Annotation of src/usr.bin/fmt/fmt.c, Revision 1.37
1.37 ! schwarze 1: /* $OpenBSD: fmt.c,v 1.36 2016/01/07 18:02:43 schwarze 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 <unistd.h>
1.34 schwarze 178: #include <wchar.h>
179: #include <wctype.h>
1.10 millert 180:
181: /* Something that, we hope, will never be a genuine line length,
182: * indentation etc.
1.1 deraadt 183: */
1.10 millert 184: #define SILLY ((size_t)-1)
1.1 deraadt 185:
1.10 millert 186: /* I used to use |strtoul| for this, but (1) not all systems have it
187: * and (2) it's probably better to use |strtol| to detect negative
188: * numbers better.
189: * If |fussyp==0| then we don't complain about non-numbers
190: * (returning 0 instead), but we do complain about bad numbers.
191: */
1.18 millert 192: static size_t
1.21 tedu 193: get_positive(const char *s, const char *err_mess, int fussyP)
194: {
195: char *t;
196: long result = strtol(s, &t, 0);
197:
198: if (*t) {
199: if (fussyP)
200: goto Lose;
201: else
202: return 0;
203: }
204: if (result <= 0) {
205: Lose:
1.35 millert 206: errx(1, "%s", err_mess);
1.21 tedu 207: }
208:
209: return (size_t) result;
1.1 deraadt 210: }
211:
1.10 millert 212: /* Global variables */
213:
1.21 tedu 214: static int centerP = 0; /* Try to center lines? */
215: static size_t goal_length = 0; /* Target length for output lines */
216: static size_t max_length = 0; /* Maximum length for output lines */
217: static int coalesce_spaces_P = 0; /* Coalesce multiple whitespace -> ' ' ? */
218: static int allow_indented_paragraphs = 0; /* Can first line have diff. ind.? */
219: static int tab_width = 8; /* Number of spaces per tab stop */
220: static size_t output_tab_width = 0; /* Ditto, when squashing leading spaces */
221: static const char *sentence_enders = ".?!"; /* Double-space after these */
222: static int grok_mail_headers = 0; /* treat embedded mail headers magically? */
223: static int format_troff = 0; /* Format troff? */
224:
1.28 schwarze 225: static int n_errors = 0; /* Number of failed files. */
1.21 tedu 226: static size_t x; /* Horizontal position in output line */
227: static size_t x0; /* Ditto, ignoring leading whitespace */
228: static size_t pending_spaces; /* Spaces to add before next word */
229: static int output_in_paragraph = 0; /* Any of current para written out yet? */
1.10 millert 230:
231: /* Prototypes */
232:
1.21 tedu 233: static void process_named_file(const char *);
234: static void process_stream(FILE *, const char *);
1.34 schwarze 235: static size_t indent_length(const char *);
1.30 deraadt 236: static int might_be_header(const char *);
1.34 schwarze 237: static void new_paragraph(size_t);
238: static void output_word(size_t, size_t, const char *, int, int, int);
1.21 tedu 239: static void output_indent(size_t);
240: static void center_stream(FILE *, const char *);
1.34 schwarze 241: static char *get_line(FILE *);
1.21 tedu 242: static void *xrealloc(void *, size_t);
243: void usage(void);
1.10 millert 244:
1.29 lum 245: #define ERRS(x) (x >= 127 ? 127 : ++x)
1.10 millert 246:
247: /* Here is perhaps the right place to mention that this code is
248: * all in top-down order. Hence, |main| comes first.
