1 /* This is the Assembler Pre-Processor
2 Copyright (C) 1987, 90, 91, 92, 93, 94, 95, 96, 97, 98, 1999
3 Free Software Foundation, Inc.
5 This file is part of GAS, the GNU Assembler.
7 GAS is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 2, or (at your option)
12 GAS is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with GAS; see the file COPYING. If not, write to the Free
19 Software Foundation, 59 Temple Place - Suite 330, Boston, MA
22 /* Modified by Allen Wirfs-Brock, Instantiations Inc 2/90 */
23 /* App, the assembler pre-processor. This pre-processor strips out excess
24 spaces, turns single-quoted characters into a decimal constant, and turns
25 # <number> <filename> <garbage> into a .line <number>\n.file <filename>
26 pair. This needs better error-handling. */
29 #include "as.h" /* For BAD_CASE() only */
33 #define const /* empty */
37 /* Whether we are scrubbing in m68k MRI mode. This is different from
38 flag_m68k_mri, because the two flags will be affected by the .mri
39 pseudo-op at different times. */
40 static int scrub_m68k_mri
;
42 /* The pseudo-op which switches in and out of MRI mode. See the
43 comment in do_scrub_chars. */
44 static const char mri_pseudo
[] = ".mri 0";
46 #if defined TC_ARM && defined OBJ_ELF
47 /* The pseudo-op for which we need to special-case `@' characters.
48 See the comment in do_scrub_chars. */
49 static const char symver_pseudo
[] = ".symver";
50 static const char * symver_state
;
54 static const char symbol_chars
[] =
55 "$._ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789";
57 #define LEX_IS_SYMBOL_COMPONENT 1
58 #define LEX_IS_WHITESPACE 2
59 #define LEX_IS_LINE_SEPARATOR 3
60 #define LEX_IS_COMMENT_START 4
61 #define LEX_IS_LINE_COMMENT_START 5
62 #define LEX_IS_TWOCHAR_COMMENT_1ST 6
63 #define LEX_IS_STRINGQUOTE 8
64 #define LEX_IS_COLON 9
65 #define LEX_IS_NEWLINE 10
66 #define LEX_IS_ONECHAR_QUOTE 11
68 #define LEX_IS_DOUBLEDASH_1ST 12
71 #define DOUBLEBAR_PARALLEL
73 #ifdef DOUBLEBAR_PARALLEL
74 #define LEX_IS_DOUBLEBAR_1ST 13
76 #define IS_SYMBOL_COMPONENT(c) (lex[c] == LEX_IS_SYMBOL_COMPONENT)
77 #define IS_WHITESPACE(c) (lex[c] == LEX_IS_WHITESPACE)
78 #define IS_LINE_SEPARATOR(c) (lex[c] == LEX_IS_LINE_SEPARATOR)
79 #define IS_COMMENT(c) (lex[c] == LEX_IS_COMMENT_START)
80 #define IS_LINE_COMMENT(c) (lex[c] == LEX_IS_LINE_COMMENT_START)
81 #define IS_NEWLINE(c) (lex[c] == LEX_IS_NEWLINE)
83 static int process_escape
PARAMS ((int));
85 /* FIXME-soon: The entire lexer/parser thingy should be
86 built statically at compile time rather than dynamically
87 each and every time the assembler is run. xoxorich. */
90 do_scrub_begin (m68k_mri
)
96 scrub_m68k_mri
= m68k_mri
;
98 lex
[' '] = LEX_IS_WHITESPACE
;
99 lex
['\t'] = LEX_IS_WHITESPACE
;
100 lex
['\r'] = LEX_IS_WHITESPACE
;
101 lex
['\n'] = LEX_IS_NEWLINE
;
102 lex
[';'] = LEX_IS_LINE_SEPARATOR
;
103 lex
