]> git.cameronkatri.com Git - mandoc.git/blobdiff - mandoc.c
more info from John Gardner about ASCII control chars in roff(7) input
[mandoc.git] / mandoc.c
index 20a6abf10dce717851011601316421419405e210..aad9a272f5c381f6b0d4c9edee89b984cdf8b61e 100644 (file)
--- a/mandoc.c
+++ b/mandoc.c
@@ -1,7 +1,7 @@
-/*     $Id: mandoc.c,v 1.64 2012/05/31 22:34:06 schwarze Exp $ */
+/*     $Id: mandoc.c,v 1.117 2020/01/19 16:44:50 schwarze Exp $ */
 /*
- * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
- * Copyright (c) 2011, 2012 Ingo Schwarze <schwarze@openbsd.org>
+ * Copyright (c) 2008-2011, 2014 Kristaps Dzonsons <kristaps@bsd.lv>
+ * Copyright (c) 2011-2015, 2017-2020 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
@@ -15,9 +15,7 @@
  * 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 <string.h>
 #include <time.h>
 
+#include "mandoc_aux.h"
 #include "mandoc.h"
+#include "roff.h"
 #include "libmandoc.h"
-
-#define DATESIZE 32
+#include "roff_int.h"
 
 static int      a2time(time_t *, const char *, const char *);
 static char    *time2a(time_t);
 
 
+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_FONTBOLD;
+                       case 'I':
+                               return ESCAPE_FONTITALIC;
+                       case 'R':
+                       case 'W':
+                               return ESCAPE_FONTCW;
+                       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)
 {
-       char             c, term;
-       int              i, rlim;
-       const char      *cp, *rstart;
-       enum mandoc_esc  gly; 
-
-       cp = *end;
-       rstart = cp;
-       if (start)
-               *start = rstart;
-       i = rlim = 0;
+       const char      *local_start;
+       int              local_sz, c, i;
+       char             term;
+       enum mandoc_esc  gly;
+
+       /*
+        * When the caller doesn't provide return storage,
+        * use local storage.
+        */
+
+       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 ((c = cp[i++])) {
+       switch ((*start)[-1]) {
        /*
         * First the glyphs.  There are several different forms of
         * these, but each eventually returns a substring of the glyph
         * name.
         */
-       case ('('):
+       case '(':
                gly = ESCAPE_SPECIAL;
-               rlim = 2;
+               *sz = 2;
                break;
-       case ('['):
+       case '[':
+               if (**start == ' ') {
+                       ++*end;
+                       return ESCAPE_ERROR;
+               }
                gly = ESCAPE_SPECIAL;
-               /*
-                * Unicode escapes are defined in groff as \[uXXXX] to
-                * \[u10FFFF], where the contained value must be a valid
-                * Unicode codepoint.  Here, however, only check whether
-                * it's not a zero-width escape.
-                */
-               if ('u' == cp[i] && ']' != cp[i + 1])
-                       gly = ESCAPE_UNICODE;
                term = ']';
                break;
-       case ('C'):
-               if ('\'' != cp[i])
-                       return(ESCAPE_ERROR);
+       case 'C':
+               if ('\'' != **start)
+                       return ESCAPE_ERROR;
+               *start = ++*end;
                gly = ESCAPE_SPECIAL;
                term = '\'';
                break;
 
+       /*
+        * 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.  
+        * character without advancing the cursor position.
         * Since we are mostly dealing with terminal mode,
         * let us just skip the next character.
         */
-       case ('z'):
-               (*end)++;
-               return(ESCAPE_SKIPCHAR);
+       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'):
-               /* FALLTHROUGH */
-       case ('g'):
-               /* FALLTHROUGH */
-       case ('k'):
-               /* FALLTHROUGH */
-       case ('M'):
-               /* FALLTHROUGH */
-       case ('m'):
-               /* FALLTHROUGH */
-       case ('n'):
-               /* FALLTHROUGH */
-       case ('V'):
-               /* FALLTHROUGH */
-       case ('Y'):
-               gly = ESCAPE_IGNORE;
-               /* FALLTHROUGH */
-       case ('f'):
-               if (ESCAPE_ERROR == gly)
-                       gly = ESCAPE_FONT;
-
-               rstart= &cp[i];
-               if (start) 
-                       *start = rstart;
-
-               switch (cp[i++]) {
-               case ('('):
-                       rlim = 2;
+       case 'F':
+       case 'f':
+       case 'g':
+       case 'k':
+       case 'M':
+       case 'm':
+       case 'n':
+       case 'O':
+       case 'V':
+       case 'Y':
+               gly = (*start)[-1] == 'f' ? ESCAPE_FONT : ESCAPE_IGNORE;
+               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:
-                       rlim = 1;
-                       i--;
+                       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 '*':
+               if (strncmp(*start, "(.T", 3) != 0)
+                       abort();
+               gly = ESCAPE_DEVICE;
+               *start = ++*end;
+               *sz = 2;
+               break;
 
