Home | History | Annotate | Line # | Download | only in indent
lexi.c revision 1.143
      1 /*	$NetBSD: lexi.c,v 1.143 2021/11/19 17:30:10 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 1.143 2021/11/19 17:30:10 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 <assert.h>
     52 #include <ctype.h>
     53 #include <stdlib.h>
     54 #include <string.h>
     55 
     56 #include "indent.h"
     57 
     58 /*
     59  * While inside lexi_alnum, this constant just marks a type, independently of
     60  * the parentheses level.
     61  */
     62 #define lsym_type lsym_type_outside_parentheses
     63 
     64 /* must be sorted alphabetically, is used in binary search */
     65 static const struct keyword {
     66     const char *name;
     67     lexer_symbol lsym;
     68 } keywords[] = {
     69     {"_Bool", lsym_type},
     70     {"_Complex", lsym_type},
     71     {"_Imaginary", lsym_type},
     72     {"auto", lsym_storage_class},
     73     {"bool", lsym_type},
     74     {"break", lsym_word},
     75     {"case", lsym_case_label},
     76     {"char", lsym_type},
     77     {"complex", lsym_type},
     78     {"const", lsym_type},
     79     {"continue", lsym_word},
     80     {"default", lsym_case_label},
     81     {"do", lsym_do},
     82     {"double", lsym_type},
     83     {"else", lsym_else},
     84     {"enum", lsym_tag},
     85     {"extern", lsym_storage_class},
     86     {"float", lsym_type},
     87     {"for", lsym_for},
     88     {"goto", lsym_word},
     89     {"if", lsym_if},
     90     {"imaginary", lsym_type},
     91     {"inline", lsym_word},
     92     {"int", lsym_type},
     93     {"long", lsym_type},
     94     {"offsetof", lsym_offsetof},
     95     {"register", lsym_storage_class},
     96     {"restrict", lsym_word},
     97     {"return", lsym_return},
     98     {"short", lsym_type},
     99     {"signed", lsym_type},
    100     {"sizeof", lsym_sizeof},
    101     {"static", lsym_storage_class},
    102     {"struct", lsym_tag},
    103     {"switch", lsym_switch},
    104     {"typedef", lsym_typedef},
    105     {"union", lsym_tag},
    106     {"unsigned", lsym_type},
    107     {"void", lsym_type},
    108     {"volatile", lsym_type},
    109     {"while", lsym_while}
    110 };
    111 
    112 static struct {
    113     const char **items;
    114     unsigned int len;
    115     unsigned int cap;
    116 } typenames;
    117 
    118 /*
    119  * The transition table below was rewritten by hand from lx's output, given
    120  * the following definitions. lx is Katherine Flavel's lexer generator.
    121  *
    122  * O  = /[0-7]/;        D  = /[0-9]/;          NZ = /[1-9]/;
    123  * H  = /[a-f0-9]/i;    B  = /[0-1]/;          HP = /0x/i;
    124  * BP = /0b/i;          E  = /e[+\-]?/i D+;    P  = /p[+\-]?/i D+;
    125  * FS = /[fl]/i;        IS = /u/i /(l|L|ll|LL)/? | /(l|L|ll|LL)/ /u/i?;
    126  *
    127  * D+           E  FS? -> $float;
    128  * D*    "." D+ E? FS? -> $float;
    129  * D+    "."    E? FS? -> $float;    HP H+           IS? -> $int;
    130  * HP H+        P  FS? -> $float;    NZ D*           IS? -> $int;
    131  * HP H* "." H+ P  FS? -> $float;    "0" O*          IS? -> $int;
    132  * HP H+ "."    P  FS  -> $float;    BP B+           IS? -> $int;
    133  */
    134 /* INDENT OFF */
    135 static const unsigned char lex_number_state[][26] = {
    136     /*                examples:
    137                                      00
    138              s                      0xx
    139              t                    00xaa
    140              a     11       101100xxa..
