Annotation of src/usr.bin/fmt/fmt.c, Revision 1.34
1.34 ! schwarze 1: /* $OpenBSD: fmt.c,v 1.33 2015/10/09 01:37:07 deraadt Exp $ */
1.1 deraadt 2:
1.10 millert 3: /* Sensible version of fmt
4: *
5: * Syntax: fmt [ options ] [ goal [ max ] ] [ filename ... ]
6: *
7: * Since the documentation for the original fmt is so poor, here
8: * is an accurate description of what this one does. It's usually
9: * the same. The *mechanism* used may differ from that suggested
10: * here. Note that we are *not* entirely compatible with fmt,
11: * because fmt gets so many things wrong.
12: *
13: * 1. Tabs are expanded, assuming 8-space tab stops.
14: * If the `-t <n>' option is given, we assume <n>-space
15: * tab stops instead.
16: * Trailing blanks are removed from all lines.
17: * x\b == nothing, for any x other than \b.
18: * Other control characters are simply stripped. This
19: * includes \r.
20: * 2. Each line is split into leading whitespace and
21: * everything else. Maximal consecutive sequences of
22: * lines with the same leading whitespace are considered
23: * to form paragraphs, except that a blank line is always
24: * a paragraph to itself.
25: * If the `-p' option is given then the first line of a
26: * paragraph is permitted to have indentation different
27: * from that of the other lines.
28: * If the `-m' option is given then a line that looks
29: * like a mail message header, if it is not immediately
30: * preceded by a non-blank non-message-header line, is
31: * taken to start a new paragraph, which also contains
32: * any subsequent lines with non-empty leading whitespace.
1.19 millert 33: * Unless the `-n' option is given, lines beginning with
34: * a . (dot) are not formatted.
1.10 millert 35: * 3. The "everything else" is split into words; a word
36: * includes its trailing whitespace, and a word at the
37: * end of a line is deemed to be followed by a single
38: * space, or two spaces if it ends with a sentence-end
39: * character. (See the `-d' option for how to change that.)
40: * If the `-s' option has been given, then a word's trailing
41: * whitespace is replaced by what it would have had if it
42: * had occurred at end of line.
43: * 4. Each paragraph is sent to standard output as follows.
44: * We output the leading whitespace, and then enough words
45: * to make the line length as near as possible to the goal
46: * without exceeding the maximum. (If a single word would
47: * exceed the maximum, we output that anyway.) Of course
48: * the trailing whitespace of the last word is ignored.
49: * We then emit a newline and start again if there are any
50: * words left.
51: * Note that for a blank line this translates as "We emit
52: * a newline".
53: * If the `-l <n>' option is given, then leading whitespace
54: * is modified slightly: <n> spaces are replaced by a tab.
55: * Indented paragraphs (see above under `-p') make matters
56: * more complicated than this suggests. Actually every paragraph
57: * has two `leading whitespace' values; the value for the first
58: * line, and the value for the most recent line. (While processing
59: * the first line, the two are equal. When `-p' has not been
60: * given, they are always equal.) The leading whitespace
61: * actually output is that of the first line (for the first
62: * line of *output*) or that of the most recent line (for
63: * all other lines of output).
64: * When `-m' has been given, message header paragraphs are
65: * taken as having first-leading-whitespace empty and
66: * subsequent-leading-whitespace two spaces.
67: *
68: * Multiple input files are formatted one at a time, so that a file
69: * never ends in the middle of a line.
70: *
71: * There's an alternative mode of operation, invoked by giving
72: * the `-c' option. In that case we just center every line,
73: * and most of the other options are ignored. This should
74: * really be in a separate program, but we must stay compatible
75: * with old `fmt'.
76: *
77: * QUERY: Should `-m' also try to do the right thing with quoted text?
78: * QUERY: `-b' to treat backslashed whitespace as old `fmt' does?
79: * QUERY: Option meaning `never join lines'?
80: * QUERY: Option meaning `split in mid-word to avoid overlong lines'?
