]> git.cameronkatri.com Git - mandoc.git/blobdiff - mandoc.c
Do not leak 64 bytes of heap memory every time a manual page calls
[mandoc.git] / mandoc.c
index 78b2967d7299ce55f522c0c0b861e886208b4b04..6adf1a4318b29f852c11ba9272a5d1953f554412 100644 (file)
--- a/mandoc.c
+++ b/mandoc.c
-/*     $Id: mandoc.c,v 1.32 2010/08/24 13:39:37 kristaps Exp $ */
+/*     $Id: mandoc.c,v 1.119 2021/08/10 12:55:03 schwarze Exp $ */
 /*
- * Copyright (c) 2008, 2009, 2010 Kristaps Dzonsons <kristaps@bsd.lv>
+ * Copyright (c) 2008-2011, 2014 Kristaps Dzonsons <kristaps@bsd.lv>
+ * Copyright (c) 2011-2015, 2017-2021 Ingo Schwarze <schwarze@openbsd.org>
  *
  * Permission to use, copy, modify, and distribute this software for any
  * purpose with or without fee is hereby granted, provided that the above
  * copyright notice and this permission notice appear in all copies.
  *
- * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
- * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
  */
-#ifdef HAVE_CONFIG_H
 #include "config.h"
-#endif
 
 #include <sys/types.h>
 
 #include <assert.h>
 #include <ctype.h>
+#include <errno.h>
+#include <limits.h>
 #include <stdlib.h>
 #include <stdio.h>
 #include <string.h>
 #include <time.h>
 
+#include "mandoc_aux.h"
 #include "mandoc.h"
+#include "roff.h"
 #include "libmandoc.h"
+#include "roff_int.h"
 
 static int      a2time(time_t *, const char *, const char *);
+static char    *time2a(time_t);
 
 
-int
-mandoc_special(char *p)
+enum mandoc_esc
+mandoc_font(const char *cp, int sz)
+{
+       switch (sz) {
+       case 0:
+               return ESCAPE_FONTPREV;
+       case 1:
+               switch (cp[0]) {
+               case 'B':
+               case '3':
+                       return ESCAPE_FONTBOLD;
+               case 'I':
+               case '2':
+                       return ESCAPE_FONTITALIC;
+               case 'P':
+                       return ESCAPE_FONTPREV;
+               case 'R':
+               case '1':
+                       return ESCAPE_FONTROMAN;
+               case '4':
+                       return ESCAPE_FONTBI;
+               default:
+                       return ESCAPE_ERROR;
+               }
+       case 2:
+               switch (cp[0]) {
+               case 'B':
+                       switch (cp[1]) {
+                       case 'I':
+                               return ESCAPE_FONTBI;
+                       default:
+                               return ESCAPE_ERROR;
+                       }
+               case 'C':
+                       switch (cp[1]) {
+                       case 'B':
+                               return ESCAPE_FONTCB;
+                       case 'I':
+                               return ESCAPE_FONTCI;
+                       case 'R':
+                       case 'W':
+                               return ESCAPE_FONTCR;
+                       default:
+                               return ESCAPE_ERROR;
+                       }
+               default:
+                       return ESCAPE_ERROR;
+               }
+       default:
+               return ESCAPE_ERROR;
+       }
+}
+
+enum mandoc_esc
+mandoc_escape(const char **end, const char **start, int *sz)
 {
-       int              len, i;
+       const char      *local_start;
+       int              local_sz, c, i;
        char             term;
-       char            *sv;
-       
-       len = 0;
-       term = '\0';
-       sv = p;
+       enum mandoc_esc  gly;
 
-       assert('\\' == *p);
-       p++;
+       /*
+        * When the caller doesn't provide return storage,
+        * use local storage.
+        */
 