[':'] = LEX_IS_COLON
;
107 lex
['"'] = LEX_IS_STRINGQUOTE
;
110 lex
['\''] = LEX_IS_ONECHAR_QUOTE
;
113 #ifdef SINGLE_QUOTE_STRINGS
114 lex
['\''] = LEX_IS_STRINGQUOTE
;
118 /* Note: if any other character can be LEX_IS_STRINGQUOTE, the loop
119 in state 5 of do_scrub_chars must be changed. */
121 /* Note that these override the previous defaults, e.g. if ';' is a
122 comment char, then it isn't a line separator. */
123 for (p
= symbol_chars
; *p
; ++p
)
125 lex
[(unsigned char) *p
] = LEX_IS_SYMBOL_COMPONENT
;
126 } /* declare symbol characters */
128 for (c
= 128; c
< 256; ++c
)
129 lex
[c
] = LEX_IS_SYMBOL_COMPONENT
;
131 #ifdef tc_symbol_chars
132 /* This macro permits the processor to specify all characters which
133 may appears in an operand. This will prevent the scrubber from
134 discarding meaningful whitespace in certain cases. The i386
135 backend uses this to support prefixes, which can confuse the
136 scrubber as to whether it is parsing operands or opcodes. */
137 for (p
= tc_symbol_chars
; *p
; ++p
)
138 lex
[(unsigned char) *p
] = LEX_IS_SYMBOL_COMPONENT
;
141 /* The m68k backend wants to be able to change comment_chars. */
142 #ifndef tc_comment_chars
143 #define tc_comment_chars comment_chars
145 for (p
= tc_comment_chars
; *p
; p
++)
147 lex
[(unsigned char) *p
] = LEX_IS_COMMENT_START
;
148 } /* declare comment chars */
150 for (p
= line_comment_chars
; *p
; p
++)
152 lex
[(unsigned char) *p
] = LEX_IS_LINE_COMMENT_START
;
153 } /* declare line comment chars */
155 for (p
= line_separator_chars
; *p
; p
++)
157 lex
[(unsigned char) *p
] = LEX_IS_LINE_SEPARATOR
;
158 } /* declare line separators */
160 /* Only allow slash-star comments if slash is not in use.
161 FIXME: This isn't right. We should always permit them. */
164 lex
['/'] = LEX_IS_TWOCHAR_COMMENT_1ST
;
169 lex
['\''] = LEX_IS_STRINGQUOTE
;
170 lex
[';'] = LEX_IS_COMMENT_START
;
171 lex
['*'] = LEX_IS_LINE_COMMENT_START
;
172 /* The MRI documentation says '!' is LEX_IS_COMMENT_START, but
173 then it can't be used in an expression. */
174 lex
['!'] = LEX_IS_LINE_COMMENT_START
;
178 lex
['-'] = LEX_IS_DOUBLEDASH_1ST
;
180 #ifdef DOUBLEBAR_PARALLEL
181 lex
['|'] = LEX_IS_DOUBLEBAR_1ST
;
184 /* must do this is we want VLIW instruction with "->" or "<-" */
185 lex
['-'] = LEX_IS_SYMBOL_COMPONENT
;
187 } /* do_scrub_begin() */
189 /* Saved state of the scrubber */
191 static int old_state
;
192 static char *out_string
;
193 static char out_buf
[20];
194 static int add_newlines
;
195 static char *saved_input
;
196 static int saved_input_len
;
197 static char input_buffer
[32 * 1024];
198 static const char *mri_state
;
199 static char mri_last_ch
;
201 /* Data structure for saving the state of app across #include's. Note that
202 app is called asynchronously to the parsing of the .include's, so our
203 state at the time .include is interpreted is completely unrelated.