1.1 deraadt 249: */
1.4 millert 250: int
1.21 tedu 251: main(int argc, char *argv[])
252: {
253: int ch; /* used for |getopt| processing */
254:
255: (void)setlocale(LC_CTYPE, "");
256:
1.33 deraadt 257: if (pledge("stdio rpath", NULL) == -1)
258: err(1, "pledge");
1.31 deraadt 259:
1.21 tedu 260: /* 1. Grok parameters. */
261: while ((ch = getopt(argc, argv, "0123456789cd:hl:mnpst:w:")) != -1) {
262: switch (ch) {
263: case 'c':
264: centerP = 1;
265: break;
266: case 'd':
267: sentence_enders = optarg;
268: break;
269: case 'l':
270: output_tab_width
271: = get_positive(optarg, "output tab width must be positive", 1);
272: break;
273: case 'm':
274: grok_mail_headers = 1;
275: break;
276: case 'n':
277: format_troff = 1;
278: break;
279: case 'p':
280: allow_indented_paragraphs = 1;
281: break;
282: case 's':
283: coalesce_spaces_P = 1;
284: break;
285: case 't':
286: tab_width = get_positive(optarg, "tab width must be positive", 1);
287: break;
288: case 'w':
289: goal_length = get_positive(optarg, "width must be positive", 1);
290: max_length = goal_length;
291: break;
292: case '0': case '1': case '2': case '3': case '4': case '5':
293: case '6': case '7': case '8': case '9':
294: /* XXX this is not a stylistically approved use of getopt() */
295: if (goal_length == 0) {
296: char *p;
297:
298: p = argv[optind - 1];
299: if (p[0] == '-' && p[1] == ch && !p[2])
300: goal_length = get_positive(++p, "width must be nonzero", 1);
301: else
302: goal_length = get_positive(argv[optind]+1,
303: "width must be nonzero", 1);
304: max_length = goal_length;
305: }
306: break;
307: case 'h':
308: default:
309: usage();
310: /* NOT REACHED */
311: }
312: }
313:
314: argc -= optind;
315: argv += optind;
316:
317: /* [ goal [ maximum ] ] */
318: if (argc > 0 && goal_length == 0 &&
319: (goal_length = get_positive(*argv,"goal length must be positive", 0)) != 0) {
320: --argc;
321: ++argv;
322: if (argc > 0 && (max_length = get_positive(*argv,"max length must be positive", 0)) != 0) {
323: --argc;
324: ++argv;
325: if (max_length < goal_length)
1.35 millert 326: errx(1, "max length must be >= goal length");
1.21 tedu 327: }
328: }
329:
330: if (goal_length == 0)
331: goal_length = 65;
332: if (max_length == 0)
333: max_length = goal_length+10;
334:
335: /* 2. Process files. */
336:
337: if (argc > 0) {
338: while (argc-- > 0)
339: process_named_file(*argv++);
340: } else {
1.33 deraadt 341: if (pledge("stdio", NULL) == -1)
342: err(1, "pledge");
1.21 tedu 343: process_stream(stdin, "standard input");
344: }
1.10 millert 345:
1.21 tedu 346: /* We're done. */
1.29 lum 347: return n_errors;
1.10 millert 348:
349: }
350:
351: /* Process a single file, given its name.
352: */
353: static void
1.21 tedu 354: process_named_file(const char *name)
355: {
356: FILE *f;
357:
358: if ((f = fopen(name, "r")) == NULL) {
1.23 cloder 359: warn("%s", name);
1.29 lum 360: ERRS(n_errors);
1.21 tedu 361: } else {
362: process_stream(f, name);
363: fclose(f);
364: }
1.10 millert 365: }
366:
367: /* Types of mail header continuation lines:
368: */
369: typedef enum {
1.21 tedu 370: hdr_ParagraphStart = -1,
371: hdr_NonHeader = 0,
372: hdr_Header = 1,
373: hdr_Continuation = 2
1.10 millert 374: } HdrType;
375:
376: /* Process a stream. This is where the real work happens,
377: * except that centering is handled separately.