-       switch (*p++) {
-#if 0
-       case ('Z'):
-               /* FALLTHROUGH */
-       case ('X'):
-               /* FALLTHROUGH */
-       case ('x'):
-               /* FALLTHROUGH */
-       case ('S'):
-               /* FALLTHROUGH */
-       case ('R'):
-               /* FALLTHROUGH */
-       case ('o'):
-               /* FALLTHROUGH */
-       case ('N'):
-               /* FALLTHROUGH */
-       case ('l'):
-               /* FALLTHROUGH */
-       case ('L'):
-               /* FALLTHROUGH */
-       case ('H'):
-               /* FALLTHROUGH */
-       case ('h'):
-               /* FALLTHROUGH */
-       case ('D'):
-               /* FALLTHROUGH */
-       case ('C'):
-               /* FALLTHROUGH */
-       case ('b'):
-               /* FALLTHROUGH */
-       case ('B'):
-               /* FALLTHROUGH */
-       case ('a'):
-               /* FALLTHROUGH */
-       case ('A'):
-               if (*p++ != '\'')
-                       return(0);
-               term = '\'';
+       if (NULL == start)
+               start = &local_start;
+       if (NULL == sz)
+               sz = &local_sz;
+
+       /*
+        * Treat "\E" just like "\";
+        * it only makes a difference in copy mode.
+        */
+
+       if (**end == 'E')
+               ++*end;
+
+       /*
+        * Beyond the backslash, at least one input character
+        * is part of the escape sequence.  With one exception
+        * (see below), that character won't be returned.
+        */
+
+       gly = ESCAPE_ERROR;
+       *start = ++*end;
+       *sz = 0;
+       term = '\0';
+
+       switch ((*start)[-1]) {
+       /*
+        * First the glyphs.  There are several different forms of
+        * these, but each eventually returns a substring of the glyph
+        * name.
+        */
+       case '(':
+               gly = ESCAPE_SPECIAL;
+               *sz = 2;
                break;
-#endif
-       case ('h'):
-               /* FALLTHROUGH */
-       case ('v'):
-               /* FALLTHROUGH */
-       case ('s'):
-               if (ASCII_HYPH == *p)
-                       *p = '-';
-
-               i = 0;
-               if ('+' == *p || '-' == *p) {
-                       p++;
-                       i = 1;
+       case '[':
+               if (**start == ' ') {
+                       ++*end;
+                       return ESCAPE_ERROR;
                }
+               gly = ESCAPE_SPECIAL;
+               term = ']';
+               break;
+       case 'C':
+               if ('\'' != **start)
+                       return ESCAPE_ERROR;
+               *start = ++*end;
+               gly = ESCAPE_SPECIAL;
+               term = '\'';
+               break;
 
