that @code{enter_macro_context} does before returning
is to mark the macro disabled for expansion (except for special macros
like @code{__TIME__}). The macro is re-enabled when its context is
later popped from the context stack, as described above. This strict
ordering ensures that a macro is disabled whilst its expansion is
being scanned, but that it is @emph{not} disabled whilst any arguments
to it are being expanded.
@section Scanning the replacement list for macros to expand
The C standard states that, after any parameters have been replaced
with their possibly-expanded arguments, the replacement list is
scanned for nested macros. Further, any identifiers in the
replacement list that are not expanded during this scan are never
again eligible for expansion in the future, if the reason they were
not expanded is that the macro in question was disabled.
Clearly this latter condition can only apply to tokens resulting from
argument pre-expansion. Other tokens never have an opportunity to be
re-tested for expansion. It is possible for identifiers that are
function-like macros to not expand initially but to expand during a
later scan. This occurs when the identifier is the last token of an
argument (and therefore originally followed by a comma or a closing
parenthesis in its macro's argument list), and when it replaces its
parameter in the macro's replacement list, the subsequent token
happens to be an opening parenthesis (itself possibly the first token
of an argument).
It is important to note that when cpplib reads the last token of a
given context, that context still remains on the stack. Only when
looking for the @emph{next} token do we pop it off the stack and drop
to a lower context. This makes backing up by one token easy, but more
importantly ensures that the macro corresponding to the current
context is still disabled when we are considering the last token of
its replacement list for expansion (or indeed expanding it). As an
example, which illustrates many of the points above, consider
@smallexample
#define foo(x) bar x
foo(foo) (2)
@end smallexample
@noindent which fully expands to @samp{bar foo (2)}. During pre-expansion
of the argument, @samp{foo} does not expand even though the macro is
enabled, since it has no following parenthesis [pre-expansion of an
argument only uses tokens from that argument; it cannot take tokens
from whatever follows the macro invocation]. This still leaves the
argument token @samp{foo} eligible for future expansion. Then, when
re-scanning after argument replacement, the token @samp{foo} is
rejected for expansion, and marked ineligible for future expansion,
since the macro is now disabled. It is disabled because the
replacement list @samp{bar foo} of the macro is still on the context
stack.
If instead the algorithm looked for an opening parenthesis first and
then tested whether the macro were disabled it would be subtly wrong.
In the example above, the replacement list of @samp{foo} would be
popped in the process of finding the parenthesis, re-enabling
@samp{foo} and expanding it a second time.
@section Looking for a function-like macro's opening parenthesis
Function-like macros only expand when immediately followed by a
parenthesis. To do this cpplib needs to temporarily disable macros
and read the next token. Unfortunately, because of spacing issues
(@pxref{Token Spacing}), there can be fake padding tokens in-between,
and if the next real token is not a parenthesis cpplib needs to be
able to back up that one token as well as retain the information in
any intervening padding tokens.
Backing up more than one token when macros are involved is not
permitted by cpplib, because in general it might involve issues like
restoring popped contexts onto the context stack, which are too hard.
Instead, searching for the parenthesis is handled by a special
function, @code{funlike_invocation_p}, which remembers padding
information as it reads tokens. If the next real token is not an
opening parenthesis, it backs up that one token, and then pushes an
extra context just containing the padding information if necessary.
@section Marking tokens ineligible for future expansion
As discussed above, cpplib needs a way of marking tokens as
unexpandable. Since the tokens cpplib handles are read-only once they
have been lexed, it instead makes a copy of the token and adds the
flag @code{NO_EXPAND} to the copy.
For efficiency and to simplify memory management by avoiding having to
remember to free these tokens, they are allocated as temporary tokens
from the lexer's current token run (@pxref{Lexing a line}) using the
function @code{_cpp_temp_token}. The tokens are then re-used once the
current line of tokens has been read in.
This might sound unsafe. However, tokens runs are not re-used at the
end of a line if it happens to be in the middle of a macro argument
list, and cpplib only wants to back-up more than one lexer token in
situations where no macro expansion is involved, so the optimization
is safe.