        /*
         * 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'):
-               /* FALLTHROUGH */
-       case ('b'):
-               /* FALLTHROUGH */
-       case ('D'):
-               /* FALLTHROUGH */
-       case ('o'):
-               /* FALLTHROUGH */
-       case ('R'):
-               /* FALLTHROUGH */
-       case ('X'):
-               /* FALLTHROUGH */
-       case ('Z'):
-               if ('\'' != cp[i++])
-                       return(ESCAPE_ERROR);
+       case 'A':
+       case 'b':
+       case 'D':
+       case 'R':
+       case 'X':
+       case 'Z':
                gly = ESCAPE_IGNORE;
-               term = '\'';
+               /* FALLTHROUGH */
+       case 'o':
+               if (**start == '\0')
+                       return ESCAPE_ERROR;
+               if (gly == ESCAPE_ERROR)
+                       gly = ESCAPE_OVERSTRIKE;
+               term = **start;
+               *start = ++*end;
                break;
 
        /*
         * These escapes are of the form \X'N', where 'X' is the trigger
         * and 'N' resolves to a numerical expression.
         */
-       case ('B'):
-               /* FALLTHROUGH */
-       case ('h'):
-               /* FALLTHROUGH */
-       case ('H'):
-               /* FALLTHROUGH */
-       case ('L'):
-               /* FALLTHROUGH */
-       case ('l'):
-               gly = ESCAPE_NUMBERED;
-               /* FALLTHROUGH */
-       case ('S'):
-               /* FALLTHROUGH */
-       case ('v'):
-               /* FALLTHROUGH */
-       case ('w'):
-               /* FALLTHROUGH */
-       case ('x'):
-               if (ESCAPE_ERROR == gly)
+       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;
+               }
+               switch ((*start)[-1]) {
+               case 'h':
+                       gly = ESCAPE_HORIZ;
+                       break;
+               case 'l':
+                       gly = ESCAPE_HLINE;
+                       break;
+               default:
                        gly = ESCAPE_IGNORE;
-               if ('\'' != cp[i++])
-                       return(ESCAPE_ERROR);
-               term = '\'';
+                       break;
+               }
+               term = **start;
+               *start = ++*end;
                break;
 
        /*
         * Special handling for the numbered character escape.
         * XXX Do any other escapes need similar handling?
         */
-       case ('N'):
-               if ('\0' == cp[i])
-                       return(ESCAPE_ERROR);
-               *end = &cp[++i];
-               if (isdigit((unsigned char)cp[i-1]))
-                       return(ESCAPE_IGNORE);
+       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)++;
-               if (start)
-                       *start = &cp[i];
-               if (sz)
-                       *sz = *end - &cp[i];
+               *sz = *end - *start;
                if ('\0' != **end)
                        (*end)++;
-               return(ESCAPE_NUMBERED);
+               return ESCAPE_NUMBERED;
 
-       /* 
+       /*
         * Sizes get a special category of their own.
         */
-       case ('s'):
+       case 's':
                gly = ESCAPE_IGNORE;
 
-               rstart = &cp[i];
-               if (start) 
-                       *start = rstart;
-
                /* See +/- counts as a sign. */
-               c = cp[i];
-               if ('+' == c || '-' == c || ASCII_HYPH == c)
-                       ++i;
+               if ('+' == **end || '-' == **end || ASCII_HYPH == **end)
+                       *start = ++*end;
 