-               switch (*p++) {
-               case ('('):
-                       len = 2;
-                       break;
-               case ('['):
-                       term = ']';
+       /*
+        * Escapes taking no arguments at all.
+        */
+       case '!':
+       case '?':
+               return ESCAPE_UNSUPP;
+       case '%':
+       case '&':
+       case ')':
+       case ',':
+       case '/':
+       case '^':
+       case 'a':
+       case 'd':
+       case 'r':
+       case 't':
+       case 'u':
+       case '{':
+       case '|':
+       case '}':
+               return ESCAPE_IGNORE;
+       case 'c':
+               return ESCAPE_NOSPACE;
+       case 'p':
+               return ESCAPE_BREAK;
+
+       /*
+        * The \z escape is supposed to output the following
+        * character without advancing the cursor position.
+        * Since we are mostly dealing with terminal mode,
+        * let us just skip the next character.
+        */
+       case 'z':
+               return ESCAPE_SKIPCHAR;
+
+       /*
+        * Handle all triggers matching \X(xy, \Xx, and \X[xxxx], where
+        * 'X' is the trigger.  These have opaque sub-strings.
+        */
+       case 'F':
+       case 'f':
+       case 'g':
+       case 'k':
+       case 'M':
+       case 'm':
+       case 'n':
+       case 'O':
+       case 'V':
+       case 'Y':
+       case '*':
+               switch ((*start)[-1]) {
+               case 'f':
+                       gly = ESCAPE_FONT;
                        break;
-               case ('\''):
-                       term = '\'';
+               case '*':
+                       gly = ESCAPE_DEVICE;
                        break;
-               case ('0'):
-                       i = 1;
-                       /* FALLTHROUGH */
                default:
-                       len = 1;
-                       p--;
+                       gly = ESCAPE_IGNORE;
                        break;
                }
-
-               if (ASCII_HYPH == *p)
-                       *p = '-';
-               if ('+' == *p || '-' == *p) {
-                       if (i)
-                               return(0);
-                       p++;
-               } 
-               
-               break;
-#if 0
-       case ('Y'):
-               /* FALLTHROUGH */
-       case ('V'):
-               /* FALLTHROUGH */
-       case ('$'):
-               /* FALLTHROUGH */
-       case ('n'):
-               /* FALLTHROUGH */
-#endif
-       case ('k'):
-               /* FALLTHROUGH */
-       case ('M'):
-               /* FALLTHROUGH */
-       case ('m'):
-               /* FALLTHROUGH */
-       case ('f'):
-               /* FALLTHROUGH */
-       case ('F'):
-               /* FALLTHROUGH */
-       case ('*'):
-               switch (*p++) {
-               case ('('):
-                       len = 2;
+               switch (**start) {
+               case '(':
+                       if ((*start)[-1] == 'O')
+                               gly = ESCAPE_ERROR;
+                       *start = ++*end;
+                       *sz = 2;
                        break;
-               case ('['):
+               case '[':
+                       if ((*start)[-1] == 'O')
+                               gly = (*start)[1] == '5' ?
+                                   ESCAPE_UNSUPP : ESCAPE_ERROR;
+                       *start = ++*end;
                        term = ']';
                        break;
                default:
-                       len = 1;
-                       p--;
+                       if ((*start)[-1] == 'O') {
+                               switch (**start) {
+                               case '0':
+                                       gly = ESCAPE_UNSUPP;
+                                       break;
+                               case '1':
+                               case '2':
+                               case '3':
+                               case '4':
+                                       break;
+                               default:
+                                       gly = ESCAPE_ERROR;
+                                       break;
+                               }
+                       }
+                       *sz = 1;
                        break;
                }
                break;
-       case ('('):
-               len = 2;
-               break;
-       case ('['):
-               term = ']';
+
+       /*
+        * These escapes are of the form \X'Y', where 'X' is the trigger
+        * and 'Y' is any string.  These have opaque sub-strings.
+        * The \B and \w escapes are handled in roff.c, roff_res().
+        */
+       case 'A':
+       case 'b':
+       case 'D':
+       case 'R':
+       case 'X':
+       case 'Z':
+               gly = ESCAPE_IGNORE;
+               /* FALLTHROUGH */
+       case 'o':
+               if (**start == '\0')
+                       return ESCAPE_ERROR;
+               if (gly == ESCAPE_ERROR)
+                       gly = ESCAPE_OVERSTRIKE;
+               term = **start;
+               *start = ++*end;
                break;
-       case ('z'):
-               len = 1;
-               if ('\\' == *p) {
-                       p += mandoc_special(p);
-                       return(*p ? (int)(p - sv) : 0);
+
+       /*
+        * These escapes are of the form \X'N', where 'X' is the trigger
+        * and 'N' resolves to a numerical expression.
+        */
+       case 'h':
+       case 'H':
+       case 'L':
+       case 'l':
+       case 'S':
+       case 'v':
+       case 'x':
+               if (strchr(" %&()*+-./0123456789:<=>", **start)) {
+                       if ('\0' != **start)
+                               ++*end;
+                       return ESCAPE_ERROR;
                }
-               break;
-       case ('w'):
-               if ('\'' == *p++) {
-                       term = '\'';
+               switch ((*start)[-1]) {
+               case 'h':
+                       gly = ESCAPE_HORIZ;
+                       break;
+               case 'l':
+                       gly = ESCAPE_HLINE;
+                       break;
+               default:
+                       gly = ESCAPE_IGNORE;
                        break;
                }
-               /* FALLTHROUGH */
-       default:
-               len = 1;
-               p--;
+               term = **start;
+               *start = ++*end;
                break;
-       }
 