@node Token Spacing
@unnumbered Token Spacing
@cindex paste avoidance
@cindex spacing
@cindex token spacing
First, consider an issue that only concerns the stand-alone
preprocessor: there needs to be a guarantee that re-reading its preprocessed
output results in an identical token stream. Without taking special
measures, this might not be the case because of macro substitution.
For example:
@smallexample
#define PLUS +
#define EMPTY
#define f(x) =x=
+PLUS -EMPTY- PLUS+ f(=)
@expansion{} + + - - + + = = =
@emph{not}
@expansion{} ++ -- ++ ===
@end smallexample
One solution would be to simply insert a space between all adjacent
tokens. However, we would like to keep space insertion to a minimum,
both for aesthetic reasons and because it causes problems for people who
still try to abuse the preprocessor for things like Fortran source and
Makefiles.
For now, just notice that when tokens are added (or removed, as shown by
the @code{EMPTY} example) from the original lexed token stream, we need
to check for accidental token pasting. We call this @dfn{paste
avoidance}. Token addition and removal can only occur because of macro
expansion, but accidental pasting can occur in many places: both before
and after each macro replacement, each argument replacement, and
additionally each token created by the @samp{#} and @samp{##} operators.
Look at how the preprocessor gets whitespace output correct
normally. The @code{cpp_token} structure contains a flags byte, and one
of those flags is @code{PREV_WHITE}. This is flagged by the lexer, and
indicates that the token was preceded by whitespace of some form other
than a new line. The stand-alone preprocessor can use this flag to
decide whether to insert a space between tokens in the output.
Now consider the result of the following macro expansion:
@smallexample
#define add(x, y, z) x + y +z;
sum = add (1,2, 3);
@expansion{} sum = 1 + 2 +3;
@end smallexample
The interesting thing here is that the tokens @samp{1} and @samp{2} are
output with a preceding space, and @samp{3} is output without a
preceding space, but when lexed none of these tokens had that property.
Careful consideration reveals that @samp{1} gets its preceding
whitespace from the space preceding @samp{add} in the macro invocation,
@emph{not} replacement list. @samp{2} gets its whitespace from the
space preceding the parameter @samp{y} in the macro replacement list,
and @samp{3} has no preceding space because parameter @samp{z} has none
in the replacement list.
Once lexed, tokens are effectively fixed and cannot be altered, since
pointers to them might be held in many places, in particular by
in-progress macro expansions. So instead of modifying the two tokens
above, the preprocessor inserts a special token, which I call a
@dfn{padding token}, into the token stream to indicate that spacing of
the subsequent token is special. The preprocessor inserts padding
tokens in front of every macro expansion and expanded macro argument.
These point to a @dfn{source token} from which the subsequent real token
should inherit its spacing. In the above example, the source tokens are
@samp{add} in the macro invocation, and @samp{y} and @samp{z} in the
macro replacement list, respectively.
It is quite easy to get multiple padding tokens in a row, for example if
a macro's first replacement token expands straight into another macro.
@smallexample
#define foo bar
#define bar baz
[foo]
@expansion{} [baz]
@end smallexample
Here, two padding tokens are generated with sources the @samp{foo} token
between the brackets, and the @samp{bar} token from foo's replacement
list, respectively. Clearly the first padding token is the one to
use, so the output code should contain a rule that the first
padding token in a sequence is the one that matters.
But what if a macro expansion is left? Adjusting the above
example slightly:
@smallexample
#define foo bar
#define bar EMPTY baz
#define EMPTY
[foo] EMPTY;
@expansion{} [ baz] ;
@end smallexample
As shown, now there should be a space before @samp{baz} and the
semicolon in the output.
The rules we decided above fail for @samp{baz}: we generate three
padding tokens, one per macro invocation, before the token @samp{baz}.
We would then have it take its spacing from the first of these, which
carries source token @samp{foo} with no leading space.