81: * (Those last two might not be useful, since we have `fold'.)
82: *
83: * Differences from old `fmt':
84: *
85: * - We have many more options. Options that aren't understood
86: * generate a lengthy usage message, rather than being
87: * treated as filenames.
88: * - Even with `-m', our handling of message headers is
89: * significantly different. (And much better.)
90: * - We don't treat `\ ' as non-word-breaking.
91: * - Downward changes of indentation start new paragraphs
92: * for us, as well as upward. (I think old `fmt' behaves
93: * in the way it does in order to allow indented paragraphs,
94: * but this is a broken way of making indented paragraphs
95: * behave right.)
96: * - Given the choice of going over or under |goal_length|
97: * by the same amount, we go over; old `fmt' goes under.
98: * - We treat `?' as ending a sentence, and not `:'. Old `fmt'
99: * does the reverse.
100: * - We return approved return codes. Old `fmt' returns
101: * 1 for some errors, and *the number of unopenable files*
102: * when that was all that went wrong.
103: * - We have fewer crashes and more helpful error messages.
104: * - We don't turn spaces into tabs at starts of lines unless
105: * specifically requested.
106: * - New `fmt' is somewhat smaller and slightly faster than
107: * old `fmt'.
108: *
109: * Bugs:
110: *
111: * None known. There probably are some, though.
112: *
113: * Portability:
114: *
115: * I believe this code to be pretty portable. It does require
116: * that you have `getopt'. If you need to include "getopt.h"
117: * for this (e.g., if your system didn't come with `getopt'
118: * and you installed it yourself) then you should arrange for
119: * NEED_getopt_h to be #defined.
120: *
121: * Everything here should work OK even on nasty 16-bit
122: * machines and nice 64-bit ones. However, it's only really
123: * been tested on my FreeBSD machine. Your mileage may vary.
124: */
125:
126: /* Copyright (c) 1997 Gareth McCaughan. All rights reserved.
127: *
128: * Redistribution and use of this code, in source or binary forms,
129: * with or without modification, are permitted subject to the following
130: * conditions:
131: *
132: * - Redistribution of source code must retain the above copyright
1.1 deraadt 133: * notice, this list of conditions and the following disclaimer.
1.10 millert 134: *
135: * - If you distribute modified source code it must also include
136: * a notice saying that it has been modified, and giving a brief
137: * description of what changes have been made.
138: *
139: * Disclaimer: I am not responsible for the results of using this code.
140: * If it formats your hard disc, sends obscene messages to
141: * your boss and kills your children then that's your problem
142: * not mine. I give absolutely no warranty of any sort as to
143: * what the program will do, and absolutely refuse to be held
144: * liable for any consequences of your using it.
145: * Thank you. Have a nice day.
146: */
147:
148: /* RCS change log:
149: * Revision 1.5 1998/03/02 18:02:21 gjm11
150: * Minor changes for portability.
151: *
152: * Revision 1.4 1997/10/01 11:51:28 gjm11
153: * Repair broken indented-paragraph handling.
154: * Add mail message header stuff.
155: * Improve comments and layout.
156: * Make usable with non-BSD systems.
157: * Add revision display to usage message.
158: *
159: * Revision 1.3 1997/09/30 16:24:47 gjm11
160: * Add copyright notice, rcsid string and log message.
161: *
162: * Revision 1.2 1997/09/30 16:13:39 gjm11
163: * Add options: -d <chars>, -l <width>, -p, -s, -t <width>, -h .
164: * Parse options with `getopt'. Clean up code generally.
165: * Make comments more accurate.
166: *
167: * Revision 1.1 1997/09/30 11:29:57 gjm11
168: * Initial revision
1.1 deraadt 169: */
170:
1.4 millert 171: #include <ctype.h>
1.18 millert 172: #include <err.h>
173: #include <locale.h>
1.1 deraadt 174: #include <stdio.h>
175: #include <stdlib.h>
176: #include <string.h>
1.18 millert 177: #include <sysexits.h>
178: #include <unistd.h>
1.34 ! schwarze 179: #include <wchar.h>
! 180: #include <wctype.h>
1.10 millert 181:
182: /* Something that, we hope, will never be a genuine line length,
183: * indentation etc.