-       if (term) {
-               for ( ; *p && term != *p; p++)
-                       if (ASCII_HYPH == *p)
-                               *p = '-';
-               return(*p ? (int)(p - sv) : 0);
-       }
-
-       for (i = 0; *p && i < len; i++, p++)
-               if (ASCII_HYPH == *p)
-                       *p = '-';
-       return(i == len ? (int)(p - sv) : 0);
-}
-
-
-void *
-mandoc_calloc(size_t num, size_t size)
-{
-       void            *ptr;
+       /*
+        * Special handling for the numbered character escape.
+        * XXX Do any other escapes need similar handling?
+        */
+       case 'N':
+               if ('\0' == **start)
+                       return ESCAPE_ERROR;
+               (*end)++;
+               if (isdigit((unsigned char)**start)) {
+                       *sz = 1;
+                       return ESCAPE_IGNORE;
+               }
+               (*start)++;
+               while (isdigit((unsigned char)**end))
+                       (*end)++;
+               *sz = *end - *start;
+               if ('\0' != **end)
+                       (*end)++;
+               return ESCAPE_NUMBERED;
 
-       ptr = calloc(num, size);
-       if (NULL == ptr) {
-               perror(NULL);
-               exit(MANDOCLEVEL_SYSERR);
-       }
+       /*
+        * Sizes get a special category of their own.
+        */
+       case 's':
+               gly = ESCAPE_IGNORE;
 
-       return(ptr);
-}
+               /* See +/- counts as a sign. */
+               if ('+' == **end || '-' == **end || ASCII_HYPH == **end)
+                       *start = ++*end;
 
+               switch (**end) {
+               case '(':
+                       *start = ++*end;
+                       *sz = 2;
+                       break;
+               case '[':
+                       *start = ++*end;
+                       term = ']';
+                       break;
+               case '\'':
+                       *start = ++*end;
+                       term = '\'';
+                       break;
+               case '3':
+               case '2':
+               case '1':
+                       *sz = (*end)[-1] == 's' &&
+                           isdigit((unsigned char)(*end)[1]) ? 2 : 1;
+                       break;
+               default:
+                       *sz = 1;
+                       break;
+               }
 
-void *
-mandoc_malloc(size_t size)
-{
-       void            *ptr;
+               break;
 
-       ptr = malloc(size);
-       if (NULL == ptr) {
-               perror(NULL);
-               exit(MANDOCLEVEL_SYSERR);
+       /*
+        * Several special characters can be encoded as
+        * one-byte escape sequences without using \[].
+        */
+       case ' ':
+       case '\'':
+       case '-':
+       case '.':
+       case '0':
+       case ':':
+       case '_':
+       case '`':
+       case 'e':
+       case '~':
+               gly = ESCAPE_SPECIAL;
+               /* FALLTHROUGH */
+       default:
+               if (gly == ESCAPE_ERROR)
+                       gly = ESCAPE_UNDEF;
+               *start = --*end;
+               *sz = 1;
+               break;
        }
 
-       return(ptr);
-}
-
-
-void *
-mandoc_realloc(void *ptr, size_t size)
-{
+       /*
+        * Read up to the terminating character,
+        * paying attention to nested escapes.
+        */
 