378: */
379: static void
1.21 tedu 380: process_stream(FILE *stream, const char *name)
381: {
1.34 schwarze 382: const char *wordp, *cp;
383: wchar_t wc;
1.21 tedu 384: size_t np;
385: size_t last_indent = SILLY; /* how many spaces in last indent? */
386: size_t para_line_number = 0; /* how many lines already read in this para? */
387: size_t first_indent = SILLY; /* indentation of line 0 of paragraph */
1.34 schwarze 388: int wcl; /* number of bytes in wide character */
389: int wcw; /* display width of wide character */
390: int word_length; /* number of bytes in word */
391: int word_width; /* display width of word */
392: int space_width; /* display width of space after word */
393: int line_width; /* display width of line */
1.21 tedu 394: HdrType prev_header_type = hdr_ParagraphStart;
395: HdrType header_type;
396:
1.10 millert 397: /* ^-- header_type of previous line; -1 at para start */
1.30 deraadt 398: const char *line;
1.10 millert 399:
1.21 tedu 400: if (centerP) {
401: center_stream(stream, name);
402: return;
403: }
404:
1.34 schwarze 405: while ((line = get_line(stream)) != NULL) {
406: np = indent_length(line);
1.21 tedu 407: header_type = hdr_NonHeader;
408: if (grok_mail_headers && prev_header_type != hdr_NonHeader) {
409: if (np == 0 && might_be_header(line))
410: header_type = hdr_Header;
411: else if (np > 0 && prev_header_type>hdr_NonHeader)
412: header_type = hdr_Continuation;
413: }
414:
415: /* We need a new paragraph if and only if:
416: * this line is blank,
417: * OR it's a troff request,
418: * OR it's a mail header,
419: * OR it's not a mail header AND the last line was one,
420: * OR the indentation has changed
421: * AND the line isn't a mail header continuation line
422: * AND this isn't the second line of an indented paragraph.
423: */
1.34 schwarze 424: if (*line == '\0' || (*line == '.' && !format_troff) ||
1.21 tedu 425: header_type == hdr_Header ||
426: (header_type == hdr_NonHeader && prev_header_type > hdr_NonHeader) ||
427: (np != last_indent && header_type != hdr_Continuation &&
428: (!allow_indented_paragraphs || para_line_number != 1)) ) {
1.34 schwarze 429: new_paragraph(np);
1.21 tedu 430: para_line_number = 0;
431: first_indent = np;
432: last_indent = np;
433:
434: /* nroff compatibility */
1.34 schwarze 435: if (*line == '.' && !format_troff) {
436: puts(line);
1.21 tedu 437: continue;
438: }
439: if (header_type == hdr_Header)
440: last_indent = 2; /* for cont. lines */
1.34 schwarze 441: if (*line == '\0') {
1.21 tedu 442: putchar('\n');
443: prev_header_type = hdr_ParagraphStart;
444: continue;
445: } else {
446: /* If this is an indented paragraph other than a mail header
447: * continuation, set |last_indent|.
448: */
449: if (np != last_indent && header_type != hdr_Continuation)
450: last_indent = np;
451: }
452: prev_header_type = header_type;
453: }
454:
1.34 schwarze 455: line_width = np;
456: for (wordp = line; *wordp != '\0'; wordp = cp) {
457: word_length = 0;
458: word_width = space_width = 0;
459: for (cp = wordp; *cp != '\0'; cp += wcl) {
460: wcl = mbtowc(&wc, cp, MB_CUR_MAX);
461: if (wcl == -1) {
462: (void)mbtowc(NULL, NULL, MB_CUR_MAX);
463: wc = L'?';
464: wcl = 1;
465: wcw = 1;
466: } else if (wc == L'\t')
467: wcw = (line_width / tab_width + 1) *
468: tab_width - line_width;
469: else if ((wcw = wcwidth(wc)) == -1)
470: wcw = 1;
1.37 ! schwarze 471: if (iswblank(wc) && wc != 0xa0) {
1.34 schwarze 472: /* Skip whitespace at start of line. */
473: if (word_length == 0) {
474: wordp += wcl;
475: continue;
476: }
477: /* Count whitespace after word. */
478: space_width += wcw;
479: } else {
480: /* Detect end of word. */
481: if (space_width > 0)
482: break;
483: /* Measure word. */
484: word_length += wcl;
485: word_width += wcw;
486: }
487: line_width += wcw;
488: }
489:
1.21 tedu 490: /* Send the word to the output machinery. */
1.34 schwarze 491: output_word(first_indent, last_indent, wordp,
492: word_length, word_width, space_width);
1.21 tedu 493: }
494: ++para_line_number;
495: }
496:
1.34 schwarze 497: new_paragraph(0);
1.21 tedu 498: if (ferror(stream)) {
1.23 cloder 499: warn("%s", name);
1.29 lum 500: ERRS(n_errors);
1.21 tedu 501: }
1.10 millert 502: }
503:
504: /* How long is the indent on this line?
