Home | History | Annotate | Line # | Download | only in indent
lexi.c revision 1.155
      1 /*	$NetBSD: lexi.c,v 1.155 2021/11/25 16:09:02 rillig Exp $	*/
      2 
      3 /*-
      4  * SPDX-License-Identifier: BSD-4-Clause
      5  *
      6  * Copyright (c) 1985 Sun Microsystems, Inc.
      7  * Copyright (c) 1980, 1993
      8  *	The Regents of the University of California.  All rights reserved.
      9  * All rights reserved.
     10  *
     11  * Redistribution and use in source and binary forms, with or without
     12  * modification, are permitted provided that the following conditions
     13  * are met:
     14  * 1. Redistributions of source code must retain the above copyright
     15  *    notice, this list of conditions and the following disclaimer.
     16  * 2. Redistributions in binary form must reproduce the above copyright
     17  *    notice, this list of conditions and the following disclaimer in the
     18  *    documentation and/or other materials provided with the distribution.
     19  * 3. All advertising materials mentioning features or use of this software
     20  *    must display the following acknowledgement:
     21  *	This product includes software developed by the University of
     22  *	California, Berkeley and its contributors.
     23  * 4. Neither the name of the University nor the names of its contributors
     24  *    may be used to endorse or promote products derived from this software
     25  *    without specific prior written permission.
     26  *
     27  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     29  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     30  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     31  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     32  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     33  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     34  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     35  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     36  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     37  * SUCH DAMAGE.
     38  */
     39 
     40 #if 0
     41 static char sccsid[] = "@(#)lexi.c	8.1 (Berkeley) 6/6/93";
     42 #endif
     43 
     44 #include <sys/cdefs.h>
     45 #if defined(__NetBSD__)
     46 __RCSID("$NetBSD: lexi.c,v 1.155 2021/11/25 16:09:02 rillig Exp $");
     47 #elif defined(__FreeBSD__)
     48 __FBSDID("$FreeBSD: head/usr.bin/indent/lexi.c 337862 2018-08-15 18:19:45Z pstef $");
     49 #endif
     50 
     51 #include <stdlib.h>
     52 #include <string.h>
     53 
     54 #include "indent.h"
     55 
     56 /*
     57  * While inside lexi_alnum, this constant just marks a type, independently of
     58  * the parentheses level.
     59  */
     60 #define lsym_type lsym_type_outside_parentheses
     61 
     62 /* must be sorted alphabetically, is used in binary search */
     63 static const struct keyword {
     64     const char *name;
     65     lexer_symbol lsym;
     66 } keywords[] = {
     67     {"_Bool", lsym_type},
     68     {"_Complex", lsym_type},
     69     {"_Imaginary", lsym_type},
     70     {"auto", lsym_storage_class},
     71     {"bool", lsym_type},
     72     {"break", lsym_word},
     73     {"case", lsym_case_label},
     74     {"char", lsym_type},
     75     {"complex", lsym_type},
     76     {"const", lsym_type},
     77     {"continue", lsym_word},
     78     {"default", lsym_case_label},
     79     {"do", lsym_do},
     80     {"double", lsym_type},
     81     {"else", lsym_else},
     82     {"enum", lsym_tag},
     83     {"extern", lsym_storage_class},
     84     {"float", lsym_type},
     85     {"for", lsym_for},
     86     {"goto", lsym_word},
     87     {"if", lsym_if},
     88     {"imaginary", lsym_type},
     89     {"inline", lsym_word},
     90     {"int", lsym_type},
     91     {"long", lsym_type},
     92     {"offsetof", lsym_offsetof},
     93     {"register", lsym_storage_class},
     94     {"restrict", lsym_word},
     95     {"return", lsym_return},
     96     {"short", lsym_type},
     97     {"signed", lsym_type},
     98     {"sizeof", lsym_sizeof},
     99     {"static", lsym_storage_class},
    100     {"struct", lsym_tag},
    101     {"switch", lsym_switch},
    102     {"typedef", lsym_typedef},
    103     {"union", lsym_tag},
    104     {"unsigned", lsym_type},
    105     {"void", lsym_type},
    106     {"volatile", lsym_type},
    107     {"while", lsym_while}
    108 };
    109 
    110 static struct {
    111     const char **items;
    112     unsigned int len;
    113     unsigned int cap;
    114 } typenames;
    115 
    116 /*
    117  * The transition table below was rewritten by hand from lx's output, given
    118  * the following definitions. lx is Katherine Flavel's lexer generator.