It is vital that cpplib get spacing correct in these examples since any
of these macro expansions could be stringified, where spacing matters.
So, this demonstrates that not just entering macro and argument
expansions, but leaving them requires special handling too. I made
cpplib insert a padding token with a @code{NULL} source token when
leaving macro expansions, as well as after each replaced argument in a
macro's replacement list. It also inserts appropriate padding tokens on
either side of tokens created by the @samp{#} and @samp{##} operators.
I expanded the rule so that, if we see a padding token with a
@code{NULL} source token, @emph{and} that source token has no leading
space, then we behave as if we have seen no padding tokens at all. A
quick check shows this rule will then get the above example correct as
well.
Now a relationship with paste avoidance is apparent: we have to be
careful about paste avoidance in exactly the same locations we have
padding tokens in order to get white space correct. This makes
implementation of paste avoidance easy: wherever the stand-alone
preprocessor is fixing up spacing because of padding tokens, and it
turns out that no space is needed, it has to take the extra step to
check that a space is not needed after all to avoid an accidental paste.
The function @code{cpp_avoid_paste} advises whether a space is required
between two consecutive tokens. To avoid excessive spacing, it tries
hard to only require a space if one is likely to be necessary, but for
reasons of efficiency it is slightly conservative and might recommend a
space where one is not strictly needed.
@node Line Numbering
@unnumbered Line numbering
@cindex line numbers
@section Just which line number anyway?
There are three reasonable requirements a cpplib client might have for
the line number of a token passed to it:
@itemize @bullet
@item
The source line it was lexed on.
@item
The line it is output on. This can be different to the line it was
lexed on if, for example, there are intervening escaped newlines or
C-style comments. For example:
@smallexample
foo /* A long
comment */ bar \
baz
@result{}
foo bar baz
@end smallexample
@item
If the token results from a macro expansion, the line of the macro name,
or possibly the line of the closing parenthesis in the case of
function-like macro expansion.
@end itemize
The @code{cpp_token} structure contains @code{line} and @code{col}
members. The lexer fills these in with the line and column of the first
character of the token. Consequently, but maybe unexpectedly, a token
from the replacement list of a macro expansion carries the location of
the token within the @code{#define} directive, because cpplib expands a
macro by returning pointers to the tokens in its replacement list. The
current implementation of cpplib assigns tokens created from built-in
macros and the @samp{#} and @samp{##} operators the location of the most
recently lexed token. This is a because they are allocated from the
lexer's token runs, and because of the way the diagnostic routines infer
the appropriate location to report.
The diagnostic routines in cpplib display the location of the most
recently @emph{lexed} token, unless they are passed a specific line and
column to report. For diagnostics regarding tokens that arise from
macro expansions, it might also be helpful for the user to see the
original location in the macro definition that the token came from.
Since that is exactly the information each token carries, such an
enhancement could be made relatively easily in future.
The stand-alone preprocessor faces a similar problem when determining
the correct line to output the token on: the position attached to a
token is fairly useless if the token came from a macro expansion. All
tokens on a logical line should be output on its first physical line, so
the token's reported location is also wrong if it is part of a physical
line other than the first.
To solve these issues, cpplib provides a callback that is generated
whenever it lexes a preprocessing token that starts a new logical line
other than a directive. It passes this token (which may be a
@code{CPP_EOF} token indicating the end of the translation unit) to the
callback routine, which can then use the line and column of this token
to produce correct output.
@section Representation of line numbers
As mentioned above, cpplib stores with each token the line number that
it was lexed on. In fact, this number is not the number of the line in
the source file, but instead bears more resemblance to the number of the
line in the translation unit.
The preprocessor maintains a monotonic increasing line count, which is
incremented at every new line character (and also at the end of any
buffer that does not end in a new line). Since a line number of zero is
useful to indicate certain special states and conditions, this variable
starts counting from one.