505: */
506: static size_t
1.34 schwarze 507: indent_length(const char *line)
1.21 tedu 508: {
509: size_t n = 0;
510:
1.34 schwarze 511: for (;;) {
512: switch(*line++) {
513: case ' ':
514: ++n;
515: continue;
516: case '\t':
517: n = (n / tab_width + 1) * tab_width;
518: continue;
519: default:
520: break;
521: }
522: break;
523: }
1.21 tedu 524: return n;
1.10 millert 525: }
526:
527: /* Might this line be a mail header?
528: * We deem a line to be a possible header if it matches the
529: * Perl regexp /^[A-Z][-A-Za-z0-9]*:\s/. This is *not* the same
530: * as in RFC whatever-number-it-is; we want to be gratuitously
531: * conservative to avoid mangling ordinary civilised text.
532: */
533: static int
1.30 deraadt 534: might_be_header(const char *line)
1.21 tedu 535: {
536:
1.30 deraadt 537: if (!isupper((unsigned char)*line++))
1.21 tedu 538: return 0;
1.30 deraadt 539: while (isalnum((unsigned char)*line) || *line == '-')
1.21 tedu 540: ++line;
1.30 deraadt 541: return (*line == ':' && isspace((unsigned char)line[1]));
1.10 millert 542: }
543:
544: /* Begin a new paragraph with an indent of |indent| spaces.
545: */
546: static void
1.34 schwarze 547: new_paragraph(size_t indent)
1.21 tedu 548: {
549:
1.34 schwarze 550: if (x0 > 0)
1.21 tedu 551: putchar('\n');
552: x = indent;
553: x0 = 0;
554: pending_spaces = 0;
555: output_in_paragraph = 0;
1.10 millert 556: }
557:
558: /* Output spaces or tabs for leading indentation.
559: */
560: static void
1.21 tedu 561: output_indent(size_t n_spaces)
562: {
563:
1.34 schwarze 564: if (n_spaces == 0)
565: return;
1.21 tedu 566: if (output_tab_width) {
567: while (n_spaces >= output_tab_width) {
568: putchar('\t');
569: n_spaces -= output_tab_width;
570: }
571: }
572: while (n_spaces-- > 0)
573: putchar(' ');
1.10 millert 574: }
575:
1.34 schwarze 576: /* Output a single word.
1.10 millert 577: * indent0 and indent1 are the indents to use on the first and subsequent
578: * lines of a paragraph. They'll often be the same, of course.
579: */
580: static void
1.34 schwarze 581: output_word(size_t indent0, size_t indent1, const char *word,
582: int length, int width, int spaces)
1.21 tedu 583: {
1.34 schwarze 584: size_t new_x = x + pending_spaces + width;
1.21 tedu 585:
586: /* If either |spaces==0| (at end of line) or |coalesce_spaces_P|
587: * (squashing internal whitespace), then add just one space;
588: * except that if the last character was a sentence-ender we
589: * actually add two spaces.
590: */
591: if (coalesce_spaces_P || spaces == 0)
592: spaces = strchr(sentence_enders, word[length-1]) ? 2 : 1;
593:
1.34 schwarze 594: if (x0 == 0)
595: output_indent(output_in_paragraph ? indent1 : indent0);
596: else if (new_x > max_length || x >= goal_length ||
597: (new_x > goal_length && new_x-goal_length > goal_length-x)) {
598: putchar('\n');
599: output_indent(indent1);
600: x0 = 0;
601: x = indent1;
602: } else {
1.21 tedu 603: x0 += pending_spaces;
604: x += pending_spaces;
1.34 schwarze 605: while (pending_spaces--)
606: putchar(' ');
1.21 tedu 607: }
1.34 schwarze 608: x0 += width;
609: x += width;
610: while(length--)
611: putchar(*word++);
612: pending_spaces = spaces;
613: output_in_paragraph = 1;
1.10 millert 614: }
615:
616: /* Process a stream, but just center its lines rather than trying to
617: * format them neatly.