-               switch (cp[i++]) {
-               case ('('):
-                       rlim = 2;
+               switch (**end) {
+               case '(':
+                       *start = ++*end;
+                       *sz = 2;
                        break;
-               case ('['):
+               case '[':
+                       *start = ++*end;
                        term = ']';
                        break;
-               case ('\''):
+               case '\'':
+                       *start = ++*end;
                        term = '\'';
                        break;
+               case '3':
+               case '2':
+               case '1':
+                       *sz = (*end)[-1] == 's' &&
+                           isdigit((unsigned char)(*end)[1]) ? 2 : 1;
+                       break;
                default:
-                       rlim = 1;
-                       i--;
+                       *sz = 1;
                        break;
                }
 
-               /* See +/- counts as a sign. */
-               c = cp[i];
-               if ('+' == c || '-' == c || ASCII_HYPH == c)
-                       ++i;
-
                break;
 
        /*
-        * Anything else is assumed to be a glyph.
+        * Several special characters can be encoded as
+        * one-byte escape sequences without using \[].
         */
-       default:
+       case ' ':
+       case '\'':
+       case '-':
+       case '.':
+       case '0':
+       case ':':
+       case '_':
+       case '`':
+       case 'e':
+       case '~':
                gly = ESCAPE_SPECIAL;
-               rlim = 1;
-               i--;
+               /* FALLTHROUGH */
+       default:
+               if (gly == ESCAPE_ERROR)
+                       gly = ESCAPE_UNDEF;
+               *start = --*end;
+               *sz = 1;
                break;
        }
 
-       assert(ESCAPE_ERROR != gly);
-
-       *end = rstart = &cp[i];
-       if (start)
-               *start = rstart;
-
        /*
         * Read up to the terminating character,
         * paying attention to nested escapes.
@@ -271,219 +387,83 @@ mandoc_escape(const char **end, const char **start, int *sz)
        if ('\0' != term) {
                while (**end != term) {
                        switch (**end) {
-                       case ('\0'):
-                               return(ESCAPE_ERROR);
-                       case ('\\'):
+                       case '\0':
+                               return ESCAPE_ERROR;
+                       case '\\':
                                (*end)++;
                                if (ESCAPE_ERROR ==
                                    mandoc_escape(end, NULL, NULL))
-                                       return(ESCAPE_ERROR);
+                                       return ESCAPE_ERROR;
                                break;
                        default:
                                (*end)++;
                                break;
                        }
                }
-               rlim = (*end)++ - rstart;
+               *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(rlim > 0);
-               if ((size_t)rlim > strlen(rstart))
-                       return(ESCAPE_ERROR);
-               *end += rlim;
+               assert(*sz > 0);
+               if ((size_t)*sz > strlen(*start))
+                       return ESCAPE_ERROR;
+               *end += *sz;
        }
-       if (sz)
-               *sz = rlim;
 
        /* Run post-processors. */
 
        switch (gly) {
-       case (ESCAPE_FONT):
-               /*
-                * Pretend that the constant-width font modes are the
-                * same as the regular font modes.
-                */
-               if (2 == rlim && 'C' == *rstart)
-                       rstart++;
-               else if (1 != rlim)
+       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;
+               }
 
-               switch (*rstart) {
-               case ('3'):
-                       /* FALLTHROUGH */
-               case ('B'):
-                       gly = ESCAPE_FONTBOLD;
-                       break;
-               case ('2'):
-                       /* FALLTHROUGH */
-               case ('I'):
-                       gly = ESCAPE_FONTITALIC;
+               /*
+                * 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;
-               case ('P'):
-                       gly = ESCAPE_FONTPREV;
+               if (*sz == 7 && ((*start)[1] != '1' || (*start)[2] != '0'))
                        break;
-               case ('1'):
-                       /* FALLTHROUGH */
-               case ('R'):
-                       gly = ESCAPE_FONTROMAN;
+               if (*sz == 6 && (*start)[1] == '0')
                        break;
-               }
-               break;
-       case (ESCAPE_SPECIAL):
-               if (1 != rlim)
+               if (*sz == 5 && (*start)[1] == 'D' &&
+                   strchr("89ABCDEF", (*start)[2]) != NULL)
                        break;
-               if ('c' == *rstart)
-                       gly = ESCAPE_NOSPACE;
+               if ((int)strspn(*start + 1, "0123456789ABCDEFabcdef")
+                   + 1 == *sz)
+                       gly = ESCAPE_UNICODE;
                break;
        default:
                break;
        }
 