-       ptr = realloc(ptr, size);
-       if (NULL == ptr) {
-               perror(NULL);
-               exit(MANDOCLEVEL_SYSERR);
+       if ('\0' != term) {
+               while (**end != term) {
+                       switch (**end) {
+                       case '\0':
+                               return ESCAPE_ERROR;
+                       case '\\':
+                               (*end)++;
+                               if (ESCAPE_ERROR ==
+                                   mandoc_escape(end, NULL, NULL))
+                                       return ESCAPE_ERROR;
+                               break;
+                       default:
+                               (*end)++;
+                               break;
+                       }
+               }
+               *sz = (*end)++ - *start;
+
+               /*
+                * The file chars.c only provides one common list
+                * of character names, but \[-] == \- is the only
+                * one of the characters with one-byte names that
+                * allows enclosing the name in brackets.
+                */
+               if (gly == ESCAPE_SPECIAL && *sz == 1 && **start != '-')
+                       return ESCAPE_ERROR;
+       } else {
+               assert(*sz > 0);
+               if ((size_t)*sz > strlen(*start))
+                       return ESCAPE_ERROR;
+               *end += *sz;
        }
 
-       return(ptr);
-}
-
+       /* Run post-processors. */
 
-char *
-mandoc_strdup(const char *ptr)
-{
-       char            *p;
+       switch (gly) {
+       case ESCAPE_FONT:
+               gly = mandoc_font(*start, *sz);
+               break;
+       case ESCAPE_SPECIAL:
+               if (**start == 'c') {
+                       if (*sz < 6 || *sz > 7 ||
+                           strncmp(*start, "char", 4) != 0 ||
+                           (int)strspn(*start + 4, "0123456789") + 4 < *sz)
+                               break;
+                       c = 0;
+                       for (i = 4; i < *sz; i++)
+                               c = 10 * c + ((*start)[i] - '0');
+                       if (c < 0x21 || (c > 0x7e && c < 0xa0) || c > 0xff)
+                               break;
+                       *start += 4;
+                       *sz -= 4;
+                       gly = ESCAPE_NUMBERED;
+                       break;
+               }
 
-       p = strdup(ptr);
-       if (NULL == p) {
-               perror(NULL);
-               exit(MANDOCLEVEL_SYSERR);
+               /*
+                * Unicode escapes are defined in groff as \[u0000]
+                * to \[u10FFFF], where the contained value must be
+                * a valid Unicode codepoint.  Here, however, only
+                * check the length and range.
+                */
+               if (**start != 'u' || *sz < 5 || *sz > 7)
+                       break;
+               if (*sz == 7 && ((*start)[1] != '1' || (*start)[2] != '0'))
+                       break;
+               if (*sz == 6 && (*start)[1] == '0')
+                       break;
+               if (*sz == 5 && (*start)[1] == 'D' &&
+                   strchr("89ABCDEF", (*start)[2]) != NULL)
+                       break;
+               if ((int)strspn(*start + 1, "0123456789ABCDEFabcdef")
+                   + 1 == *sz)
+                       gly = ESCAPE_UNICODE;
+               break;
+       case ESCAPE_DEVICE:
+               assert(*sz == 2 && (*start)[0] == '.' && (*start)[1] == 'T');
+               break;
+       default:
+               break;
        }
 
-       return(p);
+       return gly;
 }
 
-
 static int
 a2time(time_t *t, const char *fmt, const char *p)
 {
@@ -269,118 +481,189 @@ a2time(time_t *t, const char *fmt, const char *p)
 
        memset(&tm, 0, sizeof(struct tm));
 
+       pp = NULL;
+#if HAVE_STRPTIME
        pp = strptime(p, fmt, &tm);
+#endif
        if (NULL != pp && '\0' == *pp) {
                *t = mktime(&tm);
-               return(1);
+               return 1;
        }
 
-       return(0);
+       return 0;
 }
 
+static char *
+time2a(time_t t)
+{
+       struct tm       *tm;
+       char            *buf, *p;
+       size_t           ssz;
+       int              isz;
 