This variable therefore uniquely enumerates each line in the translation
unit. With some simple infrastructure, it is straight forward to map
from this to the original source file and line number pair, saving space
whenever line number information needs to be saved. The code the
implements this mapping lies in the files @file{line-map.c} and
@file{line-map.h}.
Command-line macros and assertions are implemented by pushing a buffer
containing the right hand side of an equivalent @code{#define} or
@code{#assert} directive. Some built-in macros are handled similarly.
Since these are all processed before the first line of the main input
file, it will typically have an assigned line closer to twenty than to
one.
@node Guard Macros
@unnumbered The Multiple-Include Optimization
@cindex guard macros
@cindex controlling macros
@cindex multiple-include optimization
Header files are often of the form
@smallexample
#ifndef FOO
#define FOO
@dots{}
#endif
@end smallexample
@noindent
to prevent the compiler from processing them more than once. The
preprocessor notices such header files, so that if the header file
appears in a subsequent @code{#include} directive and @code{FOO} is
defined, then it is ignored and it doesn't preprocess or even re-open
the file a second time. This is referred to as the @dfn{multiple
include optimization}.
Under what circumstances is such an optimization valid? If the file
were included a second time, it can only be optimized away if that
inclusion would result in no tokens to return, and no relevant
directives to process. Therefore the current implementation imposes
requirements and makes some allowances as follows:
@enumerate
@item
There must be no tokens outside the controlling @code{#if}-@code{#endif}
pair, but whitespace and comments are permitted.
@item
There must be no directives outside the controlling directive pair, but
the @dfn{null directive} (a line containing nothing other than a single
@samp{#} and possibly whitespace) is permitted.
@item
The opening directive must be of the form
@smallexample
#ifndef FOO
@end smallexample
or
@smallexample
#if !defined FOO [equivalently, #if !defined(FOO)]
@end smallexample
@item
In the second form above, the tokens forming the @code{#if} expression
must have come directly from the source file---no macro expansion must
have been involved. This is because macro definitions can change, and
tracking whether or not a relevant change has been made is not worth the
implementation cost.
@item
There can be no @code{#else} or @code{#elif} directives at the outer
conditional block level, because they would probably contain something
of interest to a subsequent pass.
@end enumerate
First, when pushing a new file on the buffer stack,
@code{_stack_include_file} sets the controlling macro @code{mi_cmacro} to
@code{NULL}, and sets @code{mi_valid} to @code{true}. This indicates
that the preprocessor has not yet encountered anything that would
invalidate the multiple-include optimization. As described in the next
few paragraphs, these two variables having these values effectively
indicates top-of-file.
When about to return a token that is not part of a directive,
@code{_cpp_lex_token} sets @code{mi_valid} to @code{false}. This
enforces the constraint that tokens outside the controlling conditional
block invalidate the optimization.
The @code{do_if}, when appropriate, and @code{do_ifndef} directive
handlers pass the controlling macro to the function
@code{push_conditional}. cpplib maintains a stack of nested conditional
blocks, and after processing every opening conditional this function
pushes an @code{if_stack} structure onto the stack. In this structure
it records the controlling macro for the block, provided there is one
and we're at top-of-file (as described above). If an @code{#elif} or
@code{#else} directive is encountered, the controlling macro for that
block is cleared to @code{NULL}. Otherwise, it survives until the
@code{#endif} closing the block, upon which @code{do_endif} sets
@code{mi_valid} to true and stores the controlling macro in
@code{mi_cmacro}.
@code{_cpp_handle_directive} clears @code{mi_valid} when processing any
directive other than an opening conditional and the null directive.
With this, and requiring top-of-file to record a controlling macro, and
no @code{#else} or @code{#elif} for it to survive and be copied to
@code{mi_cmacro} by @code{do_endif}, we have enforced the absence of
directives outside the main conditional block for the optimization to be
on.
Note that whilst we are inside the conditional block, @code{mi_valid} is
likely to be reset to @code{false}, but this does not matter since the
the closing @code{#endif} restores it to @code{true} if appropriate.
