/* This is the Assembler Pre-Processor
- Copyright (C) 1987, 90, 91, 92, 93, 94, 95, 96, 97, 98, 1999
+ Copyright 1987, 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
+ 1999, 2000
Free Software Foundation, Inc.
This file is part of GAS, the GNU Assembler.
#endif
#endif
+#ifdef TC_M68K
/* Whether we are scrubbing in m68k MRI mode. This is different from
flag_m68k_mri, because the two flags will be affected by the .mri
pseudo-op at different times. */
/* The pseudo-op which switches in and out of MRI mode. See the
comment in do_scrub_chars. */
static const char mri_pseudo[] = ".mri 0";
+#else
+#define scrub_m68k_mri 0
+#endif
#if defined TC_ARM && defined OBJ_ELF
-/* The pseudo-op for which we need to special-case `@' characters.
+/* The pseudo-op for which we need to special-case `@' characters.
See the comment in do_scrub_chars. */
static const char symver_pseudo[] = ".symver";
static const char * symver_state;
#define LEX_IS_DOUBLEDASH_1ST 12
#endif
#ifdef TC_M32R
+#define DOUBLEBAR_PARALLEL
+#endif
+#ifdef DOUBLEBAR_PARALLEL
#define LEX_IS_DOUBLEBAR_1ST 13
#endif
+#define LEX_IS_PARALLEL_SEPARATOR 14
#define IS_SYMBOL_COMPONENT(c) (lex[c] == LEX_IS_SYMBOL_COMPONENT)
#define IS_WHITESPACE(c) (lex[c] == LEX_IS_WHITESPACE)
#define IS_LINE_SEPARATOR(c) (lex[c] == LEX_IS_LINE_SEPARATOR)
+#define IS_PARALLEL_SEPARATOR(c) (lex[c] == LEX_IS_PARALLEL_SEPARATOR)
#define IS_COMMENT(c) (lex[c] == LEX_IS_COMMENT_START)
#define IS_LINE_COMMENT(c) (lex[c] == LEX_IS_LINE_COMMENT_START)
#define IS_NEWLINE(c) (lex[c] == LEX_IS_NEWLINE)
/* FIXME-soon: The entire lexer/parser thingy should be
built statically at compile time rather than dynamically
- each and every time the assembler is run. xoxorich. */
+ each and every time the assembler is run. xoxorich. */
-void
+void
do_scrub_begin (m68k_mri)
- int m68k_mri;
+ int m68k_mri ATTRIBUTE_UNUSED;
{
const char *p;
int c;
- scrub_m68k_mri = m68k_mri;
-
lex[' '] = LEX_IS_WHITESPACE;
lex['\t'] = LEX_IS_WHITESPACE;
lex['\r'] = LEX_IS_WHITESPACE;
lex['\n'] = LEX_IS_NEWLINE;
- lex[';'] = LEX_IS_LINE_SEPARATOR;
lex[':'] = LEX_IS_COLON;
+#ifdef TC_M68K
+ scrub_m68k_mri = m68k_mri;
+
if (! m68k_mri)
+#endif
{
lex['"'] = LEX_IS_STRINGQUOTE;
-#ifndef TC_HPPA
+#if ! defined (TC_HPPA) && ! defined (TC_I370)
+ /* I370 uses single-quotes to delimit integer, float constants */
lex['\''] = LEX_IS_ONECHAR_QUOTE;
#endif
lex[(unsigned char) *p] = LEX_IS_LINE_SEPARATOR;
} /* declare line separators */
+#ifdef tc_parallel_separator_chars
+ /* This macro permits the processor to specify all characters which
+ separate parallel insns on the same line. */
+ for (p = tc_parallel_separator_chars; *p; p++)
+ {
+ lex[(unsigned char) *p] = LEX_IS_PARALLEL_SEPARATOR;
+ } /* declare parallel separators */
+#endif
+
/* Only allow slash-star comments if slash is not in use.
FIXME: This isn't right. We should always permit them. */
if (lex['/'] == 0)
lex['/'] = LEX_IS_TWOCHAR_COMMENT_1ST;
}
+#ifdef TC_M68K
if (m68k_mri)
{
lex['\''] = LEX_IS_STRINGQUOTE;
then it can't be used in an expression. */
lex['!'] = LEX_IS_LINE_COMMENT_START;
}
+#endif
#ifdef TC_V850
lex['-'] = LEX_IS_DOUBLEDASH_1ST;
#endif
-#ifdef TC_M32R
+#ifdef DOUBLEBAR_PARALLEL
lex['|'] = LEX_IS_DOUBLEBAR_1ST;
#endif
#ifdef TC_D30V
static int add_newlines;
static char *saved_input;
static int saved_input_len;
+static char input_buffer[32 * 1024];
static const char *mri_state;
static char mri_last_ch;
state at the time .include is interpreted is completely unrelated.