-/*
- * Convert from a manual date string (see mdoc(7) and man(7)) into a
- * date according to the stipulated date type.
- */
-time_t
-mandoc_a2time(int flags, const char *p)
+       buf = NULL;
+       tm = localtime(&t);
+       if (tm == NULL)
+               goto fail;
+
+       /*
+        * Reserve space:
+        * up to 9 characters for the month (September) + blank
+        * up to 2 characters for the day + comma + blank
+        * 4 characters for the year and a terminating '\0'
+        */
+
+       p = buf = mandoc_malloc(10 + 4 + 4 + 1);
+
+       if ((ssz = strftime(p, 10 + 1, "%B ", tm)) == 0)
+               goto fail;
+       p += (int)ssz;
+
+       /*
+        * The output format is just "%d" here, not "%2d" or "%02d".
+        * That's also the reason why we can't just format the
+        * date as a whole with "%B %e, %Y" or "%B %d, %Y".
+        * Besides, the present approach is less prone to buffer
+        * overflows, in case anybody should ever introduce the bug
+        * of looking at LC_TIME.
+        */
+
+       isz = snprintf(p, 4 + 1, "%d, ", tm->tm_mday);
+       if (isz < 0 || isz > 4)
+               goto fail;
+       p += isz;
+
+       if (strftime(p, 4 + 1, "%Y", tm) == 0)
+               goto fail;
+       return buf;
+
+fail:
+       free(buf);
+       return mandoc_strdup("");
+}
+
+char *
+mandoc_normdate(struct roff_node *nch, struct roff_node *nbl)
 {
+       char            *cp;
        time_t           t;
 
-       if (MTIME_MDOCDATE & flags) {
-               if (0 == strcmp(p, "$" "Mdocdate$"))
-                       return(time(NULL));
-               if (a2time(&t, "$" "Mdocdate: %b %d %Y $", p))
-                       return(t);
+       /* No date specified. */
+
+       if (nch == NULL) {
+               if (nbl == NULL)
+                       mandoc_msg(MANDOCERR_DATE_MISSING, 0, 0, NULL);
+               else
+                       mandoc_msg(MANDOCERR_DATE_MISSING, nbl->line,
+                           nbl->pos, "%s", roff_name[nbl->tok]);
+               return mandoc_strdup("");
+       }
+       if (*nch->string == '\0') {
+               mandoc_msg(MANDOCERR_DATE_MISSING, nch->line,
+                   nch->pos, "%s", roff_name[nbl->tok]);
+               return mandoc_strdup("");
+       }
+       if (strcmp(nch->string, "$" "Mdocdate$") == 0)
+               return time2a(time(NULL));
+
+       /* Valid mdoc(7) date format. */
+
+       if (a2time(&t, "$" "Mdocdate: %b %d %Y $", nch->string) ||
+           a2time(&t, "%b %d, %Y", nch->string)) {
+               cp = time2a(t);
+               if (t > time(NULL) + 86400)
+                       mandoc_msg(MANDOCERR_DATE_FUTURE, nch->line,
+                           nch->pos, "%s %s", roff_name[nbl->tok], cp);
+               else if (*nch->string != '$' &&
+                   strcmp(nch->string, cp) != 0)
+                       mandoc_msg(MANDOCERR_DATE_NORM, nch->line,
+                           nch->pos, "%s %s", roff_name[nbl->tok], cp);
+               return cp;
        }
 
-       if (MTIME_CANONICAL & flags || MTIME_REDUCED & flags) 
-               if (a2time(&t, "%b %d, %Y", p))
-                       return(t);
+       /* In man(7), do not warn about the legacy format. */
 