    141              r   11ee0001101lbuuxx.a.pp
    142              t.01.e+008bLuxll0Ll.aa.p+0
    143     states:  ABCDEFGHIJKLMNOPQRSTUVWXYZ */
    144     [0] =   "uuiifuufiuuiiuiiiiiuiuuuuu",	/* (other) */
    145     [1] =   "CEIDEHHHIJQ  U  Q  VUVVZZZ",	/* 0 */
    146     [2] =   "DEIDEHHHIJQ  U  Q  VUVVZZZ",	/* 1 */
    147     [3] =   "DEIDEHHHIJ   U     VUVVZZZ",	/* 2 3 4 5 6 7 */
    148     [4] =   "DEJDEHHHJJ   U     VUVVZZZ",	/* 8 9 */
    149     [5] =   "             U     VUVV   ",	/* A a C c D d */
    150     [6] =   "  K          U     VUVV   ",	/* B b */
    151     [7] =   "  FFF   FF   U     VUVV   ",	/* E e */
    152     [8] =   "    f  f     U     VUVV  f",	/* F f */
    153     [9] =   "  LLf  fL  PR   Li  L    f",	/* L */
    154     [10] =  "  OOf  fO   S P O i O    f",	/* l */
    155     [11] =  "                    FFX   ",	/* P p */
    156     [12] =  "  MM    M  i  iiM   M     ",	/* U u */
    157     [13] =  "  N                       ",	/* X x */
    158     [14] =  "     G                 Y  ",	/* + - */
    159     [15] =  "B EE    EE   T      W     ",	/* . */
    160     /*       ABCDEFGHIJKLMNOPQRSTUVWXYZ */
    161 };
    162 /* INDENT ON */
    163 
    164 static const unsigned char lex_number_row[] = {
    165     ['0'] = 1,
    166     ['1'] = 2,
    167     ['2'] = 3, ['3'] = 3, ['4'] = 3, ['5'] = 3, ['6'] = 3, ['7'] = 3,
    168     ['8'] = 4, ['9'] = 4,
    169     ['A'] = 5, ['a'] = 5, ['C'] = 5, ['c'] = 5, ['D'] = 5, ['d'] = 5,
    170     ['B'] = 6, ['b'] = 6,
    171     ['E'] = 7, ['e'] = 7,
    172     ['F'] = 8, ['f'] = 8,
    173     ['L'] = 9,
    174     ['l'] = 10,
    175     ['P'] = 11, ['p'] = 11,
    176     ['U'] = 12, ['u'] = 12,
    177     ['X'] = 13, ['x'] = 13,
    178     ['+'] = 14, ['-'] = 14,
    179     ['.'] = 15,
    180 };
    181 
    182 static void
    183 check_size_token(size_t desired_size)
    184 {
    185     if (token.e + desired_size >= token.l)
    186 	buf_expand(&token, desired_size);
    187 }
    188 
    189 static void
    190 token_add_char(char ch)
    191 {
    192     check_size_token(1);
    193     *token.e++ = ch;
    194 }
    195 
    196 #ifdef debug
    197 static const char *
    198 lsym_name(lexer_symbol sym)
    199 {
    200     static const char *const name[] = {
    201 	"eof",
    202 	"preprocessing",
    203 	"newline",
    204 	"form_feed",
    205 	"comment",
    206 	"lparen_or_lbracket",
    207 	"rparen_or_rbracket",
    208 	"lbrace",
    209 	"rbrace",
    210 	"period",
    211 	"unary_op",
    212 	"binary_op",
    213 	"postfix_op",
    214 	"question",
    215 	"colon",
    216 	"comma",
    217 	"semicolon",
    218 	"typedef",
    219 	"storage_class",
    220 	"type_outside_parentheses",
    221 	"type_in_parentheses",
    222 	"tag",
    223 	"case_label",
    224 	"string_prefix",
    225 	"sizeof",
    226 	"offsetof",
    227 	"word",
    228 	"funcname",
    229 	"do",
    230 	"else",
    231 	"for",
    232 	"if",
    233 	"switch",
    234 	"while",
    235 	"return",
    236     };
    237 
    238     return name[sym];
    239 }
    240 
    241 static void
    242 debug_print_buf(const char *name, const struct buffer *buf)
    243 {
    244     if (buf->s < buf->e) {
    245 	debug_printf("%s ", name);
    246 	debug_vis_range("\"", buf->s, buf->e, "\"\n");
    247     }
    248 }
    249 
    250 #define debug_ps_bool(name) \
    251         if (ps.name != prev_ps.name) \
    252 	    debug_println("[%c] ps." #name, ps.name ? 'x' : ' ')
    253 #define debug_ps_int(name) \
    254 	if (ps.name != prev_ps.name) \
    255 	    debug_println("%3d ps." #name, ps.name)
    256 
    257 static void
    258 debug_lexi(lexer_symbol lsym)
    259 {
    260     /*
    261      * Watch out for 'rolled back parser state' in the debug output; the
    262      * differences around these are unreliable.