204 That's why we have to save it all. */
211 char out_buf
[sizeof (out_buf
)];
216 const char * mri_state
;
218 #if defined TC_ARM && defined OBJ_ELF
219 const char * symver_state
;
226 register struct app_save
*saved
;
228 saved
= (struct app_save
*) xmalloc (sizeof (*saved
));
229 saved
->state
= state
;
230 saved
->old_state
= old_state
;
231 saved
->out_string
= out_string
;
232 memcpy (saved
->out_buf
, out_buf
, sizeof (out_buf
));
233 saved
->add_newlines
= add_newlines
;
234 if (saved_input
== NULL
)
235 saved
->saved_input
= NULL
;
238 saved
->saved_input
= xmalloc (saved_input_len
);
239 memcpy (saved
->saved_input
, saved_input
, saved_input_len
);
240 saved
->saved_input_len
= saved_input_len
;
242 saved
->scrub_m68k_mri
= scrub_m68k_mri
;
243 saved
->mri_state
= mri_state
;
244 saved
->mri_last_ch
= mri_last_ch
;
245 #if defined TC_ARM && defined OBJ_ELF
246 saved
->symver_state
= symver_state
;
249 /* do_scrub_begin() is not useful, just wastes time. */
254 return (char *) saved
;
261 register struct app_save
*saved
= (struct app_save
*) arg
;
263 /* There is no do_scrub_end (). */
264 state
= saved
->state
;
265 old_state
= saved
->old_state
;
266 out_string
= saved
->out_string
;
267 memcpy (out_buf
, saved
->out_buf
, sizeof (out_buf
));
268 add_newlines
= saved
->add_newlines
;
269 if (saved
->saved_input
== NULL
)
273 assert (saved
->saved_input_len
<= (int) (sizeof input_buffer
));
274 memcpy (input_buffer
, saved
->saved_input
, saved
->saved_input_len
);
275 saved_input
= input_buffer
;
276 saved_input_len
= saved
->saved_input_len
;
277 free (saved
->saved_input
);
279 scrub_m68k_mri
= saved
->scrub_m68k_mri
;
280 mri_state
= saved
->mri_state
;
281 mri_last_ch
= saved
->mri_last_ch
;
282 #if defined TC_ARM && defined OBJ_ELF
283 symver_state
= saved
->symver_state
;
289 /* @@ This assumes that \n &c are the same on host and target. This is not
316 /* This function is called to process input characters. The GET
317 parameter is used to retrieve more input characters. GET should
318 set its parameter to point to a buffer, and return the length of
319 the buffer; it should return 0 at end of file. The scrubbed output
320 characters are put into the buffer starting at TOSTART; the TOSTART
321 buffer is TOLEN bytes in length. The function returns the number
322 of scrubbed characters put into TOSTART. This will be TOLEN unless
323 end of file was seen. This function is arranged as a state
324 machine, and saves its state so that it may return at any point.
325 This is the way the old code used to work. */
328 do_scrub_chars (get
, tostart
, tolen
)
329 int (*get
) PARAMS ((char *, int));
334 char *toend
= tostart
+ tolen
;
338 register int ch
, ch2
= 0;
340 /*State 0: beginning of normal line
341 1: After first whitespace on line (flush more white)
342 2: After first non-white (opcode) on line (keep 1white)
343 3: after second white on line (into operands) (flush white)
344 4: after putting out a .line, put out digits
345 5: parsing a string, then go to old-state
346 6: putting out \ escape in a "d string.
347 7: After putting out a .appfile, put out string.
348 8: After putting out a .appfile string, flush until newline.
349 9: After seeing symbol char in state 3 (keep 1white after symchar)
350 10: After seeing whitespace in state 9 (keep white before symchar)
351 11: After seeing a symbol character in state 0 (eg a label definition)
352 -1: output string in out_string and go to the state in old_state
353 -2: flush text until a '*' '/' is seen, then go to state old_state
355 12: After seeing a dash, looking for a second dash as a start of comment.
357 #ifdef DOUBLEBAR_PARALLEL
358 13: After seeing a vertical bar, looking for a second vertical bar as a parallel expression seperator.
362 /* I added states 9 and 10 because the MIPS ECOFF assembler uses
363 constructs like ``.loc 1 20''. This was turning into ``.loc
364 120''. States 9 and 10 ensure that a space is never dropped in
365 between characters which could appear in a identifier. Ian
366 Taylor, ian@cygnus.com.
368 I added state 11 so that something like "Lfoo add %r25,%r26,%r27" works
369 correctly on the PA (and any other target where colons are optional).