Finally, since @code{_cpp_lex_direct} pops the file off the buffer stack
at @code{EOF} without returning a token, if the @code{#endif} directive
was not followed by any tokens, @code{mi_valid} is @code{true} and
@code{_cpp_pop_file_buffer} remembers the controlling macro associated
with the file. Subsequent calls to @code{stack_include_file} result in
no buffer being pushed if the controlling macro is defined, effecting
the optimization.
A quick word on how we handle the
@smallexample
#if !defined FOO
@end smallexample
@noindent
case. @code{_cpp_parse_expr} and @code{parse_defined} take steps to see
whether the three stages @samp{!}, @samp{defined-expression} and
@samp{end-of-directive} occur in order in a @code{#if} expression. If
so, they return the guard macro to @code{do_if} in the variable
@code{mi_ind_cmacro}, and otherwise set it to @code{NULL}.
@code{enter_macro_context} sets @code{mi_valid} to false, so if a macro
was expanded whilst parsing any part of the expression, then the
top-of-file test in @code{push_conditional} fails and the optimization
is turned off.
@node Files
@unnumbered File Handling
@cindex files
Fairly obviously, the file handling code of cpplib resides in the file
@file{cppfiles.c}. It takes care of the details of file searching,
opening, reading and caching, for both the main source file and all the
headers it recursively includes.
The basic strategy is to minimize the number of system calls. On many
systems, the basic @code{open ()} and @code{fstat ()} system calls can
be quite expensive. For every @code{#include}-d file, we need to try
all the directories in the search path until we find a match. Some
projects, such as glibc, pass twenty or thirty include paths on the
command line, so this can rapidly become time consuming.
For a header file we have not encountered before we have little choice
but to do this. However, it is often the case that the same headers are
repeatedly included, and in these cases we try to avoid repeating the
filesystem queries whilst searching for the correct file.
For each file we try to open, we store the constructed path in a splay
tree. This path first undergoes simplification by the function
@code{_cpp_simplify_pathname}. For example,
@file{/usr/include/bits/../foo.h} is simplified to
@file{/usr/include/foo.h} before we enter it in the splay tree and try
to @code{open ()} the file. CPP will then find subsequent uses of
@file{foo.h}, even as @file{/usr/include/foo.h}, in the splay tree and
save system calls.
Further, it is likely the file contents have also been cached, saving a
@code{read ()} system call. We don't bother caching the contents of
header files that are re-inclusion protected, and whose re-inclusion
macro is defined when we leave the header file for the first time. If
the host supports it, we try to map suitably large files into memory,
rather than reading them in directly.
The include paths are internally stored on a null-terminated
singly-linked list, starting with the @code{"header.h"} directory search
chain, which then links into the @code{<header.h>} directory chain.
Files included with the @code{<foo.h>} syntax start the lookup directly
in the second half of this chain. However, files included with the
@code{"foo.h"} syntax start at the beginning of the chain, but with one
extra directory prepended. This is the directory of the current file;
the one containing the @code{#include} directive. Prepending this
directory on a per-file basis is handled by the function
@code{search_from}.
Note that a header included with a directory component, such as
@code{#include "mydir/foo.h"} and opened as
@file{/usr/local/include/mydir/foo.h}, will have the complete path minus
the basename @samp{foo.h} as the current directory.
Enough information is stored in the splay tree that CPP can immediately
tell whether it can skip the header file because of the multiple include
optimization, whether the file didn't exist or couldn't be opened for
some reason, or whether the header was flagged not to be re-used, as it
is with the obsolete @code{#import} directive.
For the benefit of MS-DOS filesystems with an 8.3 filename limitation,
CPP offers the ability to treat various include file names as aliases
for the real header files with shorter names. The map from one to the
other is found in a special file called @samp{header.gcc}, stored in the
command line (or system) include directories to which the mapping
applies. This may be higher up the directory tree than the full path to
the file minus the base name.
@node Concept Index
@unnumbered Concept Index
@printindex cp
@bye
|