    263      */
    264     static struct parser_state prev_ps;
    265 
    266     debug_println("");
    267     debug_printf("line %d: %s", line_no, lsym_name(lsym));
    268     debug_vis_range(" \"", token.s, token.e, "\"\n");
    269 
    270     debug_print_buf("label", &lab);
    271     debug_print_buf("code", &code);
    272     debug_print_buf("comment", &com);
    273 
    274     debug_println("    ps.prev_token = %s", lsym_name(ps.prev_token));
    275     debug_ps_bool(next_col_1);
    276     debug_ps_bool(curr_col_1);
    277     debug_ps_bool(next_unary);
    278     if (strcmp(ps.procname, prev_ps.procname) != 0)
    279 	debug_println("    ps.procname = '%s'", ps.procname);
    280     debug_ps_bool(want_blank);
    281     debug_ps_int(paren_level);
    282     debug_ps_int(p_l_follow);
    283     if (ps.paren_level != prev_ps.paren_level) {
    284 	debug_printf("    ps.paren_indents:");
    285 	for (int i = 0; i < ps.paren_level; i++)
    286 	    debug_printf(" %d", ps.paren_indents[i]);
    287 	debug_println("");
    288     }
    289     debug_ps_int(cast_mask);
    290     debug_ps_int(not_cast_mask);
    291 
    292     debug_ps_int(comment_delta);
    293     debug_ps_int(n_comment_delta);
    294     debug_ps_int(com_ind);
    295 
    296     debug_ps_bool(block_init);
    297     debug_ps_int(block_init_level);
    298     debug_ps_bool(init_or_struct);
    299 
    300     debug_ps_int(ind_level);
    301     debug_ps_int(ind_level_follow);
    302 
    303     debug_ps_int(decl_level);
    304     debug_ps_bool(decl_on_line);
    305     debug_ps_bool(in_decl);
    306     debug_ps_int(just_saw_decl);
    307     debug_ps_bool(in_parameter_declaration);
    308     debug_ps_bool(decl_indent_done);
    309 
    310     debug_ps_bool(in_stmt);
    311     debug_ps_bool(ind_stmt);
    312     debug_ps_bool(is_case_label);
    313 
    314     debug_ps_bool(search_stmt);
    315 
    316     prev_ps = ps;
    317 }
    318 #endif
    319 
    320 /* ARGSUSED */
    321 static lexer_symbol
    322 lexi_end(lexer_symbol lsym)
    323 {
    324 #ifdef debug
    325     debug_lexi(lsym);
    326 #endif
    327     return lsym;
    328 }
    329 
    330 static void
    331 lex_number(void)
    332 {
    333     for (unsigned char s = 'A'; s != 'f' && s != 'i' && s != 'u';) {
    334 	unsigned char ch = (unsigned char)inp_peek();
    335 	if (ch >= array_length(lex_number_row) || lex_number_row[ch] == 0)
    336 	    break;
    337 
    338 	unsigned char row = lex_number_row[ch];
    339 	if (lex_number_state[row][s - 'A'] == ' ') {
    340 	    /*-
    341 	     * lex_number_state[0][s - 'A'] now indicates the type:
    342 	     * f = floating, i = integer, u = unknown
    343 	     */
    344 	    return;
    345 	}
    346 
    347 	s = lex_number_state[row][s - 'A'];
    348 	token_add_char(inp_next());
    349     }
    350 }
    351 
    352 static void
    353 lex_word(void)
    354 {
    355     while (isalnum((unsigned char)inp_peek()) ||
    356 	    inp_peek() == '\\' ||
    357 	    inp_peek() == '_' || inp_peek() == '$') {
    358 
    359 	if (inp_peek() == '\\') {
    360 	    if (inp_lookahead(1) == '\n') {
    361 		inp_skip();
    362 		inp_skip();
    363 	    } else
    364 		break;
    365 	}
    366 
    367 	token_add_char(inp_next());
    368     }
    369 }
    370 
    371 static void
    372 lex_char_or_string(void)
    373 {
    374     for (char delim = token.e[-1];;) {
    375 	if (inp_peek() == '\n') {
    376 	    diag(1, "Unterminated literal");
    377 	    return;
    378 	}
    379 
    380 	token_add_char(inp_next());
    381 	if (token.e[-1] == delim)