370 Jeff Law, law@cs.utah.edu.
372 I added state 13 so that something like "cmp r1, r2 || trap #1" does not
373 get squashed into "cmp r1,r2||trap#1", with the all important space
374 between the 'trap' and the '#1' being eliminated. nickc@cygnus.com */
376 /* This macro gets the next input character. */
380 ? * (unsigned char *) (from++) \
381 : (saved_input = NULL, \
382 fromlen = (*get) (input_buffer, sizeof input_buffer), \
383 from = input_buffer, \
384 fromend = from + fromlen, \
387 : * (unsigned char *) (from++))))
389 /* This macro pushes a character back on the input stream. */
391 #define UNGET(uch) (*--from = (uch))
393 /* This macro puts a character into the output buffer. If this
394 character fills the output buffer, this macro jumps to the label
395 TOFULL. We use this rather ugly approach because we need to
396 handle two different termination conditions: EOF on the input
397 stream, and a full output buffer. It would be simpler if we
398 always read in the entire input stream before processing it, but
399 I don't want to make such a significant change to the assembler's
411 if (saved_input
!= NULL
)
414 fromend
= from
+ saved_input_len
;
418 fromlen
= (*get
) (input_buffer
, sizeof input_buffer
);
422 fromend
= from
+ fromlen
;
427 /* The cases in this switch end with continue, in order to
428 branch back to the top of this while loop and generate the
429 next output character in the appropriate state. */
434 if (*out_string
== '\0')
451 as_warn (_("end of file in comment"));
460 while ((ch
= GET ()) == '*')
465 as_warn (_("end of file in comment"));
483 else if (ch
>= '0' && ch
<= '9')
487 while (ch
!= EOF
&& IS_WHITESPACE (ch
))
493 out_string
= "\n\tappfile ";
495 out_string
= "\n\t.appfile ";
502 while (ch
!= EOF
&& ch
!= '\n')
511 /* We are going to copy everything up to a quote character,
512 with special handling for a backslash. We try to
513 optimize the copying in the simple case without using the
514 GET and PUT macros. */
519 for (s
= from
; s
< fromend
; s
++)
522 /* This condition must be changed if the type of any
523 other character can be LEX_IS_STRINGQUOTE. */
531 if (len
> toend
- to
)
535 memcpy (to
, from
, len
);
544 as_warn (_("end of file in string: inserted '\"'"));
549 else if (lex
[ch
] == LEX_IS_STRINGQUOTE
)
554 #ifndef NO_STRING_ESCAPES
561 else if (scrub_m68k_mri
&& ch
== '\n')
563 /* Just quietly terminate the string. This permits lines like
564 bne label loop if we haven't reach end yet
581 /* Handle strings broken across lines, by turning '\n' into
608 #if defined(IGNORE_NONSTANDARD_ESCAPES) | defined(ONLY_STANDARD_ESCAPES)
610 as_warn (_("Unknown escape '\\%c' in string: Ignored"), ch
);
612 #else /* ONLY_STANDARD_ESCAPES */
614 /* Accept \x as x for any x */
616 #endif /* ONLY_STANDARD_ESCAPES */
619 as_warn (_("End of file in string: '\"' inserted"));
638 while (ch
!= '\n' && ch
!= EOF
);
646 /* OK, we are somewhere in states 0 through 4 or 9 through 11 */
653 #if defined TC_ARM && defined OBJ_ELF
654 /* We need to watch out for .symver directives. See the comment later
656 if (symver_state
== NULL
)
658 if ((state
== 0 || state
== 1) && ch
== symver_pseudo
[0])
659 symver_state
= symver_pseudo
+ 1;
663 /* We advance to the next state if we find the right
665 if (ch
!= '\0' && (*symver_state
== ch
))
667 else if (*symver_state
!= '\0')
668 /* We did not get the expected character, or we didn't
669 get a valid terminating character after seeing the
670 entire pseudo-op, so we must go back to the beginning. */
674 /* We've read the entire pseudo-op. If this is the end
675 of the line, go back to the beginning. */
680 #endif /* TC_ARM && OBJ_ELF */
683 /* We want to have pseudo-ops which control whether we are in
684 MRI mode or not. Unfortunately, since m68k MRI mode affects
685 the scrubber, that means that we need a special purpose
687 if (mri_state
== NULL
)
689 if ((state
== 0 || state
== 1)
690 && ch
== mri_pseudo
[0])
691 mri_state
= mri_pseudo
+ 1;
695 /* We advance to the next state if we find the right
696 character, or if we need a space character and we get any
697 whitespace character, or if we need a '0' and we get a
698 '1' (this is so that we only need one state to handle
699 ``.mri 0'' and ``.mri 1''). */
702 || (*mri_state
== ' '
703 && lex
[ch
] == LEX_IS_WHITESPACE
)
704 || (*mri_state
== '0'
710 else if (*mri_state
!= '\0'
711 || (lex
[ch
] != LEX_IS_WHITESPACE
712 && lex
[ch
] != LEX_IS_NEWLINE