    119  *
    120  * O  = /[0-7]/;        D  = /[0-9]/;          NZ = /[1-9]/;
    121  * H  = /[a-f0-9]/i;    B  = /[0-1]/;          HP = /0x/i;
    122  * BP = /0b/i;          E  = /e[+\-]?/i D+;    P  = /p[+\-]?/i D+;
    123  * FS = /[fl]/i;        IS = /u/i /(l|L|ll|LL)/? | /(l|L|ll|LL)/ /u/i?;
    124  *
    125  * D+           E  FS? -> $float;
    126  * D*    "." D+ E? FS? -> $float;
    127  * D+    "."    E? FS? -> $float;    HP H+           IS? -> $int;
    128  * HP H+        P  FS? -> $float;    NZ D*           IS? -> $int;
    129  * HP H* "." H+ P  FS? -> $float;    "0" O*          IS? -> $int;
    130  * HP H+ "."    P  FS  -> $float;    BP B+           IS? -> $int;
    131  */
    132 /* INDENT OFF */
    133 static const unsigned char lex_number_state[][26] = {
    134     /*                examples:
    135                                      00
    136              s                      0xx
    137              t                    00xaa
    138              a     11       101100xxa..
    139              r   11ee0001101lbuuxx.a.pp
    140              t.01.e+008bLuxll0Ll.aa.p+0
    141     states:  ABCDEFGHIJKLMNOPQRSTUVWXYZ */
    142     [0] =   "uuiifuufiuuiiuiiiiiuiuuuuu",	/* (other) */
    143     [1] =   "CEIDEHHHIJQ  U  Q  VUVVZZZ",	/* 0 */
    144     [2] =   "DEIDEHHHIJQ  U  Q  VUVVZZZ",	/* 1 */
    145     [3] =   "DEIDEHHHIJ   U     VUVVZZZ",	/* 2 3 4 5 6 7 */
    146     [4] =   "DEJDEHHHJJ   U     VUVVZZZ",	/* 8 9 */
    147     [5] =   "             U     VUVV   ",	/* A a C c D d */
    148     [6] =   "  K          U     VUVV   ",	/* B b */
    149     [7] =   "  FFF   FF   U     VUVV   ",	/* E e */
    150     [8] =   "    f  f     U     VUVV  f",	/* F f */
    151     [9] =   "  LLf  fL  PR   Li  L    f",	/* L */
    152     [10] =  "  OOf  fO   S P O i O    f",	/* l */
    153     [11] =  "                    FFX   ",	/* P p */
    154     [12] =  "  MM    M  i  iiM   M     ",	/* U u */
    155     [13] =  "  N                       ",	/* X x */
    156     [14] =  "     G                 Y  ",	/* + - */
    157     [15] =  "B EE    EE   T      W     ",	/* . */
    158     /*       ABCDEFGHIJKLMNOPQRSTUVWXYZ */
    159 };
    160 /* INDENT ON */
    161 
    162 static const unsigned char lex_number_row[] = {
    163     ['0'] = 1,
    164     ['1'] = 2,
    165     ['2'] = 3, ['3'] = 3, ['4'] = 3, ['5'] = 3, ['6'] = 3, ['7'] = 3,
    166     ['8'] = 4, ['9'] = 4,
    167     ['A'] = 5, ['a'] = 5, ['C'] = 5, ['c'] = 5, ['D'] = 5, ['d'] = 5,
    168     ['B'] = 6, ['b'] = 6,
    169     ['E'] = 7, ['e'] = 7,
    170     ['F'] = 8, ['f'] = 8,
    171     ['L'] = 9,
    172     ['l'] = 10,
    173     ['P'] = 11, ['p'] = 11,
    174     ['U'] = 12, ['u'] = 12,
    175     ['X'] = 13, ['x'] = 13,
    176     ['+'] = 14, ['-'] = 14,
    177     ['.'] = 15,
    178 };
    179 
    180 static void
    181 check_size_token(size_t desired_size)
    182 {
    183     if (token.e + desired_size >= token.l)
    184 	buf_expand(&token, desired_size);
    185 }
    186 
    187 static void
    188 token_add_char(char ch)
    189 {
    190     check_size_token(1);
    191     *token.e++ = ch;
    192 }
    193 
    194 #ifdef debug
    195 static const char *
    196 lsym_name(lexer_symbol sym)
    197 {