That's why we have to save it all. */
-struct app_save
- {
- int state;
- int old_state;
- char * out_string;
- char out_buf[sizeof (out_buf)];
- int add_newlines;
- char * saved_input;
- int saved_input_len;
- int scrub_m68k_mri;
- const char * mri_state;
- char mri_last_ch;
+struct app_save {
+ int state;
+ int old_state;
+ char * out_string;
+ char out_buf[sizeof (out_buf)];
+ int add_newlines;
+ char * saved_input;
+ int saved_input_len;
+#ifdef TC_M68K
+ int scrub_m68k_mri;
+#endif
+ const char * mri_state;
+ char mri_last_ch;
#if defined TC_ARM && defined OBJ_ELF
- const char * symver_state;
+ const char * symver_state;
#endif
- };
+};
char *
app_push ()
saved->out_string = out_string;
memcpy (saved->out_buf, out_buf, sizeof (out_buf));
saved->add_newlines = add_newlines;
- saved->saved_input = saved_input;
- saved->saved_input_len = saved_input_len;
+ if (saved_input == NULL)
+ saved->saved_input = NULL;
+ else
+ {
+ saved->saved_input = xmalloc (saved_input_len);
+ memcpy (saved->saved_input, saved_input, saved_input_len);
+ saved->saved_input_len = saved_input_len;
+ }
+#ifdef TC_M68K
saved->scrub_m68k_mri = scrub_m68k_mri;
+#endif
saved->mri_state = mri_state;
saved->mri_last_ch = mri_last_ch;
#if defined TC_ARM && defined OBJ_ELF
saved->symver_state = symver_state;
#endif
- /* do_scrub_begin() is not useful, just wastes time. */
+ /* do_scrub_begin() is not useful, just wastes time. */
state = 0;
saved_input = NULL;
return (char *) saved;
}
-void
+void
app_pop (arg)
char *arg;
{
register struct app_save *saved = (struct app_save *) arg;
- /* There is no do_scrub_end (). */
+ /* There is no do_scrub_end (). */
state = saved->state;
old_state = saved->old_state;
out_string = saved->out_string;
memcpy (out_buf, saved->out_buf, sizeof (out_buf));
add_newlines = saved->add_newlines;
- saved_input = saved->saved_input;
- saved_input_len = saved->saved_input_len;
+ if (saved->saved_input == NULL)
+ saved_input = NULL;
+ else
+ {
+ assert (saved->saved_input_len <= (int) (sizeof input_buffer));
+ memcpy (input_buffer, saved->saved_input, saved->saved_input_len);
+ saved_input = input_buffer;
+ saved_input_len = saved->saved_input_len;
+ free (saved->saved_input);
+ }
+#ifdef TC_M68K
scrub_m68k_mri = saved->scrub_m68k_mri;
+#endif
mri_state = saved->mri_state;
mri_last_ch = saved->mri_last_ch;
#if defined TC_ARM && defined OBJ_ELF
/* @@ This assumes that \n &c are the same on host and target. This is not
necessarily true. */
-static int
+static int
process_escape (ch)
int ch;
{
int
do_scrub_chars (get, tostart, tolen)
- int (*get) PARAMS ((char **));
+ int (*get) PARAMS ((char *, int));
char *tostart;
int tolen;
{
#ifdef TC_V850
12: After seeing a dash, looking for a second dash as a start of comment.
#endif
-#ifdef TC_M32R
+#ifdef DOUBLEBAR_PARALLEL
13: After seeing a vertical bar, looking for a second vertical bar as a parallel expression seperator.
#endif
*/
/* I added states 9 and 10 because the MIPS ECOFF assembler uses
constructs like ``.loc 1 20''. This was turning into ``.loc
120''. States 9 and 10 ensure that a space is never dropped in
- between characters which could appear in a identifier. Ian
+ between characters which could appear in an identifier. Ian
Taylor, ian@cygnus.com.