))
714 /* We did not get the expected character, or we didn't
715 get a valid terminating character after seeing the
716 entire pseudo-op, so we must go back to the
722 /* We've read the entire pseudo-op. mips_last_ch is
723 either '0' or '1' indicating whether to enter or
725 do_scrub_begin (mri_last_ch
== '1');
728 /* We continue handling the character as usual. The
729 main gas reader must also handle the .mri pseudo-op
730 to control expression parsing and the like. */
739 as_warn (_("end of file not at end of a line; newline inserted"));
748 case LEX_IS_WHITESPACE
:
753 while (ch
!= EOF
&& IS_WHITESPACE (ch
));
759 /* Preserve a single whitespace character at the
760 beginning of a line. */
767 #ifdef KEEP_WHITE_AROUND_COLON
768 if (lex
[ch
] == LEX_IS_COLON
)
770 /* only keep this white if there's no white *after* the colon */
773 if (!IS_WHITESPACE (ch2
))
784 || IS_LINE_SEPARATOR (ch
))
788 /* In MRI mode, we keep these spaces. */
796 /* If we're in state 2 or 11, we've seen a non-white
797 character followed by whitespace. If the next character
798 is ':', this is whitespace after a label name which we
799 normally must ignore. In MRI mode, though, spaces are
800 not permitted between the label and the colon. */
801 if ((state
== 2 || state
== 11)
802 && lex
[ch
] == LEX_IS_COLON
814 goto recycle
; /* Punted leading sp */
816 /* We can arrive here if we leave a leading whitespace
817 character at the beginning of a line. */
823 /* Optimize common case by skipping UNGET/GET. */
824 PUT (' '); /* Sp after opco */
833 /* In MRI mode, we keep these spaces. */
838 goto recycle
; /* Sp in operands */
843 /* In MRI mode, we keep these spaces. */
849 state
= 10; /* Sp after symbol char */
853 #ifdef LABELS_WITHOUT_COLONS
860 /* We know that ch is not ':', since we tested that
861 case above. Therefore this is not a label, so it
862 must be the opcode, and we've just seen the
863 whitespace after it. */
867 PUT (' '); /* Sp after label definition. */
874 case LEX_IS_TWOCHAR_COMMENT_1ST
:
883 if (ch2
!= EOF
&& IS_NEWLINE (ch2
))
886 while (ch2
!= EOF
&& ch2
!= '*');
891 if (ch2
== EOF
|| ch2
== '/')
894 /* This UNGET will ensure that we count newlines
900 as_warn (_("end of file in multiline comment"));
909 if (state
== 9 || state
== 10)
915 case LEX_IS_STRINGQUOTE
:
918 /* Preserve the whitespace in foo "bar" */
923 /* PUT didn't jump out. We could just break, but we
924 know what will happen, so optimize a bit. */
937 case LEX_IS_ONECHAR_QUOTE
:
940 /* Preserve the whitespace in foo 'b' */
949 as_warn (_("end of file after a one-character quote; \\0 inserted"));
957 as_warn (_("end of file in escape character"));
961 ch
= process_escape (ch
);
963 sprintf (out_buf
, "%d", (int) (unsigned char) ch
);
965 /* None of these 'x constants for us. We want 'x'. */
966 if ((ch
= GET ()) != '\'')
968 #ifdef REQUIRE_CHAR_CLOSE_QUOTE
969 as_warn (_("Missing close quote: (assumed)"));
975 if (strlen (out_buf
) == 1)
985 out_string
= out_buf
;
991 #ifdef KEEP_WHITE_AROUND_COLON
994 if (state
== 9 || state
== 10)
1002 case LEX_IS_NEWLINE
:
1003 /* Roll out a bunch of newlines from inside comments, etc. */
1009 /* fall thru into... */
1011 case LEX_IS_LINE_SEPARATOR
:
1017 case LEX_IS_DOUBLEDASH_1ST
:
1024 /* read and skip to end of line */
1029 while (ch
!= EOF
&& ch
!= '\n');
1032 as_warn (_("end of file in comment; newline inserted"));
1038 #ifdef DOUBLEBAR_PARALLEL
1039 case LEX_IS_DOUBLEBAR_1ST
:
1046 /* Reset back to state 1 and pretend that we are parsing a line from
1047 just after the first white space. */
1053 case LEX_IS_LINE_COMMENT_START
:
1054 /* FIXME-someday: The two character comment stuff was badly
1055 thought out. On i386, we want '/' as line comment start
1056 AND we want C style comments. hence this hack. The
1057 whole lexical process should be reworked. xoxorich. */
1073 if (state
== 0 || state
== 1) /* Only comment at start of line. */
1083 while (ch
!= EOF
&& IS_WHITESPACE (ch
));
1086 as_warn (_("end of file in comment; newline inserted"));
1090 if (ch
< '0' || ch
> '9' || state
!= 0 || startch
!= '#')
1092 /* Not a cpp line. */
1093 while (ch
!= EOF
&& !IS_NEWLINE (ch
))
1096 as_warn (_("EOF in Comment: Newline inserted"));
1101 /* Loks like `# 123 "filename"' from cpp. */
1106 out_string
= "\tappline ";
1108 out_string
= "\t.appline ";
1109 PUT (*out_string
++);
1114 /* All insns end in a char for which LEX_IS_SYMBOL_COMPONENT is true.