1.1 deraadt 184: */
1.10 millert 185: #define SILLY ((size_t)-1)
1.1 deraadt 186:
1.10 millert 187: /* I used to use |strtoul| for this, but (1) not all systems have it
188: * and (2) it's probably better to use |strtol| to detect negative
189: * numbers better.
190: * If |fussyp==0| then we don't complain about non-numbers
191: * (returning 0 instead), but we do complain about bad numbers.
192: */
1.18 millert 193: static size_t
1.21 tedu 194: get_positive(const char *s, const char *err_mess, int fussyP)
195: {
196: char *t;
197: long result = strtol(s, &t, 0);
198:
199: if (*t) {
200: if (fussyP)
201: goto Lose;
202: else
203: return 0;
204: }
205: if (result <= 0) {
206: Lose:
207: errx(EX_USAGE, "%s", err_mess);
208: }
209:
210: return (size_t) result;
1.1 deraadt 211: }
212:
1.10 millert 213: /* Global variables */
214:
1.21 tedu 215: static int centerP = 0; /* Try to center lines? */
216: static size_t goal_length = 0; /* Target length for output lines */
217: static size_t max_length = 0; /* Maximum length for output lines */
218: static int coalesce_spaces_P = 0; /* Coalesce multiple whitespace -> ' ' ? */
219: static int allow_indented_paragraphs = 0; /* Can first line have diff. ind.? */
220: static int tab_width = 8; /* Number of spaces per tab stop */
221: static size_t output_tab_width = 0; /* Ditto, when squashing leading spaces */
222: static const char *sentence_enders = ".?!"; /* Double-space after these */
223: static int grok_mail_headers = 0; /* treat embedded mail headers magically? */
224: static int format_troff = 0; /* Format troff? */
225:
1.28 schwarze 226: static int n_errors = 0; /* Number of failed files. */
1.21 tedu 227: static size_t x; /* Horizontal position in output line */
228: static size_t x0; /* Ditto, ignoring leading whitespace */
229: static size_t pending_spaces; /* Spaces to add before next word */
230: static int output_in_paragraph = 0; /* Any of current para written out yet? */
1.10 millert 231:
232: /* Prototypes */
233:
1.21 tedu 234: static void process_named_file(const char *);
235: static void process_stream(FILE *, const char *);
1.34 ! schwarze 236: static size_t indent_length(const char *);
1.30 deraadt 237: static int might_be_header(const char *);
1.34 ! schwarze 238: static void new_paragraph(size_t);
! 239: static void output_word(size_t, size_t, const char *, int, int, int);
1.21 tedu 240: static void output_indent(size_t);
241: static void center_stream(FILE *, const char *);
1.34 ! schwarze 242: static char *get_line(FILE *);
1.21 tedu 243: static void *xrealloc(void *, size_t);
244: void usage(void);
1.10 millert 245:
1.29 lum 246: #define ERRS(x) (x >= 127 ? 127 : ++x)
1.10 millert 247:
248: /* Here is perhaps the right place to mention that this code is
249: * all in top-down order. Hence, |main| comes first.