-       return(gly);
-}
-
-void *
-mandoc_calloc(size_t num, size_t size)
-{
-       void            *ptr;
-
-       ptr = calloc(num, size);
-       if (NULL == ptr) {
-               perror(NULL);
-               exit((int)MANDOCLEVEL_SYSERR);
-       }
-
-       return(ptr);
-}
-
-
-void *
-mandoc_malloc(size_t size)
-{
-       void            *ptr;
-
-       ptr = malloc(size);
-       if (NULL == ptr) {
-               perror(NULL);
-               exit((int)MANDOCLEVEL_SYSERR);
-       }
-
-       return(ptr);
-}
-
-
-void *
-mandoc_realloc(void *ptr, size_t size)
-{
-
-       ptr = realloc(ptr, size);
-       if (NULL == ptr) {
-               perror(NULL);
-               exit((int)MANDOCLEVEL_SYSERR);
-       }
-
-       return(ptr);
-}
-
-char *
-mandoc_strndup(const char *ptr, size_t sz)
-{
-       char            *p;
-
-       p = mandoc_malloc(sz + 1);
-       memcpy(p, ptr, sz);
-       p[(int)sz] = '\0';
-       return(p);
-}
-
-char *
-mandoc_strdup(const char *ptr)
-{
-       char            *p;
-
-       p = strdup(ptr);
-       if (NULL == p) {
-               perror(NULL);
-               exit((int)MANDOCLEVEL_SYSERR);
-       }
-
-       return(p);
-}
-
-/*
- * Parse a quoted or unquoted roff-style request or macro argument.
- * Return a pointer to the parsed argument, which is either the original
- * pointer or advanced by one byte in case the argument is quoted.
- * Null-terminate the argument in place.
- * Collapse pairs of quotes inside quoted arguments.
- * Advance the argument pointer to the next argument,
- * or to the null byte terminating the argument line.
- */
-char *
-mandoc_getarg(struct mparse *parse, char **cpp, int ln, int *pos)
-{
-       char     *start, *cp;
-       int       quoted, pairs, white;
-
-       /* Quoting can only start with a new word. */
-       start = *cpp;
-       quoted = 0;
-       if ('"' == *start) {
-               quoted = 1;
-               start++;
-       } 
-
-       pairs = 0;
-       white = 0;
-       for (cp = start; '\0' != *cp; cp++) {
-               /* Move left after quoted quotes and escaped backslashes. */
-               if (pairs)
-                       cp[-pairs] = cp[0];
-               if ('\\' == cp[0]) {
-                       if ('\\' == cp[1]) {
-                               /* Poor man's copy mode. */
-                               pairs++;
-                               cp++;
-                       } else if (0 == quoted && ' ' == cp[1])
-                               /* Skip escaped blanks. */
-                               cp++;
-               } else if (0 == quoted) {
-                       if (' ' == cp[0]) {
-                               /* Unescaped blanks end unquoted args. */
-                               white = 1;
-                               break;
-                       }
-               } else if ('"' == cp[0]) {
-                       if ('"' == cp[1]) {
-                               /* Quoted quotes collapse. */
-                               pairs++;
-                               cp++;
-                       } else {
-                               /* Unquoted quotes end quoted args. */
-                               quoted = 2;
-                               break;
-                       }
-               }
-       }
-
-       /* Quoted argument without a closing quote. */
-       if (1 == quoted)
-               mandoc_msg(MANDOCERR_BADQUOTE, parse, ln, *pos, NULL);
-
-       /* Null-terminate this argument and move to the next one. */
-       if (pairs)
-               cp[-pairs] = '\0';
-       if ('\0' != *cp) {
-               *cp++ = '\0';
-               while (' ' == *cp)
-                       cp++;
-       }
-       *pos += (int)(cp - start) + (quoted ? 1 : 0);
-       *cpp = cp;
-
-       if ('\0' == *cp && (white || ' ' == cp[-1]))
-               mandoc_msg(MANDOCERR_EOLNSPACE, parse, ln, *pos, NULL);
-
-       return(start);
+       return gly;
 }
 
 static int
@@ -495,15 +475,15 @@ a2time(time_t *t, const char *fmt, const char *p)
        memset(&tm, 0, sizeof(struct tm));
 
        pp = NULL;
-#ifdef HAVE_STRPTIME
+#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 *
@@ -514,7 +494,10 @@ time2a(time_t t)
        size_t           ssz;
        int              isz;
 