I added state 11 so that something like "Lfoo add %r25,%r26,%r27" works
/* This macro gets the next input character. */
-#define GET() \
- (from < fromend \
- ? * (unsigned char *) (from++) \
- : ((saved_input != NULL \
- ? (free (saved_input), \
- saved_input = NULL, \
- 0) \
- : 0), \
- fromlen = (*get) (&from), \
- fromend = from + fromlen, \
- (fromlen == 0 \
- ? EOF \
+#define GET() \
+ (from < fromend \
+ ? * (unsigned char *) (from++) \
+ : (saved_input = NULL, \
+ fromlen = (*get) (input_buffer, sizeof input_buffer), \
+ from = input_buffer, \
+ fromend = from + fromlen, \
+ (fromlen == 0 \
+ ? EOF \
: * (unsigned char *) (from++))))
/* This macro pushes a character back on the input stream. */
I don't want to make such a significant change to the assembler's
memory usage. */
-#define PUT(pch) \
- do \
- { \
- *to++ = (pch); \
- if (to >= toend) \
- goto tofull; \
- } \
+#define PUT(pch) \
+ do \
+ { \
+ *to++ = (pch); \
+ if (to >= toend) \
+ goto tofull; \
+ } \
while (0)
if (saved_input != NULL)
}
else
{
- fromlen = (*get) (&from);
+ fromlen = (*get) (input_buffer, sizeof input_buffer);
if (fromlen == 0)
return 0;
+ from = input_buffer;
fromend = from + fromlen;
}
ch = GET ();
if (ch == EOF)
{
- as_warn (_("end of file in string: inserted '\"'"));
+ as_warn (_("end of file in string; inserted '\"'"));
state = old_state;
UNGET ('\n');
PUT ('"');
break;
#if defined(IGNORE_NONSTANDARD_ESCAPES) | defined(ONLY_STANDARD_ESCAPES)
default:
- as_warn (_("Unknown escape '\\%c' in string: Ignored"), ch);
+ as_warn (_("unknown escape '\\%c' in string; ignored"), ch);
break;
#else /* ONLY_STANDARD_ESCAPES */
default:
#endif /* ONLY_STANDARD_ESCAPES */
case EOF:
- as_warn (_("End of file in string: '\"' inserted"));
+ as_warn (_("end of file in string; '\"' inserted"));
PUT ('"');
continue;
}
break;
}
+#ifdef KEEP_WHITE_AROUND_COLON
+ if (lex[ch] == LEX_IS_COLON)
+ {
+ /* Only keep this white if there's no white *after* the
+ colon. */
+ ch2 = GET ();
+ UNGET (ch2);
+ if (!IS_WHITESPACE (ch2))
+ {
+ state = 9;
+ UNGET (ch);
+ PUT (' ');
+ break;
+ }
+ }
+#endif
if (IS_COMMENT (ch)
|| ch == '/'
- || IS_LINE_SEPARATOR (ch))
+ || IS_LINE_SEPARATOR (ch)
+ || IS_PARALLEL_SEPARATOR (ch))
{
if (scrub_m68k_mri)
{
state = 10; /* Sp after symbol char */
goto recycle;
case 11:
- if (flag_m68k_mri
-#ifdef LABELS_WITHOUT_COLONS
- || 1
-#endif
- )
+ if (LABELS_WITHOUT_COLONS || flag_m68k_mri)
state = 1;
else
{
ch = ' ';
goto recycle;
}
+#ifdef DOUBLESLASH_LINE_COMMENTS
+ else if (ch2 == '/')
+ {
+ do
+ {
+ ch = GET ();
+ }
+ while (ch != EOF && !IS_NEWLINE (ch));
+ if (ch == EOF)
+ as_warn ("end of file in comment; newline inserted");
+ state = 0;
+ PUT ('\n');
+ break;
+ }
+#endif
else
{
if (ch2 != EOF)
if ((ch = GET ()) != '\'')
{
#ifdef REQUIRE_CHAR_CLOSE_QUOTE
- as_warn (_("Missing close quote: (assumed)"));
+ as_warn (_("missing close quote; (assumed)"));
#else
if (ch != EOF)
UNGET (ch);
#endif
case LEX_IS_COLON:
+#ifdef KEEP_WHITE_AROUND_COLON
+ state = 9;
+#else
if (state == 9 || state == 10)
state = 3;
else if (state != 3)
state = 1;
+#endif
PUT (ch);
break;
--add_newlines;
UNGET (ch);
}
- /* fall thru into... */
+ /* Fall through. */
case LEX_IS_LINE_SEPARATOR:
state = 0;
PUT (ch);
break;
+ case LEX_IS_PARALLEL_SEPARATOR:
+ state = 1;
+ PUT (ch);
+ break;
+
#ifdef TC_V850
case LEX_IS_DOUBLEDASH_1ST:
- ch2 = GET();
+ ch2 = GET ();
if (ch2 != '-')
{
UNGET (ch2);
goto de_fault;
}
- /* read and skip to end of line */
+ /* Read and skip to end of line. */
do
{
ch = GET ();
state = 0;