1.1 deraadt 250: */
1.4 millert 251: int
1.21 tedu 252: main(int argc, char *argv[])
253: {
254: int ch; /* used for |getopt| processing */
255:
256: (void)setlocale(LC_CTYPE, "");
257:
1.33 deraadt 258: if (pledge("stdio rpath", NULL) == -1)
259: err(1, "pledge");
1.31 deraadt 260:
1.21 tedu 261: /* 1. Grok parameters. */
262: while ((ch = getopt(argc, argv, "0123456789cd:hl:mnpst:w:")) != -1) {
263: switch (ch) {
264: case 'c':
265: centerP = 1;
266: break;
267: case 'd':
268: sentence_enders = optarg;
269: break;
270: case 'l':
271: output_tab_width
272: = get_positive(optarg, "output tab width must be positive", 1);
273: break;
274: case 'm':
275: grok_mail_headers = 1;
276: break;
277: case 'n':
278: format_troff = 1;
279: break;
280: case 'p':
281: allow_indented_paragraphs = 1;
282: break;
283: case 's':
284: coalesce_spaces_P = 1;
285: break;
286: case 't':
287: tab_width = get_positive(optarg, "tab width must be positive", 1);
288: break;
289: case 'w':
290: goal_length = get_positive(optarg, "width must be positive", 1);
291: max_length = goal_length;
292: break;
293: case '0': case '1': case '2': case '3': case '4': case '5':
294: case '6': case '7': case '8': case '9':
295: /* XXX this is not a stylistically approved use of getopt() */
296: if (goal_length == 0) {
297: char *p;
298:
299: p = argv[optind - 1];
300: if (p[0] == '-' && p[1] == ch && !p[2])
301: goal_length = get_positive(++p, "width must be nonzero", 1);
302: else
303: goal_length = get_positive(argv[optind]+1,
304: "width must be nonzero", 1);
305: max_length = goal_length;
306: }
307: break;
308: case 'h':
309: default:
310: usage();
311: /* NOT REACHED */
312: }
313: }
314:
315: argc -= optind;
316: argv += optind;
317:
318: /* [ goal [ maximum ] ] */
319: if (argc > 0 && goal_length == 0 &&
320: (goal_length = get_positive(*argv,"goal length must be positive", 0)) != 0) {
321: --argc;
322: ++argv;
323: if (argc > 0 && (max_length = get_positive(*argv,"max length must be positive", 0)) != 0) {
324: --argc;
325: ++argv;
326: if (max_length < goal_length)
327: errx(EX_USAGE, "max length must be >= goal length");
328: }
329: }
330:
331: if (goal_length == 0)
332: goal_length = 65;
333: if (max_length == 0)
334: max_length = goal_length+10;
335:
336: /* 2. Process files. */
337:
338: if (argc > 0) {
339: while (argc-- > 0)
340: process_named_file(*argv++);
341: } else {
1.33 deraadt 342: if (pledge("stdio", NULL) == -1)
343: err(1, "pledge");
1.21 tedu 344: process_stream(stdin, "standard input");
345: }
1.10 millert 346:
1.21 tedu 347: /* We're done. */
1.29 lum 348: return n_errors;
1.10 millert 349:
350: }
351:
352: /* Process a single file, given its name.
353: */
354: static void
1.21 tedu 355: process_named_file(const char *name)
356: {
357: FILE *f;
358:
359: if ((f = fopen(name, "r")) == NULL) {
1.23 cloder 360: warn("%s", name);
1.29 lum 361: ERRS(n_errors);
1.21 tedu 362: } else {
363: process_stream(f, name);
364: fclose(f);
365: }
1.10 millert 366: }
367:
368: /* Types of mail header continuation lines:
369: */
370: typedef enum {
1.21 tedu 371: hdr_ParagraphStart = -1,
372: hdr_NonHeader = 0,
373: hdr_Header = 1,
374: hdr_Continuation = 2
1.10 millert 375: } HdrType;
376:
377: /* Process a stream. This is where the real work happens,
378: * except that centering is handled separately.