-       if (MTIME_ISO_8601 & flags) 
-               if (a2time(&t, "%Y-%m-%d", p))
-                       return(t);
+       if (a2time(&t, "%Y-%m-%d", nch->string) == 0)
+               mandoc_msg(MANDOCERR_DATE_BAD, nch->line, nch->pos,
+                   "%s %s", roff_name[nbl->tok], nch->string);
+       else if (t > time(NULL) + 86400)
+               mandoc_msg(MANDOCERR_DATE_FUTURE, nch->line, nch->pos,
+                   "%s %s", roff_name[nbl->tok], nch->string);
+       else if (nbl->tok == MDOC_Dd)
+               mandoc_msg(MANDOCERR_DATE_LEGACY, nch->line, nch->pos,
+                   "Dd %s", nch->string);
 
-       if (MTIME_REDUCED & flags) {
-               if (a2time(&t, "%d, %Y", p))
-                       return(t);
-               if (a2time(&t, "%Y", p))
-                       return(t);
-       }
+       /* Use any non-mdoc(7) date verbatim. */
 
-       return(0);
+       return mandoc_strdup(nch->string);
 }
 
-
 int
-mandoc_eos(const char *p, size_t sz, int enclosed)
+mandoc_eos(const char *p, size_t sz)
 {
-       const char *q;
-       int found;
+       const char      *q;
+       int              enclosed, found;
 
        if (0 == sz)
-               return(0);
+               return 0;
 
        /*
         * End-of-sentence recognition must include situations where
         * some symbols, such as `)', allow prior EOS punctuation to
-        * propogate outward.
+        * propagate outward.
         */
 
-       found = 0;
+       enclosed = found = 0;
        for (q = p + (int)sz - 1; q >= p; q--) {
                switch (*q) {
-               case ('\"'):
-                       /* FALLTHROUGH */
-               case ('\''):
-                       /* FALLTHROUGH */
-               case (']'):
-                       /* FALLTHROUGH */
-               case (')'):
+               case '\"':
+               case '\'':
+               case ']':
+               case ')':
                        if (0 == found)
                                enclosed = 1;
                        break;
-               case ('.'):
-                       /* FALLTHROUGH */
-               case ('!'):
-                       /* FALLTHROUGH */
-               case ('?'):
+               case '.':
+               case '!':
+               case '?':
                        found = 1;
                        break;
                default:
-                       return(found && (!enclosed || isalnum((unsigned char)*q)));
+                       return found &&
+                           (!enclosed || isalnum((unsigned char)*q));
                }
        }
 
-       return(found && !enclosed);
+       return found && !enclosed;
 }
 
-
+/*
+ * Convert a string to a long that may not be <0.
+ * If the string is invalid, or is less than 0, return -1.
+ */
 int
-mandoc_hyph(const char *start, const char *c)
+mandoc_strntoi(const char *p, size_t sz, int base)
 {
+       char             buf[32];
+       char            *ep;
+       long             v;
 
-       /*
-        * Choose whether to break at a hyphenated character.  We only
-        * do this if it's free-standing within a word.
-        */
+       if (sz > 31)
+               return -1;
+
+       memcpy(buf, p, sz);
+       buf[(int)sz] = '\0';
+
+       errno = 0;
+       v = strtol(buf, &ep, base);
+
+       if (buf[0] == '\0' || *ep != '\0')
+               return -1;
+
+       if (v > INT_MAX)
+               v = INT_MAX;
+       if (v < INT_MIN)
+               v = INT_MIN;
 
-       /* Skip first/last character of buffer. */
-       if (c == start || '\0' == *(c + 1))
-               return(0);
-       /* Skip first/last character of word. */
-       if ('\t' == *(c + 1) || '\t' == *(c - 1))
-               return(0);
-       if (' ' == *(c + 1) || ' ' == *(c - 1))
-               return(0);
-       /* Skip double invocations. */
-       if ('-' == *(c + 1) || '-' == *(c - 1))
-               return(0);
-       /* Skip escapes. */
-       if ('\\' == *(c - 1))
-               return(0);
-
-       return(1);
+       return (int)v;
 }