    198     static const char *const name[] = {
    199 	"eof",
    200 	"preprocessing",
    201 	"newline",
    202 	"form_feed",
    203 	"comment",
    204 	"lparen_or_lbracket",
    205 	"rparen_or_rbracket",
    206 	"lbrace",
    207 	"rbrace",
    208 	"period",
    209 	"unary_op",
    210 	"binary_op",
    211 	"postfix_op",
    212 	"question",
    213 	"colon",
    214 	"comma",
    215 	"semicolon",
    216 	"typedef",
    217 	"storage_class",
    218 	"type_outside_parentheses",
    219 	"type_in_parentheses",
    220 	"tag",
    221 	"case_label",
    222 	"string_prefix",
    223 	"sizeof",
    224 	"offsetof",
    225 	"word",
    226 	"funcname",
    227 	"do",
    228 	"else",
    229 	"for",
    230 	"if",
    231 	"switch",
    232 	"while",
    233 	"return",
    234     };
    235 
    236     return name[sym];
    237 }
    238 
    239 static void
    240 debug_print_buf(const char *name, const struct buffer *buf)
    241 {
    242     if (buf->s < buf->e) {
    243 	debug_printf("%s ", name);
    244 	debug_vis_range("\"", buf->s, buf->e, "\"\n");
    245     }
    246 }
    247 
    248 #define debug_ps_bool(name) \
    249         if (ps.name != prev_ps.name) \
    250 	    debug_println("[%c] ps." #name, ps.name ? 'x' : ' ')
    251 #define debug_ps_int(name) \
    252 	if (ps.name != prev_ps.name) \
    253 	    debug_println("%3d ps." #name, ps.name)
    254 
    255 static void
    256 debug_lexi(lexer_symbol lsym)
    257 {
    258     /*
    259      * Watch out for 'rolled back parser state' in the debug output; the
    260      * differences around these are unreliable.
    261      */
    262     static struct parser_state prev_ps;
    263 
    264     debug_println("");
    265     debug_printf("line %d: %s", line_no, lsym_name(lsym));
    266     debug_vis_range(" \"", token.s, token.e, "\"\n");
    267 
    268     debug_print_buf("label", &lab);
    269     debug_print_buf("code", &code);
    270     debug_print_buf("comment", &com);
    271 
    272     debug_println("    ps.prev_token = %s", lsym_name(ps.prev_token));
    273     debug_ps_bool(next_col_1);
    274     debug_ps_bool(curr_col_1);
    275     debug_ps_bool(next_unary);
    276     debug_ps_bool(is_function_definition);
    277     debug_ps_bool(want_blank);
    278     debug_ps_int(paren_level);
    279     debug_ps_int(p_l_follow);
    280     if (ps.paren_level != prev_ps.paren_level) {
    281 	debug_printf("    ps.paren_indents:");
    282 	for (int i = 0; i < ps.paren_level; i++)
    283 	    debug_printf(" %d", ps.paren_indents[i]);
    284 	debug_println("");
    285     }
    286     debug_ps_int(cast_mask);
    287     debug_ps_int(not_cast_mask);
    288 
    289     debug_ps_int(comment_delta);
    290     debug_ps_int(n_comment_delta);
    291     debug_ps_int(com_ind);
    292 
    293     debug_ps_bool(block_init);
    294     debug_ps_int(block_init_level);
    295     debug_ps_bool(init_or_struct);
    296 
    297     debug_ps_int(ind_level);
    298     debug_ps_int(ind_level_follow);
    299 
    300     debug_ps_int(decl_level);
    301     debug_ps_bool(decl_on_line);
    302     debug_ps_bool(in_decl);
    303     debug_ps_int(just_saw_decl);
    304     debug_ps_bool(in_parameter_declaration);
    305     debug_ps_bool(decl_indent_done);