379: */
380: static void
1.21 tedu 381: process_stream(FILE *stream, const char *name)
382: {
1.34 ! schwarze 383: const char *wordp, *cp;
! 384: wchar_t wc;
1.21 tedu 385: size_t np;
386: size_t last_indent = SILLY; /* how many spaces in last indent? */
387: size_t para_line_number = 0; /* how many lines already read in this para? */
388: size_t first_indent = SILLY; /* indentation of line 0 of paragraph */
1.34 ! schwarze 389: int wcl; /* number of bytes in wide character */
! 390: int wcw; /* display width of wide character */
! 391: int word_length; /* number of bytes in word */
! 392: int word_width; /* display width of word */
! 393: int space_width; /* display width of space after word */
! 394: int line_width; /* display width of line */
1.21 tedu 395: HdrType prev_header_type = hdr_ParagraphStart;
396: HdrType header_type;
397:
1.10 millert 398: /* ^-- header_type of previous line; -1 at para start */
1.30 deraadt 399: const char *line;
1.10 millert 400:
1.21 tedu 401: if (centerP) {
402: center_stream(stream, name);
403: return;
404: }
405:
1.34 ! schwarze 406: while ((line = get_line(stream)) != NULL) {
! 407: np = indent_length(line);
1.21 tedu 408: header_type = hdr_NonHeader;
409: if (grok_mail_headers && prev_header_type != hdr_NonHeader) {
410: if (np == 0 && might_be_header(line))
411: header_type = hdr_Header;
412: else if (np > 0 && prev_header_type>hdr_NonHeader)
413: header_type = hdr_Continuation;
414: }
415:
416: /* We need a new paragraph if and only if:
417: * this line is blank,
418: * OR it's a troff request,
419: * OR it's a mail header,
420: * OR it's not a mail header AND the last line was one,
421: * OR the indentation has changed
422: * AND the line isn't a mail header continuation line
423: * AND this isn't the second line of an indented paragraph.
424: */
1.34 ! schwarze 425: if (*line == '\0' || (*line == '.' && !format_troff) ||
1.21 tedu 426: header_type == hdr_Header ||
427: (header_type == hdr_NonHeader && prev_header_type > hdr_NonHeader) ||
428: (np != last_indent && header_type != hdr_Continuation &&
429: (!allow_indented_paragraphs || para_line_number != 1)) ) {
1.34 ! schwarze 430: new_paragraph(np);
1.21 tedu 431: para_line_number = 0;
432: first_indent = np;
433: last_indent = np;
434:
435: /* nroff compatibility */
1.34 ! schwarze 436: if (*line == '.' && !format_troff) {
! 437: puts(line);
1.21 tedu 438: continue;
439: }
440: if (header_type == hdr_Header)
441: last_indent = 2; /* for cont. lines */
1.34 ! schwarze 442: if (*line == '\0') {
1.21 tedu 443: putchar('\n');
444: prev_header_type = hdr_ParagraphStart;
445: continue;
446: } else {
447: /* If this is an indented paragraph other than a mail header
448: * continuation, set |last_indent|.
449: */
450: if (np != last_indent && header_type != hdr_Continuation)
451: last_indent = np;
452: }
453: prev_header_type = header_type;
454: }
455:
1.34 ! schwarze 456: line_width = np;
! 457: for (wordp = line; *wordp != '\0'; wordp = cp) {
! 458: word_length = 0;
! 459: word_width = space_width = 0;
! 460: for (cp = wordp; *cp != '\0'; cp += wcl) {
! 461: wcl = mbtowc(&wc, cp, MB_CUR_MAX);
! 462: if (wcl == -1) {
! 463: (void)mbtowc(NULL, NULL, MB_CUR_MAX);
! 464: wc = L'?';
! 465: wcl = 1;
! 466: wcw = 1;
! 467: } else if (wc == L'\t')
! 468: wcw = (line_width / tab_width + 1) *
! 469: tab_width - line_width;
! 470: else if ((wcw = wcwidth(wc)) == -1)
! 471: wcw = 1;
! 472: if (iswblank(wc)) {
! 473: /* Skip whitespace at start of line. */
! 474: if (word_length == 0) {
! 475: wordp += wcl;
! 476: continue;
! 477: }
! 478: /* Count whitespace after word. */
! 479: space_width += wcw;
! 480: } else {
! 481: /* Detect end of word. */
! 482: if (space_width > 0)
! 483: break;
! 484: /* Measure word. */
! 485: word_length += wcl;
! 486: word_width += wcw;
! 487: }
! 488: line_width += wcw;
! 489: }
! 490:
1.21 tedu 491: /* Send the word to the output machinery. */
1.34 ! schwarze 492: output_word(first_indent, last_indent, wordp,
! 493: word_length, word_width, space_width);
1.21 tedu 494: }
495: ++para_line_number;
496: }
497:
1.34 ! schwarze 498: new_paragraph(0);
1.21 tedu 499: if (ferror(stream)) {
1.23 cloder 500: warn("%s", name);
1.29 lum 501: ERRS(n_errors);
1.21 tedu 502: }
1.10 millert 503: }
504:
505: /* How long is the indent on this line?