PUT ('\n');
break;
-#endif
-#ifdef TC_M32R
+#endif
+#ifdef DOUBLEBAR_PARALLEL
case LEX_IS_DOUBLEBAR_1ST:
- ch2 = GET();
+ ch2 = GET ();
if (ch2 != '|')
{
UNGET (ch2);
PUT ('|');
PUT ('|');
break;
-#endif
+#endif
case LEX_IS_LINE_COMMENT_START:
/* FIXME-someday: The two character comment stuff was badly
thought out. On i386, we want '/' as line comment start
while (ch != EOF && !IS_NEWLINE (ch))
ch = GET ();
if (ch == EOF)
- as_warn (_("EOF in Comment: Newline inserted"));
+ as_warn (_("end of file in comment; newline inserted"));
state = 0;
PUT ('\n');
break;
}
- /* Loks like `# 123 "filename"' from cpp. */
+ /* Looks like `# 123 "filename"' from cpp. */
UNGET (ch);
old_state = 4;
state = -1;
Trap is the only short insn that has a first operand that is
neither register nor label.
We must prevent exef0f ||trap #1 to degenerate to exef0f ||trap#1 .
- We can't make '#' LEX_IS_SYMBOL_COMPONENT because it is already
- LEX_IS_LINE_COMMENT_START. However, it is the only character in
- line_comment_chars for d10v, hence we can recognize it as such. */
+ We can't make '#' LEX_IS_SYMBOL_COMPONENT because it is
+ already LEX_IS_LINE_COMMENT_START. However, it is the
+ only character in line_comment_chars for d10v, hence we
+ can recognize it as such. */
/* An alternative approach would be to reset the state to 1 when
we see '||', '<'- or '->', but that seems to be overkill. */
- if (state == 10) PUT (' ');
+ if (state == 10)
+ PUT (' ');
#endif
/* We have a line comment character which is not at the
start of a line. If this is also a normal comment
#if defined TC_ARM && defined OBJ_ELF
/* On the ARM, `@' is the comment character.
Unfortunately this is also a special character in ELF .symver
- directives (and .type, though we deal with those another way). So
- we check if this line is such a directive, and treat the character
- as default if so. This is a hack. */
+ directives (and .type, though we deal with those another way).
+ So we check if this line is such a directive, and treat
+ the character as default if so. This is a hack. */
if ((symver_state != NULL) && (*symver_state == 0))
goto de_fault;
+#endif
+#ifdef WARN_COMMENTS
+ if (!found_comment)
+ as_where (&found_comment_file, &found_comment);
#endif
do
{
{
int type;
- ch2 = * (unsigned char *) s;
+ ch2 = *(unsigned char *) s;
type = lex[ch2];
if (type != 0
&& type != LEX_IS_SYMBOL_COMPONENT)
case 2: *to++ = *from++;
case 1: *to++ = *from++;
}
- }
+ }
ch = GET ();
}
}
}
else if (state == 10)
{
+ if (ch == '\\')
+ {
+ /* Special handling for backslash: a backslash may
+ be the beginning of a formal parameter (of a
+ macro) following another symbol character, with
+ whitespace in between. If that is the case, we
+ output a space before the parameter. Strictly
+ speaking, correct handling depends upon what the
+ macro parameter expands into; if the parameter
+ expands into something which does not start with
+ an operand character, then we don't want to keep
+ the space. We don't have enough information to
+ make the right choice, so here we are making the
+ choice which is more likely to be correct. */
+ PUT (' ');
+ }
+
state = 3;
}
PUT (ch);
processed. */
if (fromend > from)
{
- char *save;
-
- save = (char *) xmalloc (fromend - from);
- memcpy (save, from, fromend - from);
- if (saved_input != NULL)
- free (saved_input);
- saved_input = save;
+ saved_input = from;
saved_input_len = fromend - from;
}
else
- {
- if (saved_input != NULL)
- {
- free (saved_input);
- saved_input = NULL;
- }
- }
+ saved_input = NULL;
+
return to - tostart;
}