618: */
619: static void
1.21 tedu 620: center_stream(FILE *stream, const char *name)
621: {
1.36 schwarze 622: char *line, *cp;
623: wchar_t wc;
624: size_t l; /* Display width of the line. */
625: int wcw; /* Display width of one character. */
626: int wcl; /* Length in bytes of one character. */
1.21 tedu 627:
1.34 schwarze 628: while ((line = get_line(stream)) != NULL) {
1.36 schwarze 629: l = 0;
630: for (cp = line; *cp != '\0'; cp += wcl) {
631: if (*cp == '\t')
632: *cp = ' ';
633: if ((wcl = mbtowc(&wc, cp, MB_CUR_MAX)) == -1) {
634: (void)mbtowc(NULL, NULL, MB_CUR_MAX);
635: *cp = '?';
636: wcl = 1;
637: wcw = 1;
638: } else if ((wcw = wcwidth(wc)) == -1)
639: wcw = 1;
640: if (l == 0 && iswspace(wc))
641: line += wcl;
642: else
643: l += wcw;
644: }
1.21 tedu 645: while (l < goal_length) {
646: putchar(' ');
647: l += 2;
648: }
1.34 schwarze 649: puts(line);
1.21 tedu 650: }
651:
652: if (ferror(stream)) {
1.23 cloder 653: warn("%s", name);
1.29 lum 654: ERRS(n_errors);
1.21 tedu 655: }
1.10 millert 656: }
657:
1.34 schwarze 658: /* Get a single line from a stream. Strip control
1.10 millert 659: * characters and trailing whitespace, and handle backspaces.
1.34 schwarze 660: * Return the address of the buffer containing the line.
1.10 millert 661: * This can cope with arbitrarily long lines, and with lines
662: * without terminating \n.
663: * If there are no characters left or an error happens, we
1.34 schwarze 664: * return NULL.
1.10 millert 665: */
666: static char *
1.34 schwarze 667: get_line(FILE *stream)
1.21 tedu 668: {
669: int ch;
670: int troff = 0;
671: static char *buf = NULL;
672: static size_t length = 0;
673: size_t len = 0;
674:
675: if (buf == NULL) {
676: length = 100;
1.34 schwarze 677: buf = xrealloc(NULL, length);
1.21 tedu 678: }
679:
680: while ((ch = getc(stream)) != '\n' && ch != EOF) {
1.34 schwarze 681: if ((len == 0) && (ch == '.' && !format_troff))
1.21 tedu 682: troff = 1;
1.34 schwarze 683: if (troff || ch == '\t' || !iscntrl(ch)) {
684: if (len >= length) {
1.21 tedu 685: length *= 2;
686: buf = xrealloc(buf, length);
687: }
688: buf[len++] = ch;
689: } else if (ch == '\b') {
690: if (len)
691: --len;
692: }
693: }
1.34 schwarze 694: while (len > 0 && isspace((unsigned char)buf[len-1]))
695: --len;
696: buf[len] = '\0';
697: return (len > 0 || ch != EOF) ? buf : NULL;
1.10 millert 698: }
699:
700: /* (Re)allocate some memory, exiting with an error if we can't.
701: */
702: static void *
1.21 tedu 703: xrealloc(void *ptr, size_t nbytes)
704: {
705: void *p;
706:
707: p = realloc(ptr, nbytes);
708: if (p == NULL)
1.35 millert 709: errx(1, "out of memory");
1.21 tedu 710: return p;
711: }
712:
713: void
714: usage(void)
715: {
1.22 mickey 716: extern char *__progname;
1.21 tedu 717:
718: fprintf(stderr,
1.25 jmc 719: "usage: %s [-cmnps] [-d chars] [-l number] [-t number]\n"
720: "\t[goal [maximum] | -width | -w width] [file ...]\n",
721: __progname);
1.22 mickey 722: exit (1);
1.1 deraadt 723: }