506: */
507: static size_t
1.34 ! schwarze 508: indent_length(const char *line)
1.21 tedu 509: {
510: size_t n = 0;
511:
1.34 ! schwarze 512: for (;;) {
! 513: switch(*line++) {
! 514: case ' ':
! 515: ++n;
! 516: continue;
! 517: case '\t':
! 518: n = (n / tab_width + 1) * tab_width;
! 519: continue;
! 520: default:
! 521: break;
! 522: }
! 523: break;
! 524: }
1.21 tedu 525: return n;
1.10 millert 526: }
527:
528: /* Might this line be a mail header?
529: * We deem a line to be a possible header if it matches the
530: * Perl regexp /^[A-Z][-A-Za-z0-9]*:\s/. This is *not* the same
531: * as in RFC whatever-number-it-is; we want to be gratuitously
532: * conservative to avoid mangling ordinary civilised text.
533: */
534: static int
1.30 deraadt 535: might_be_header(const char *line)
1.21 tedu 536: {
537:
1.30 deraadt 538: if (!isupper((unsigned char)*line++))
1.21 tedu 539: return 0;
1.30 deraadt 540: while (isalnum((unsigned char)*line) || *line == '-')
1.21 tedu 541: ++line;
1.30 deraadt 542: return (*line == ':' && isspace((unsigned char)line[1]));
1.10 millert 543: }
544:
545: /* Begin a new paragraph with an indent of |indent| spaces.
546: */
547: static void
1.34 ! schwarze 548: new_paragraph(size_t indent)
1.21 tedu 549: {
550:
1.34 ! schwarze 551: if (x0 > 0)
1.21 tedu 552: putchar('\n');
553: x = indent;
554: x0 = 0;
555: pending_spaces = 0;
556: output_in_paragraph = 0;
1.10 millert 557: }
558:
559: /* Output spaces or tabs for leading indentation.
560: */
561: static void
1.21 tedu 562: output_indent(size_t n_spaces)
563: {
564:
1.34 ! schwarze 565: if (n_spaces == 0)
! 566: return;
1.21 tedu 567: if (output_tab_width) {
568: while (n_spaces >= output_tab_width) {
569: putchar('\t');
570: n_spaces -= output_tab_width;
571: }
572: }
573: while (n_spaces-- > 0)
574: putchar(' ');
1.10 millert 575: }
576:
1.34 ! schwarze 577: /* Output a single word.
1.10 millert 578: * indent0 and indent1 are the indents to use on the first and subsequent
579: * lines of a paragraph. They'll often be the same, of course.
580: */
581: static void
1.34 ! schwarze 582: output_word(size_t indent0, size_t indent1, const char *word,
! 583: int length, int width, int spaces)
1.21 tedu 584: {
1.34 ! schwarze 585: size_t new_x = x + pending_spaces + width;
1.21 tedu 586:
587: /* If either |spaces==0| (at end of line) or |coalesce_spaces_P|
588: * (squashing internal whitespace), then add just one space;
589: * except that if the last character was a sentence-ender we
590: * actually add two spaces.