    306 
    307     debug_ps_bool(in_stmt_or_decl);
    308     debug_ps_bool(in_stmt_cont);
    309     debug_ps_bool(is_case_label);
    310 
    311     debug_ps_bool(search_stmt);
    312 
    313     prev_ps = ps;
    314 }
    315 #endif
    316 
    317 static lexer_symbol
    318 lexi_end(lexer_symbol lsym)
    319 {
    320 #ifdef debug
    321     debug_lexi(lsym);
    322 #endif
    323     return lsym;
    324 }
    325 
    326 static void
    327 lex_number(void)
    328 {
    329     for (unsigned char s = 'A'; s != 'f' && s != 'i' && s != 'u';) {
    330 	unsigned char ch = (unsigned char)inp_peek();
    331 	if (ch >= array_length(lex_number_row) || lex_number_row[ch] == 0)
    332 	    break;
    333 
    334 	unsigned char row = lex_number_row[ch];
    335 	if (lex_number_state[row][s - 'A'] == ' ') {
    336 	    /*-
    337 	     * lex_number_state[0][s - 'A'] now indicates the type:
    338 	     * f = floating, i = integer, u = unknown
    339 	     */
    340 	    return;
    341 	}
    342 
    343 	s = lex_number_state[row][s - 'A'];
    344 	token_add_char(inp_next());
    345     }
    346 }
    347 
    348 static bool
    349 is_identifier_start(char ch)
    350 {
    351     return ch_isalpha(ch) || ch == '_' || ch == '$';
    352 }
    353 
    354 static bool
    355 is_identifier_part(char ch)
    356 {
    357     return ch_isalnum(ch) || ch == '_' || ch == '$';
    358 }
    359 
    360 static void
    361 lex_word(void)
    362 {
    363     for (;;) {
    364 	if (is_identifier_part(inp_peek()))
    365 	    token_add_char(inp_next());
    366 	else if (inp_peek() == '\\' && inp_lookahead(1) == '\n') {
    367 	    inp_skip();
    368 	    inp_skip();
    369 	} else
    370 	    return;
    371     }
    372 }
    373 
    374 static void
    375 lex_char_or_string(void)
    376 {
    377     for (char delim = token.e[-1];;) {
    378 	if (inp_peek() == '\n') {
    379 	    diag(1, "Unterminated literal");
    380 	    return;
    381 	}
    382 
    383 	token_add_char(inp_next());
    384 	if (token.e[-1] == delim)
    385 	    return;
    386 
    387 	if (token.e[-1] == '\\') {
    388 	    if (inp_peek() == '\n')
    389 		++line_no;
    390 	    token_add_char(inp_next());
    391 	}
    392     }
    393 }
    394 
    395 /* Guess whether the current token is a declared type. */
    396 static bool
    397 probably_typename(void)
    398 {
    399     if (ps.prev_token == lsym_storage_class)
    400 	return true;
    401     if (ps.block_init)
    402 	return false;
    403     if (ps.in_stmt_or_decl)	/* XXX: this condition looks incorrect */
    404 	return false;
    405     if (inp_peek() == '*' && inp_lookahead(1) != '=')
    406 	goto maybe;
    407     /* XXX: is_identifier_start */
    408     if (ch_isalpha(inp_peek()))
    409 	goto maybe;
    410     return false;
    411 maybe:
    412     return ps.prev_token == lsym_semicolon ||
    413 	ps.prev_token == lsym_lbrace ||
    414 	ps.prev_token == lsym_rbrace;
    415 }
    416 
    417 static int
    418 bsearch_typenames(const char *key)
    419 {
    420     const char **arr = typenames.items;
    421     int lo = 0;
    422     int hi = (int)typenames.len - 1;
    423 
    424     while (lo <= hi) {
    425 	int mid = (int)((unsigned)(lo + hi) >> 1);