1115 Trap is the only short insn that has a first operand that is
1116 neither register nor label.
1117 We must prevent exef0f ||trap #1 to degenerate to exef0f ||trap#1 .
1118 We can't make '#' LEX_IS_SYMBOL_COMPONENT because it is already
1119 LEX_IS_LINE_COMMENT_START. However, it is the only character in
1120 line_comment_chars for d10v, hence we can recognize it as such. */
1121 /* An alternative approach would be to reset the state to 1 when
1122 we see '||', '<'- or '->', but that seems to be overkill. */
1123 if (state
== 10) PUT (' ');
1125 /* We have a line comment character which is not at the
1126 start of a line. If this is also a normal comment
1127 character, fall through. Otherwise treat it as a default
1129 if (strchr (tc_comment_chars
, ch
) == NULL
1130 && (! scrub_m68k_mri
1131 || (ch
!= '!' && ch
!= '*')))
1134 && (ch
== '!' || ch
== '*' || ch
== '#')
1139 case LEX_IS_COMMENT_START
:
1140 #if defined TC_ARM && defined OBJ_ELF
1141 /* On the ARM, `@' is the comment character.
1142 Unfortunately this is also a special character in ELF .symver
1143 directives (and .type, though we deal with those another way). So
1144 we check if this line is such a directive, and treat the character
1145 as default if so. This is a hack. */
1146 if ((symver_state
!= NULL
) && (*symver_state
== 0))
1153 while (ch
!= EOF
&& !IS_NEWLINE (ch
));
1155 as_warn (_("end of file in comment; newline inserted"));
1160 case LEX_IS_SYMBOL_COMPONENT
:
1163 /* This is a symbol character following another symbol
1164 character, with whitespace in between. We skipped
1165 the whitespace earlier, so output it now. */
1175 /* This is a common case. Quickly copy CH and all the
1176 following symbol component or normal characters. */
1178 && mri_state
== NULL
1179 #if defined TC_ARM && defined OBJ_ELF
1180 && symver_state
== NULL
1187 for (s
= from
; s
< fromend
; s
++)
1191 ch2
= * (unsigned char *) s
;
1194 && type
!= LEX_IS_SYMBOL_COMPONENT
)
1199 /* Handle the last character normally, for
1204 if (len
> (toend
- to
) - 1)
1205 len
= (toend
- to
) - 1;
1211 memcpy (to
, from
, len
);
1219 case 8: *to
++ = *from
++;
1220 case 7: *to
++ = *from
++;
1221 case 6: *to
++ = *from
++;
1222 case 5: *to
++ = *from
++;
1223 case 4: *to
++ = *from
++;
1224 case 3: *to
++ = *from
++;
1225 case 2: *to
++ = *from
++;
1226 case 1: *to
++ = *from
++;
1236 /* Some relatively `normal' character. */
1239 state
= 11; /* Now seeing label definition */
1241 else if (state
== 1)
1243 state
= 2; /* Ditto */
1245 else if (state
== 9)
1247 if (lex
[ch
] != LEX_IS_SYMBOL_COMPONENT
)
1250 else if (state
== 10)
1254 /* Special handling for backslash: a backslash may
1255 be the beginning of a formal parameter (of a
1256 macro) following another symbol character, with
1257 whitespace in between. If that is the case, we
1258 output a space before the parameter. Strictly
1259 speaking, correct handling depends upon what the
1260 macro parameter expands into; if the parameter
1261 expands into something which does not start with
1262 an operand character, then we don't want to keep
1263 the space. We don't have enough information to
1264 make the right choice, so here we are making the
1265 choice which is more likely to be correct. */
1279 /* We have reached the end of the input. */
1280 return to
- tostart
;
1283 /* The output buffer is full. Save any input we have not yet
1288 saved_input_len
= fromend
- from
;
1293 return to
- tostart
;