    382 	    return;
    383 
    384 	if (token.e[-1] == '\\') {
    385 	    if (inp_peek() == '\n')
    386 		++line_no;
    387 	    token_add_char(inp_next());
    388 	}
    389     }
    390 }
    391 
    392 /* Guess whether the current token is a declared type. */
    393 static bool
    394 probably_typename(void)
    395 {
    396     if (ps.block_init || ps.in_stmt)
    397 	return false;
    398     if (inp_peek() == '*' && inp_lookahead(1) != '=')
    399 	goto maybe;
    400     if (isalpha((unsigned char)inp_peek()))
    401 	goto maybe;
    402     return false;
    403 maybe:
    404     return ps.prev_token == lsym_semicolon ||
    405 	ps.prev_token == lsym_lbrace ||
    406 	ps.prev_token == lsym_rbrace;
    407 }
    408 
    409 static int
    410 bsearch_typenames(const char *key)
    411 {
    412     const char **arr = typenames.items;
    413     int lo = 0;
    414     int hi = (int)typenames.len - 1;
    415 
    416     while (lo <= hi) {
    417 	int mid = (int)((unsigned)(lo + hi) >> 1);
    418 	int cmp = strcmp(arr[mid], key);
    419 	if (cmp < 0)
    420 	    lo = mid + 1;
    421 	else if (cmp > 0)
    422 	    hi = mid - 1;
    423 	else
    424 	    return mid;
    425     }
    426     return -(lo + 1);
    427 }
    428 
    429 static bool
    430 is_typename(void)
    431 {
    432     if (opt.auto_typedefs &&
    433 	token.e - token.s >= 2 && memcmp(token.e - 2, "_t", 2) == 0)
    434 	return true;
    435 
    436     return bsearch_typenames(token.s) >= 0;
    437 }
    438 
    439 static int
    440 cmp_keyword_by_name(const void *key, const void *elem)
    441 {
    442     return strcmp(key, ((const struct keyword *)elem)->name);
    443 }
    444 
    445 /* Read an alphanumeric token into 'token', or return lsym_eof. */
    446 static lexer_symbol
    447 lexi_alnum(void)
    448 {
    449     if (isdigit((unsigned char)inp_peek()) ||
    450 	    (inp_peek() == '.' && isdigit((unsigned char)inp_lookahead(1)))) {
    451 	lex_number();
    452     } else if (isalnum((unsigned char)inp_peek()) ||
    453 	    inp_peek() == '_' || inp_peek() == '$') {
    454 	lex_word();
    455     } else
    456 	return lsym_eof;	/* just as a placeholder */
    457 
    458     *token.e = '\0';
    459 
    460     if (token.s[0] == 'L' && token.s[1] == '\0' &&
    461 	    (inp_peek() == '"' || inp_peek() == '\''))
    462 	return lsym_string_prefix;
    463 
    464     while (ch_isblank(inp_peek()))
    465 	inp_skip();
    466 
    467     if (ps.prev_token == lsym_tag && ps.p_l_follow == 0) {
    468 	ps.next_unary = true;
    469 	return lsym_type_outside_parentheses;
    470     }
    471 
    472     /* Operator after identifier is binary unless last token was 'struct'. */
    473     ps.next_unary = ps.prev_token == lsym_tag;
    474 
    475     const struct keyword *kw = bsearch(token.s, keywords,
    476 	array_length(keywords), sizeof(keywords[0]), cmp_keyword_by_name);
    477     bool is_type = false;
    478     if (kw == NULL) {
    479 	if (is_typename()) {
    480 	    is_type = true;
    481 	    ps.next_unary = true;
    482 	    goto found_typename;
    483 	}
    484 
    485     } else {			/* we have a keyword */
    486 	is_type = kw->lsym == lsym_type;
    487 	ps.next_unary = true;
    488 	if (kw->lsym != lsym_tag && kw->lsym != lsym_type)
    489 	    return kw->lsym;
    490 
    491 found_typename:
    492 	if (ps.p_l_follow > 0) {
    493 	    /* inside parentheses: cast, param list, offsetof or sizeof */
    494 	    ps.cast_mask |= (1 << ps.p_l_follow) & ~ps.not_cast_mask;
    495 	}
    496 	if (ps.prev_token != lsym_period && ps.prev_token != lsym_unary_op) {