    426 	int cmp = strcmp(arr[mid], key);
    427 	if (cmp < 0)
    428 	    lo = mid + 1;
    429 	else if (cmp > 0)
    430 	    hi = mid - 1;
    431 	else
    432 	    return mid;
    433     }
    434     return -(lo + 1);
    435 }
    436 
    437 static bool
    438 is_typename(void)
    439 {
    440     if (opt.auto_typedefs &&
    441 	token.e - token.s >= 2 && memcmp(token.e - 2, "_t", 2) == 0)
    442 	return true;
    443 
    444     return bsearch_typenames(token.s) >= 0;
    445 }
    446 
    447 static int
    448 cmp_keyword_by_name(const void *key, const void *elem)
    449 {
    450     return strcmp(key, ((const struct keyword *)elem)->name);
    451 }
    452 
    453 static bool
    454 probably_looking_at_definition(void)
    455 {
    456     for (const char *p = inp_p(), *e = inp_line_end(); p < e;)
    457 	if (*p++ == ')' && (*p == ';' || *p == ','))
    458 	    return false;
    459     return true;
    460 }
    461 
    462 /* Read an alphanumeric token into 'token', or return lsym_eof. */
    463 static lexer_symbol
    464 lexi_alnum(void)
    465 {
    466     if (ch_isdigit(inp_peek()) ||
    467 	    (inp_peek() == '.' && ch_isdigit(inp_lookahead(1)))) {
    468 	lex_number();
    469     } else if (is_identifier_part(inp_peek())) {
    470 	lex_word();
    471     } else
    472 	return lsym_eof;	/* just as a placeholder */
    473 
    474     *token.e = '\0';
    475 
    476     if (token.s[0] == 'L' && token.s[1] == '\0' &&
    477 	    (inp_peek() == '"' || inp_peek() == '\''))
    478 	return lsym_string_prefix;
    479 
    480     while (ch_isblank(inp_peek()))
    481 	inp_skip();
    482 
    483     ps.next_unary = ps.prev_token == lsym_tag;	/* for 'struct s *' */
    484 
    485     if (ps.prev_token == lsym_tag && ps.p_l_follow == 0)
    486 	return lsym_type_outside_parentheses;
    487 
    488     const struct keyword *kw = bsearch(token.s, keywords,
    489 	array_length(keywords), sizeof(keywords[0]), cmp_keyword_by_name);
    490     bool is_type = false;
    491     if (kw == NULL) {
    492 	if (is_typename()) {
    493 	    is_type = true;
    494 	    ps.next_unary = true;
    495 	    goto found_typename;
    496 	}
    497 
    498     } else {			/* we have a keyword */
    499 	is_type = kw->lsym == lsym_type;
    500 	ps.next_unary = true;
    501 	if (kw->lsym != lsym_tag && kw->lsym != lsym_type)
    502 	    return kw->lsym;
    503 
    504 found_typename:
    505 	if (ps.p_l_follow > 0) {
    506 	    /* inside parentheses: cast, param list, offsetof or sizeof */
    507 	    ps.cast_mask |= (1 << ps.p_l_follow) & ~ps.not_cast_mask;
    508 	}
    509 	if (ps.prev_token != lsym_period && ps.prev_token != lsym_unary_op) {
    510 	    if (kw != NULL && kw->lsym == lsym_tag)
    511 		return lsym_tag;
    512 	    if (ps.p_l_follow == 0)
    513 		return lsym_type_outside_parentheses;
    514 	}
    515     }
    516 
    517     if (inp_peek() == '(' && ps.tos <= 1 && ps.ind_level == 0 &&
    518 	!ps.in_parameter_declaration && !ps.block_init) {
    519 
    520 	if (probably_looking_at_definition()) {
    521 	    ps.is_function_definition = true;
    522 	    if (ps.in_decl)
    523 		ps.in_parameter_declaration = true;
    524 	    return lsym_funcname;
    525 	}
    526 
