diff options
author | Marc Espie <espie@cvs.openbsd.org> | 2000-09-12 19:13:03 +0000 |
---|---|---|
committer | Marc Espie <espie@cvs.openbsd.org> | 2000-09-12 19:13:03 +0000 |
commit | 1b50fce4c0ed748c156af3ac629e50cb5e4d0ef4 (patch) | |
tree | edee61faabd18b4a5d84e8cdb679d7a8d1fe171f /gnu/usr.bin/binutils/gas/app.c | |
parent | 9f1193e30b5f04af9ea81c644eec79b7b535b890 (diff) |
Help stupid cvs fixing basic conflicts.
Diffstat (limited to 'gnu/usr.bin/binutils/gas/app.c')
-rw-r--r-- | gnu/usr.bin/binutils/gas/app.c | 378 |
1 files changed, 288 insertions, 90 deletions
diff --git a/gnu/usr.bin/binutils/gas/app.c b/gnu/usr.bin/binutils/gas/app.c index ade363e1b26..2613e7440b3 100644 --- a/gnu/usr.bin/binutils/gas/app.c +++ b/gnu/usr.bin/binutils/gas/app.c @@ -1,5 +1,5 @@ /* This is the Assembler Pre-Processor - Copyright (C) 1987, 90, 91, 92, 93, 94, 95, 1996 + Copyright (C) 1987, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 2000 Free Software Foundation, Inc. This file is part of GAS, the GNU Assembler. @@ -15,8 +15,9 @@ GNU General Public License for more details. You should have received a copy of the GNU General Public License - along with GAS; see the file COPYING. If not, write to - the Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ + along with GAS; see the file COPYING. If not, write to the Free + Software Foundation, 59 Temple Place - Suite 330, Boston, MA + 02111-1307, USA. */ /* Modified by Allen Wirfs-Brock, Instantiations Inc 2/90 */ /* App, the assembler pre-processor. This pre-processor strips out excess @@ -33,15 +34,26 @@ #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. */ static int scrub_m68k_mri; +#else +#define scrub_m68k_mri 0 +#endif /* 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"; +#if defined TC_ARM && defined OBJ_ELF +/* 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; +#endif + static char lex[256]; static const char symbol_chars[] = "$._ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789"; @@ -52,11 +64,19 @@ static const char symbol_chars[] = #define LEX_IS_COMMENT_START 4 #define LEX_IS_LINE_COMMENT_START 5 #define LEX_IS_TWOCHAR_COMMENT_1ST 6 -#define LEX_IS_TWOCHAR_COMMENT_2ND 7 #define LEX_IS_STRINGQUOTE 8 #define LEX_IS_COLON 9 #define LEX_IS_NEWLINE 10 #define LEX_IS_ONECHAR_QUOTE 11 +#ifdef TC_V850 +#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 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) @@ -72,23 +92,28 @@ static int process_escape PARAMS ((int)); void do_scrub_begin (m68k_mri) - int m68k_mri; + int m68k_mri ATTRIBUTE_UNUSED; { const char *p; - - scrub_m68k_mri = m68k_mri; + int c; 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 @@ -107,6 +132,19 @@ do_scrub_begin (m68k_mri) lex[(unsigned char) *p] = LEX_IS_SYMBOL_COMPONENT; } /* declare symbol characters */ + for (c = 128; c < 256; ++c) + lex[c] = LEX_IS_SYMBOL_COMPONENT; + +#ifdef tc_symbol_chars + /* This macro permits the processor to specify all characters which + may appears in an operand. This will prevent the scrubber from + discarding meaningful whitespace in certain cases. The i386 + backend uses this to support prefixes, which can confuse the + scrubber as to whether it is parsing operands or opcodes. */ + for (p = tc_symbol_chars; *p; ++p) + lex[(unsigned char) *p] = LEX_IS_SYMBOL_COMPONENT; +#endif + /* The m68k backend wants to be able to change comment_chars. */ #ifndef tc_comment_chars #define tc_comment_chars comment_chars @@ -126,18 +164,14 @@ do_scrub_begin (m68k_mri) lex[(unsigned char) *p] = LEX_IS_LINE_SEPARATOR; } /* declare line separators */ - /* Only allow slash-star comments if slash is not in use */ + /* 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; } - /* FIXME-soon. This is a bad hack but otherwise, we can't do - c-style comments when '/' is a line comment char. xoxorich. */ - if (lex['*'] == 0) - { - lex['*'] = LEX_IS_TWOCHAR_COMMENT_2ND; - } +#ifdef TC_M68K if (m68k_mri) { lex['\''] = LEX_IS_STRINGQUOTE; @@ -147,6 +181,18 @@ do_scrub_begin (m68k_mri) 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 DOUBLEBAR_PARALLEL + lex['|'] = LEX_IS_DOUBLEBAR_1ST; +#endif +#ifdef TC_D30V + /* must do this is we want VLIW instruction with "->" or "<-" */ + lex['-'] = LEX_IS_SYMBOL_COMPONENT; +#endif } /* do_scrub_begin() */ /* Saved state of the scrubber */ @@ -157,6 +203,7 @@ static char out_buf[20]; 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; @@ -167,16 +214,21 @@ static 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; - int scrub_m68k_mri; - const char *mri_state; - char mri_last_ch; + 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; +#endif }; char * @@ -190,11 +242,22 @@ 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. */ @@ -216,11 +279,24 @@ app_pop (arg) 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 + symver_state = saved->symver_state; +#endif free (arg); } /* app_pop() */ @@ -265,7 +341,7 @@ process_escape (ch) int do_scrub_chars (get, tostart, tolen) - int (*get) PARAMS ((char **)); + int (*get) PARAMS ((char *, int)); char *tostart; int tolen; { @@ -290,6 +366,12 @@ do_scrub_chars (get, tostart, tolen) 11: After seeing a symbol character in state 0 (eg a label definition) -1: output string in out_string and go to the state in old_state -2: flush text until a '*' '/' is seen, then go to state old_state +#ifdef TC_V850 + 12: After seeing a dash, looking for a second dash as a start of comment. +#endif +#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 @@ -300,23 +382,24 @@ do_scrub_chars (get, tostart, tolen) I added state 11 so that something like "Lfoo add %r25,%r26,%r27" works correctly on the PA (and any other target where colons are optional). - Jeff Law, law@cs.utah.edu. */ + Jeff Law, law@cs.utah.edu. + + I added state 13 so that something like "cmp r1, r2 || trap #1" does not + get squashed into "cmp r1,r2||trap#1", with the all important space + between the 'trap' and the '#1' being eliminated. nickc@cygnus.com */ /* This macro gets the next input character. */ -#define GET() \ - (from < fromend \ - ? *from++ \ - : ((saved_input != NULL \ - ? (free (saved_input), \ - saved_input = NULL, \ - 0) \ - : 0), \ - fromlen = (*get) (&from), \ - fromend = from + fromlen, \ - (fromlen == 0 \ - ? EOF \ - : *from++))) +#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. */ @@ -347,9 +430,10 @@ do_scrub_chars (get, tostart, tolen) } else { - fromlen = (*get) (&from); + fromlen = (*get) (input_buffer, sizeof input_buffer); if (fromlen == 0) return 0; + from = input_buffer; fromend = from + fromlen; } @@ -379,7 +463,7 @@ do_scrub_chars (get, tostart, tolen) if (ch == EOF) { - as_warn ("end of file in comment"); + as_warn (_("end of file in comment")); goto fromeof; } @@ -393,7 +477,7 @@ do_scrub_chars (get, tostart, tolen) if (ch == EOF) { - as_warn ("end of file in comment"); + as_warn (_("end of file in comment")); goto fromeof; } @@ -404,7 +488,7 @@ do_scrub_chars (get, tostart, tolen) } state = old_state; - PUT (' '); + UNGET (' '); continue; case 4: @@ -472,7 +556,7 @@ do_scrub_chars (get, tostart, tolen) 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 ('"'); @@ -538,7 +622,7 @@ do_scrub_chars (get, tostart, tolen) 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: @@ -547,7 +631,7 @@ do_scrub_chars (get, tostart, tolen) #endif /* ONLY_STANDARD_ESCAPES */ case EOF: - as_warn ("End of file in string: '\"' inserted"); + as_warn (_("End of file in string: '\"' inserted")); PUT ('"'); continue; } @@ -581,6 +665,35 @@ do_scrub_chars (get, tostart, tolen) recycle: +#if defined TC_ARM && defined OBJ_ELF + /* We need to watch out for .symver directives. See the comment later + in this function. */ + if (symver_state == NULL) + { + if ((state == 0 || state == 1) && ch == symver_pseudo[0]) + symver_state = symver_pseudo + 1; + } + else + { + /* We advance to the next state if we find the right + character. */ + if (ch != '\0' && (*symver_state == ch)) + ++symver_state; + else if (*symver_state != '\0') + /* We did not get the expected character, or we didn't + get a valid terminating character after seeing the + entire pseudo-op, so we must go back to the beginning. */ + symver_state = NULL; + else + { + /* We've read the entire pseudo-op. If this is the end + of the line, go back to the beginning. */ + if (IS_NEWLINE (ch)) + symver_state = NULL; + } + } +#endif /* TC_ARM && OBJ_ELF */ + #ifdef TC_M68K /* We want to have pseudo-ops which control whether we are in MRI mode or not. Unfortunately, since m68k MRI mode affects @@ -625,6 +738,7 @@ do_scrub_chars (get, tostart, tolen) either '0' or '1' indicating whether to enter or leave MRI mode. */ do_scrub_begin (mri_last_ch == '1'); + mri_state = NULL; /* We continue handling the character as usual. The main gas reader must also handle the .mri pseudo-op @@ -637,7 +751,7 @@ do_scrub_chars (get, tostart, tolen) { if (state != 0) { - as_warn ("end of file not at end of a line; newline inserted"); + as_warn (_("end of file not at end of a line; newline inserted")); state = 0; PUT ('\n'); } @@ -665,6 +779,21 @@ do_scrub_chars (get, tostart, tolen) 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)) @@ -735,11 +864,7 @@ do_scrub_chars (get, tostart, tolen) 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 { @@ -759,7 +884,7 @@ do_scrub_chars (get, tostart, tolen) case LEX_IS_TWOCHAR_COMMENT_1ST: ch2 = GET (); - if (ch2 != EOF && lex[ch2] == LEX_IS_TWOCHAR_COMMENT_2ND) + if (ch2 == '*') { for (;;) { @@ -769,22 +894,21 @@ do_scrub_chars (get, tostart, tolen) if (ch2 != EOF && IS_NEWLINE (ch2)) add_newlines++; } - while (ch2 != EOF && - (lex[ch2] != LEX_IS_TWOCHAR_COMMENT_2ND)); + while (ch2 != EOF && ch2 != '*'); - while (ch2 != EOF && - (lex[ch2] == LEX_IS_TWOCHAR_COMMENT_2ND)) - { - ch2 = GET (); - } + while (ch2 == '*') + ch2 = GET (); - if (ch2 == EOF - || lex[ch2] == LEX_IS_TWOCHAR_COMMENT_1ST) + if (ch2 == EOF || ch2 == '/') break; + + /* This UNGET will ensure that we count newlines + correctly. */ UNGET (ch2); } + if (ch2 == EOF) - as_warn ("end of file in multiline comment"); + as_warn (_("end of file in multiline comment")); ch = ' '; goto recycle; @@ -833,7 +957,7 @@ do_scrub_chars (get, tostart, tolen) ch = GET (); if (ch == EOF) { - as_warn ("end of file after a one-character quote; \\0 inserted"); + as_warn (_("end of file after a one-character quote; \\0 inserted")); ch = 0; } if (ch == '\\') @@ -841,7 +965,7 @@ do_scrub_chars (get, tostart, tolen) ch = GET (); if (ch == EOF) { - as_warn ("end of file in escape character"); + as_warn (_("end of file in escape character")); ch = '\\'; } else @@ -853,7 +977,7 @@ do_scrub_chars (get, tostart, tolen) 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); @@ -875,10 +999,14 @@ do_scrub_chars (get, tostart, tolen) #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; @@ -896,6 +1024,43 @@ do_scrub_chars (get, tostart, tolen) PUT (ch); break; +#ifdef TC_V850 + case LEX_IS_DOUBLEDASH_1ST: + ch2 = GET(); + if (ch2 != '-') + { + UNGET (ch2); + goto de_fault; + } + /* read and skip to end of line */ + do + { + ch = GET (); + } + while (ch != EOF && ch != '\n'); + if (ch == EOF) + { + as_warn (_("end of file in comment; newline inserted")); + } + state = 0; + PUT ('\n'); + break; +#endif +#ifdef DOUBLEBAR_PARALLEL + case LEX_IS_DOUBLEBAR_1ST: + ch2 = GET(); + if (ch2 != '|') + { + UNGET (ch2); + goto de_fault; + } + /* Reset back to state 1 and pretend that we are parsing a line from + just after the first white space. */ + state = 1; + PUT ('|'); + PUT ('|'); + break; +#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 @@ -906,6 +1071,7 @@ do_scrub_chars (get, tostart, tolen) ch2 = GET (); if (ch2 == '*') { + old_state = 3; state = -2; break; } @@ -928,7 +1094,7 @@ do_scrub_chars (get, tostart, tolen) while (ch != EOF && IS_WHITESPACE (ch)); if (ch == EOF) { - as_warn ("end of file in comment; newline inserted"); + as_warn (_("end of file in comment; newline inserted")); PUT ('\n'); break; } @@ -938,7 +1104,7 @@ do_scrub_chars (get, tostart, tolen) while (ch != EOF && !IS_NEWLINE (ch)) ch = GET (); if (ch == EOF) - as_warn ("EOF in Comment: Newline inserted"); + as_warn (_("EOF in Comment: Newline inserted")); state = 0; PUT ('\n'); break; @@ -955,6 +1121,18 @@ do_scrub_chars (get, tostart, tolen) break; } +#ifdef TC_D10V + /* All insns end in a char for which LEX_IS_SYMBOL_COMPONENT is true. + 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. */ + /* 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 (' '); +#endif /* We have a line comment character which is not at the start of a line. If this is also a normal comment character, fall through. Otherwise treat it as a default @@ -970,13 +1148,22 @@ do_scrub_chars (get, tostart, tolen) goto de_fault; /* Fall through. */ case LEX_IS_COMMENT_START: +#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. */ + if ((symver_state != NULL) && (*symver_state == 0)) + goto de_fault; +#endif do { ch = GET (); } while (ch != EOF && !IS_NEWLINE (ch)); if (ch == EOF) - as_warn ("end of file in comment; newline inserted"); + as_warn (_("end of file in comment; newline inserted")); state = 0; PUT ('\n'); break; @@ -998,7 +1185,12 @@ do_scrub_chars (get, tostart, tolen) /* This is a common case. Quickly copy CH and all the following symbol component or normal characters. */ - if (to + 1 < toend && mri_state == NULL) + if (to + 1 < toend + && mri_state == NULL +#if defined TC_ARM && defined OBJ_ELF + && symver_state == NULL +#endif + ) { char *s; int len; @@ -1007,7 +1199,7 @@ do_scrub_chars (get, tostart, tolen) { int type; - ch2 = *s; + ch2 = * (unsigned char *) s; type = lex[ch2]; if (type != 0 && type != LEX_IS_SYMBOL_COMPONENT) @@ -1068,6 +1260,23 @@ do_scrub_chars (get, tostart, tolen) } 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); @@ -1086,23 +1295,12 @@ do_scrub_chars (get, tostart, tolen) 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; } |