    497 	    if (kw != NULL && kw->lsym == lsym_tag)
    498 		return lsym_tag;
    499 	    if (ps.p_l_follow == 0)
    500 		return lsym_type_outside_parentheses;
    501 	}
    502     }
    503 
    504     if (inp_peek() == '(' && ps.tos <= 1 && ps.ind_level == 0 &&
    505 	!ps.in_parameter_declaration && !ps.block_init) {
    506 
    507 	for (const char *p = inp_p(), *e = inp_line_end(); p < e;)
    508 	    if (*p++ == ')' && (*p == ';' || *p == ','))
    509 		goto no_function_definition;
    510 
    511 	strncpy(ps.procname, token.s, sizeof ps.procname - 1);
    512 	if (ps.in_decl)
    513 	    ps.in_parameter_declaration = true;
    514 	return lsym_funcname;
    515 no_function_definition:;
    516 
    517     } else if (ps.p_l_follow == 0 && probably_typename()) {
    518 	ps.next_unary = true;
    519 	return lsym_type_outside_parentheses;
    520     }
    521 
    522     return is_type ? lsym_type_in_parentheses : lsym_word;
    523 }
    524 
    525 /* Reads the next token, placing it in the global variable "token". */
    526 lexer_symbol
    527 lexi(void)
    528 {
    529     token.e = token.s;
    530     ps.curr_col_1 = ps.next_col_1;
    531     ps.next_col_1 = false;
    532 
    533     while (ch_isblank(inp_peek())) {
    534 	ps.curr_col_1 = false;
    535 	inp_skip();
    536     }
    537 
    538     lexer_symbol alnum_lsym = lexi_alnum();
    539     if (alnum_lsym != lsym_eof)
    540 	return lexi_end(alnum_lsym);
    541 
    542     /* Scan a non-alphanumeric token */
    543 
    544     check_size_token(3);	/* for things like "<<=" */
    545     *token.e++ = inp_next();
    546     *token.e = '\0';
    547 
    548     lexer_symbol lsym;
    549     bool unary_delim = false;	/* whether the current token forces a
    550 				 * following operator to be unary */
    551 
    552     switch (token.e[-1]) {
    553     case '\n':
    554 	unary_delim = ps.next_unary;
    555 	ps.next_col_1 = true;
    556 	/* if data has been exhausted, the newline is a dummy. */
    557 	lsym = had_eof ? lsym_eof : lsym_newline;
    558 	break;
    559 
    560     case '\'':
    561     case '"':
    562 	lex_char_or_string();
    563 	lsym = lsym_word;
    564 	break;
    565 
    566     case '(':
    567     case '[':
    568 	unary_delim = true;
    569 	lsym = lsym_lparen_or_lbracket;
    570 	break;
    571 
    572     case ')':
    573     case ']':
    574 	lsym = lsym_rparen_or_rbracket;
    575 	break;
    576 
    577     case '#':
    578 	unary_delim = ps.next_unary;
    579 	lsym = lsym_preprocessing;
    580 	break;
    581 
    582     case '?':
    583 	unary_delim = true;
    584 	lsym = lsym_question;
    585 	break;
    586 
    587     case ':':
    588 	lsym = lsym_colon;
    589 	unary_delim = true;
    590 	break;
    591 
    592     case ';':
    593 	unary_delim = true;
    594 	lsym = lsym_semicolon;
    595 	break;
    596 
    597     case '{':
    598 	unary_delim = true;
    599 	lsym = lsym_lbrace;
    600 	break;
    601 
    602     case '}':
    603 	unary_delim = true;
    604 	lsym = lsym_rbrace;
    605 	break;
    606 
    607     case '\f':
    608 	unary_delim = ps.next_unary;
    609 	ps.next_col_1 = true;
    610 	lsym = lsym_form_feed;
    611 	break;
    612 
    613     case ',':
    614 	unary_delim = true;
    615 	lsym = lsym_comma;
    616 	break;
    617 
    618     case '.':
    619 	unary_delim = false;
    620 	lsym = lsym_period;
    621 	break;
    622 
    623     case '-':
    624     case '+':
    625 	lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op;
    626 	unary_delim = true;
    627 