    527     } else if (ps.p_l_follow == 0 && probably_typename()) {
    528 	ps.next_unary = true;
    529 	return lsym_type_outside_parentheses;
    530     }
    531 
    532     return is_type ? lsym_type_in_parentheses : lsym_word;
    533 }
    534 
    535 /* Reads the next token, placing it in the global variable "token". */
    536 lexer_symbol
    537 lexi(void)
    538 {
    539     token.e = token.s;
    540     ps.curr_col_1 = ps.next_col_1;
    541     ps.next_col_1 = false;
    542 
    543     while (ch_isblank(inp_peek())) {
    544 	ps.curr_col_1 = false;
    545 	inp_skip();
    546     }
    547 
    548     lexer_symbol alnum_lsym = lexi_alnum();
    549     if (alnum_lsym != lsym_eof)
    550 	return lexi_end(alnum_lsym);
    551 
    552     /* Scan a non-alphanumeric token */
    553 
    554     check_size_token(3);	/* for things like "<<=" */
    555     *token.e++ = inp_next();
    556     *token.e = '\0';
    557 
    558     lexer_symbol lsym;
    559     bool unary_delim = false;	/* whether the current token forces a
    560 				 * following operator to be unary */
    561 
    562     switch (token.e[-1]) {
    563     case '\n':
    564 	unary_delim = ps.next_unary;
    565 	ps.next_col_1 = true;
    566 	/* if data has been exhausted, the newline is a dummy. */
    567 	lsym = had_eof ? lsym_eof : lsym_newline;
    568 	break;
    569 
    570     case '\'':
    571     case '"':
    572 	lex_char_or_string();
    573 	lsym = lsym_word;
    574 	break;
    575 
    576     case '(':
    577     case '[':
    578 	unary_delim = true;
    579 	lsym = lsym_lparen_or_lbracket;
    580 	break;
    581 
    582     case ')':
    583     case ']':
    584 	lsym = lsym_rparen_or_rbracket;
    585 	break;
    586 
    587     case '#':
    588 	unary_delim = ps.next_unary;
    589 	lsym = lsym_preprocessing;
    590 	break;
    591 
    592     case '?':
    593 	unary_delim = true;
    594 	lsym = lsym_question;
    595 	break;
    596 
    597     case ':':
    598 	lsym = lsym_colon;
    599 	unary_delim = true;
    600 	break;
    601 
    602     case ';':
    603 	unary_delim = true;
    604 	lsym = lsym_semicolon;
    605 	break;
    606 
    607     case '{':
    608 	unary_delim = true;
    609 	lsym = lsym_lbrace;
    610 	break;
    611 
    612     case '}':
    613 	unary_delim = true;
    614 	lsym = lsym_rbrace;
    615 	break;
    616 
    617     case '\f':
    618 	unary_delim = ps.next_unary;
    619 	ps.next_col_1 = true;
    620 	lsym = lsym_form_feed;
    621 	break;
    622 
    623     case ',':
    624 	unary_delim = true;
    625 	lsym = lsym_comma;
    626 	break;
    627 
    628     case '.':
    629 	unary_delim = false;
    630 	lsym = lsym_period;
    631 	break;
    632 
    633     case '-':
    634     case '+':
    635 	lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op;
    636 	unary_delim = true;
    637 
    638 	if (inp_peek() == token.e[-1]) {	/* ++, -- */
    639 	    *token.e++ = inp_next();
    640 	    if (ps.prev_token == lsym_word ||
    641 		    ps.prev_token == lsym_rparen_or_rbracket) {
    642 		lsym = ps.next_unary ? lsym_unary_op : lsym_postfix_op;
    643 		unary_delim = false;
    644 	    }
    645 
    646 	} else if (inp_peek() == '=') {	/* += */