+       buf = NULL;
        tm = localtime(&t);
+       if (tm == NULL)
+               goto fail;
 
        /*
         * Reserve space:
@@ -522,55 +505,100 @@ time2a(time_t t)
         * 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 (0 == (ssz = strftime(p, 10 + 1, "%B ", tm)))
+       if ((ssz = strftime(p, 10 + 1, "%B ", tm)) == 0)
                goto fail;
        p += (int)ssz;
 
-       if (-1 == (isz = snprintf(p, 4 + 1, "%d, ", tm->tm_mday)))
+       /*
+        * 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 (0 == strftime(p, 4 + 1, "%Y", tm))
+       if (strftime(p, 4 + 1, "%Y", tm) == 0)
                goto fail;
-       return(buf);
+       return buf;
 
 fail:
        free(buf);
-       return(NULL);
+       return mandoc_strdup("");
 }
 
 char *
-mandoc_normdate(struct mparse *parse, char *in, int ln, int pos)
+mandoc_normdate(struct roff_node *nch, struct roff_node *nbl)
 {
-       char            *out;
+       char            *cp;
        time_t           t;
 
-       if (NULL == in || '\0' == *in ||
-           0 == strcmp(in, "$" "Mdocdate$")) {
-               mandoc_msg(MANDOCERR_NODATE, parse, ln, pos, NULL);
-               time(&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("");
        }
-       else if (a2time(&t, "%Y-%m-%d", in))
-               t = 0;
-       else if (!a2time(&t, "$" "Mdocdate: %b %d %Y $", in) &&
-           !a2time(&t, "%b %d, %Y", in)) {
-               mandoc_msg(MANDOCERR_BADDATE, parse, ln, pos, NULL);
-               t = 0;
+       if (*nch->string == '\0') {
+               mandoc_msg(MANDOCERR_DATE_MISSING, nch->line,
+                   nch->pos, "%s", roff_name[nbl->tok]);
+               return mandoc_strdup("");
        }
-       out = t ? time2a(t) : NULL;
-       return(out ? out : mandoc_strdup(in));
+       if (strcmp(nch->string, "$" "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;
+       }
+
+       /* In man(7), do not warn about the legacy format. */
+
+       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);
+
+       /* Use any non-mdoc(7) date verbatim. */
+
+       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
@@ -578,58 +606,28 @@ mandoc_eos(const char *p, size_t sz, int enclosed)
         * 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);
-}
-
-/*
- * Find out whether a line is a macro line or not.  If it is, adjust the
- * current position and return one; if it isn't, return zero and don't
- * change the current position.
- */
-int
-mandoc_getcontrol(const char *cp, int *ppos)
-{
-       int             pos;
-
-       pos = *ppos;
-
-       if ('\\' == cp[pos] && '.' == cp[pos + 1])
-               pos += 2;
-       else if ('.' == cp[pos] || '\'' == cp[pos])
-               pos++;
-       else
-               return(0);
-
-       while (' ' == cp[pos] || '\t' == cp[pos])
-               pos++;
-
-       *ppos = pos;
-       return(1);
+       return found && !enclosed;
 }
 
 /*
@@ -644,7 +642,7 @@ mandoc_strntoi(const char *p, size_t sz, int base)
        long             v;
 
        if (sz > 31)
-               return(-1);
+               return -1;
 
        memcpy(buf, p, sz);
        buf[(int)sz] = '\0';
@@ -653,12 +651,12 @@ mandoc_strntoi(const char *p, size_t sz, int base)
        v = strtol(buf, &ep, base);
 
        if (buf[0] == '\0' || *ep != '\0')
-               return(-1);
+               return -1;
 
        if (v > INT_MAX)
                v = INT_MAX;
        if (v < INT_MIN)
                v = INT_MIN;
 
-       return((int)v);
+       return (int)v;
 }