    628 	if (inp_peek() == token.e[-1]) {	/* ++, -- */
    629 	    *token.e++ = inp_next();
    630 	    if (ps.prev_token == lsym_word ||
    631 		    ps.prev_token == lsym_rparen_or_rbracket) {
    632 		lsym = ps.next_unary ? lsym_unary_op : lsym_postfix_op;
    633 		unary_delim = false;
    634 	    }
    635 
    636 	} else if (inp_peek() == '=') {	/* += */
    637 	    *token.e++ = inp_next();
    638 
    639 	} else if (inp_peek() == '>') {	/* -> */
    640 	    *token.e++ = inp_next();
    641 	    unary_delim = false;
    642 	    lsym = lsym_unary_op;
    643 	    ps.want_blank = false;
    644 	}
    645 	break;
    646 
    647     case '=':
    648 	if (ps.init_or_struct)
    649 	    ps.block_init = true;
    650 	if (inp_peek() == '=') {	/* == */
    651 	    *token.e++ = inp_next();
    652 	    *token.e = '\0';
    653 	}
    654 	lsym = lsym_binary_op;
    655 	unary_delim = true;
    656 	break;
    657 
    658     case '>':
    659     case '<':
    660     case '!':			/* ops like <, <<, <=, !=, etc */
    661 	if (inp_peek() == '>' || inp_peek() == '<' || inp_peek() == '=')
    662 	    *token.e++ = inp_next();
    663 	if (inp_peek() == '=')
    664 	    *token.e++ = inp_next();
    665 	lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op;
    666 	unary_delim = true;
    667 	break;
    668 
    669     case '*':
    670 	unary_delim = true;
    671 	if (!ps.next_unary) {
    672 	    if (inp_peek() == '=')
    673 		*token.e++ = inp_next();
    674 	    lsym = lsym_binary_op;
    675 	    break;
    676 	}
    677 
    678 	while (inp_peek() == '*' || isspace((unsigned char)inp_peek())) {
    679 	    if (inp_peek() == '*')
    680 		token_add_char('*');
    681 	    inp_skip();
    682 	}
    683 
    684 	if (ps.in_decl) {
    685 	    const char *tp = inp_p();
    686 
    687 	    while (isalpha((unsigned char)*tp) ||
    688 		    isspace((unsigned char)*tp)) {
    689 		if (++tp >= inp_line_end()) {
    690 		    const char *p_before = inp_p();
    691 		    inp_read_line();
    692 		    if (inp_p() != p_before)
    693 			abort();
    694 		}
    695 	    }
    696 	    if (*tp == '(')
    697 		ps.procname[0] = ' ';
    698 	}
    699 
    700 	lsym = lsym_unary_op;
    701 	break;
    702 
    703     default:
    704 	if (token.e[-1] == '/' && (inp_peek() == '*' || inp_peek() == '/')) {
    705 	    *token.e++ = inp_next();
    706 	    lsym = lsym_comment;
    707 	    unary_delim = ps.next_unary;
    708 	    break;
    709 	}
    710 
    711 	/* handle '||', '&&', etc., and also things as in 'int *****i' */
    712 	while (token.e[-1] == inp_peek() || inp_peek() == '=')
    713 	    token_add_char(inp_next());
    714 
    715 	lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op;
    716 	unary_delim = true;
    717     }
    718 
    719     ps.next_unary = unary_delim;
    720 
    721     check_size_token(1);
    722     *token.e = '\0';
    723 
    724     return lexi_end(lsym);
    725 }
    726 
    727 void
    728 register_typename(const char *name)
    729 {
    730     if (typenames.len >= typenames.cap) {
    731 	typenames.cap = 16 + 2 * typenames.cap;
    732 	typenames.items = xrealloc(typenames.items,
    733 	    sizeof(typenames.items[0]) * typenames.cap);
    734     }
    735 
    736     int pos = bsearch_typenames(name);
    737     if (pos >= 0)
    738 	return;			/* already in the list */
    739 
    740     pos = -(pos + 1);
    741     memmove(typenames.items + pos + 1, typenames.items + pos,
    742 	sizeof(typenames.items[0]) * (typenames.len++ - (unsigned)pos));
    743     typenames.items[pos] = xstrdup(name);
    744 }
    745