591: */
592: if (coalesce_spaces_P || spaces == 0)
593: spaces = strchr(sentence_enders, word[length-1]) ? 2 : 1;
594:
1.34 ! schwarze 595: if (x0 == 0)
! 596: output_indent(output_in_paragraph ? indent1 : indent0);
! 597: else if (new_x > max_length || x >= goal_length ||
! 598: (new_x > goal_length && new_x-goal_length > goal_length-x)) {
! 599: putchar('\n');
! 600: output_indent(indent1);
! 601: x0 = 0;
! 602: x = indent1;
! 603: } else {
1.21 tedu 604: x0 += pending_spaces;
605: x += pending_spaces;
1.34 ! schwarze 606: while (pending_spaces--)
! 607: putchar(' ');
1.21 tedu 608: }
1.34 ! schwarze 609: x0 += width;
! 610: x += width;
! 611: while(length--)
! 612: putchar(*word++);
! 613: pending_spaces = spaces;
! 614: output_in_paragraph = 1;
1.10 millert 615: }
616:
617: /* Process a stream, but just center its lines rather than trying to
618: * format them neatly.
619: */
620: static void
1.21 tedu 621: center_stream(FILE *stream, const char *name)
622: {
623: char *line;
624: size_t l;
625:
1.34 ! schwarze 626: while ((line = get_line(stream)) != NULL) {
! 627: while (isspace((unsigned char)*line))
1.21 tedu 628: ++line;
1.34 ! schwarze 629: l = strlen(line);
1.21 tedu 630: while (l < goal_length) {
631: putchar(' ');
632: l += 2;
633: }
1.34 ! schwarze 634: puts(line);
1.21 tedu 635: }
636:
637: if (ferror(stream)) {
1.23 cloder 638: warn("%s", name);
1.29 lum 639: ERRS(n_errors);
1.21 tedu 640: }
1.10 millert 641: }
642:
1.34 ! schwarze 643: /* Get a single line from a stream. Strip control
1.10 millert 644: * characters and trailing whitespace, and handle backspaces.
1.34 ! schwarze 645: * Return the address of the buffer containing the line.
1.10 millert 646: * This can cope with arbitrarily long lines, and with lines
647: * without terminating \n.
648: * If there are no characters left or an error happens, we
1.34 ! schwarze 649: * return NULL.
1.10 millert 650: */
651: static char *
1.34 ! schwarze 652: get_line(FILE *stream)
1.21 tedu 653: {
654: int ch;
655: int troff = 0;
656: static char *buf = NULL;
657: static size_t length = 0;
658: size_t len = 0;
659:
660: if (buf == NULL) {
661: length = 100;
1.34 ! schwarze 662: buf = xrealloc(NULL, length);
1.21 tedu 663: }
664:
665: while ((ch = getc(stream)) != '\n' && ch != EOF) {
1.34 ! schwarze 666: if ((len == 0) && (ch == '.' && !format_troff))
1.21 tedu 667: troff = 1;
1.34 ! schwarze 668: if (troff || ch == '\t' || !iscntrl(ch)) {
! 669: if (len >= length) {
1.21 tedu 670: length *= 2;
671: buf = xrealloc(buf, length);
672: }
673: buf[len++] = ch;
674: } else if (ch == '\b') {
675: if (len)
676: --len;
677: }
678: }
1.34 ! schwarze 679: while (len > 0 && isspace((unsigned char)buf[len-1]))
! 680: --len;
! 681: buf[len] = '\0';
! 682: return (len > 0 || ch != EOF) ? buf : NULL;
1.10 millert 683: }
684:
685: /* (Re)allocate some memory, exiting with an error if we can't.
686: */
687: static void *
1.21 tedu 688: xrealloc(void *ptr, size_t nbytes)
689: {
690: void *p;
691:
692: p = realloc(ptr, nbytes);
693: if (p == NULL)
694: errx(EX_OSERR, "out of memory");
695: return p;
696: }
697:
698: void
699: usage(void)
700: {
1.22 mickey 701: extern char *__progname;
1.21 tedu 702:
703: fprintf(stderr,
1.25 jmc 704: "usage: %s [-cmnps] [-d chars] [-l number] [-t number]\n"
705: "\t[goal [maximum] | -width | -w width] [file ...]\n",
706: __progname);
1.22 mickey 707: exit (1);
1.1 deraadt 708: }