    647 	    *token.e++ = inp_next();
    648 
    649 	} else if (inp_peek() == '>') {	/* -> */
    650 	    *token.e++ = inp_next();
    651 	    unary_delim = false;
    652 	    lsym = lsym_unary_op;
    653 	    ps.want_blank = false;
    654 	}
    655 	break;
    656 
    657     case '=':
    658 	if (ps.init_or_struct)
    659 	    ps.block_init = true;
    660 	if (inp_peek() == '=') {	/* == */
    661 	    *token.e++ = inp_next();
    662 	    *token.e = '\0';
    663 	}
    664 	lsym = lsym_binary_op;
    665 	unary_delim = true;
    666 	break;
    667 
    668     case '>':
    669     case '<':
    670     case '!':			/* ops like <, <<, <=, !=, etc */
    671 	if (inp_peek() == '>' || inp_peek() == '<' || inp_peek() == '=')
    672 	    *token.e++ = inp_next();
    673 	if (inp_peek() == '=')
    674 	    *token.e++ = inp_next();
    675 	lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op;
    676 	unary_delim = true;
    677 	break;
    678 
    679     case '*':
    680 	unary_delim = true;
    681 	if (!ps.next_unary) {
    682 	    if (inp_peek() == '=')
    683 		*token.e++ = inp_next();
    684 	    lsym = lsym_binary_op;
    685 	    break;
    686 	}
    687 
    688 	while (inp_peek() == '*' || ch_isspace(inp_peek())) {
    689 	    if (inp_peek() == '*')
    690 		token_add_char('*');
    691 	    inp_skip();
    692 	}
    693 
    694 	if (ps.in_decl) {
    695 	    const char *tp = inp_p(), *e = inp_line_end();
    696 
    697 	    while (tp < e) {
    698 		if (ch_isspace(*tp))
    699 		    tp++;
    700 		else if (is_identifier_start(*tp)) {
    701 		    tp++;
    702 		    while (tp < e && is_identifier_part(*tp))
    703 			tp++;
    704 		} else
    705 		    break;
    706 	    }
    707 
    708 	    if (tp < e && *tp == '(')
    709 		ps.is_function_definition = true;
    710 	}
    711 
    712 	lsym = lsym_unary_op;
    713 	break;
    714 
    715     default:
    716 	if (token.e[-1] == '/' && (inp_peek() == '*' || inp_peek() == '/')) {
    717 	    *token.e++ = inp_next();
    718 	    lsym = lsym_comment;
    719 	    unary_delim = ps.next_unary;
    720 	    break;
    721 	}
    722 
    723 	/* handle '||', '&&', etc., and also things as in 'int *****i' */
    724 	while (token.e[-1] == inp_peek() || inp_peek() == '=')
    725 	    token_add_char(inp_next());
    726 
    727 	lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op;
    728 	unary_delim = true;
    729     }
    730 
    731     ps.next_unary = unary_delim;
    732 
    733     check_size_token(1);
    734     *token.e = '\0';
    735 
    736     return lexi_end(lsym);
    737 }
    738 
    739 void
    740 register_typename(const char *name)
    741 {
    742     if (typenames.len >= typenames.cap) {
    743 	typenames.cap = 16 + 2 * typenames.cap;
    744 	typenames.items = xrealloc(typenames.items,
    745 	    sizeof(typenames.items[0]) * typenames.cap);
    746     }
    747 
    748     int pos = bsearch_typenames(name);
    749     if (pos >= 0)
    750 	return;			/* already in the list */
    751 
    752     pos = -(pos + 1);
    753     memmove(typenames.items + pos + 1, typenames.items + pos,
    754 	sizeof(typenames.items[0]) * (typenames.len++ - (unsigned)pos));
    755     typenames.items[pos] = xstrdup(name);
    756 }
    757