]> git.cameronkatri.com Git - mandoc.git/blobdiff - term.c
Reduce memory and time consumption on certain malformed input files
[mandoc.git] / term.c
diff --git a/term.c b/term.c
index c6a4015019c2fc6ffcd68203d15b6f883ea5dd82..e6e3f641b7a5393b7ac08a5890658e9fbeef744f 100644 (file)
--- a/term.c
+++ b/term.c
@@ -1,4 +1,4 @@
-/*     $Id: term.c,v 1.223 2014/04/23 21:06:41 schwarze Exp $ */
+/*     $Id: term.c,v 1.242 2014/12/24 23:32:42 schwarze Exp $ */
 /*
  * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
  * Copyright (c) 2010-2014 Ingo Schwarze <schwarze@openbsd.org>
@@ -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>
 
@@ -44,11 +42,8 @@ void
 term_free(struct termp *p)
 {
 
-       if (p->buf)
-               free(p->buf);
-       if (p->symtab)
-               mchars_free(p->symtab);
-
+       free(p->buf);
+       free(p->fontq);
        free(p);
 }
 
@@ -106,7 +101,6 @@ term_flushln(struct termp *p)
        size_t           j;     /* temporary loop index for p->buf */
        size_t           jhy;   /* last hyph before overflow w/r/t j */
        size_t           maxvis; /* output position of visible boundary */
-       size_t           mmax; /* used in calculating bp */
 
        /*
         * First, establish the maximum columns of "visible" content.
@@ -119,13 +113,16 @@ term_flushln(struct termp *p)
         * is negative, it gets sign extended.  Subtracting that
         * very large size_t effectively adds a small number to dv.
         */
-       assert  (p->rmargin >= p->offset);
-       dv     = p->rmargin - p->offset;
+       dv = p->rmargin > p->offset ? p->rmargin - p->offset : 0;
        maxvis = (int)dv > p->overstep ? dv - (size_t)p->overstep : 0;
-       dv     = p->maxrmargin - p->offset;
-       mmax   = (int)dv > p->overstep ? dv - (size_t)p->overstep : 0;
 
-       bp = TERMP_NOBREAK & p->flags ? mmax : maxvis;
+       if (p->flags & TERMP_NOBREAK) {
+               dv = p->maxrmargin > p->offset ?
+                    p->maxrmargin - p->offset : 0;
+               bp = (int)dv > p->overstep ?
+                    dv - (size_t)p->overstep : 0;
+       } else
+               bp = maxvis;
 
        /*
         * Calculate the required amount of padding.
@@ -195,7 +192,8 @@ term_flushln(struct termp *p)
                        p->viscol = 0;
                        if (TERMP_BRIND & p->flags) {
                                vbl = p->rmargin;
-                               vend += p->rmargin - p->offset;
+                               vend += p->rmargin;
+                               vend -= p->offset;
                        } else
                                vbl = p->offset;
 
@@ -222,7 +220,7 @@ term_flushln(struct termp *p)
                                break;
                        if (' ' == p->buf[i]) {
                                j = i;
-                               while (' ' == p->buf[i])
+                               while (i < p->col && ' ' == p->buf[i])
                                        i++;
                                dv = (i - j) * (*p->width)(p, ' ');
                                vbl += dv;
@@ -260,8 +258,10 @@ term_flushln(struct termp *p)
         * If there was trailing white space, it was not printed;
         * so reset the cursor position accordingly.
         */
-       if (vis)
+       if (vis > vbl)
                vis -= vbl;
+       else
+               vis = 0;
 
        p->col = 0;
        p->overstep = 0;
@@ -273,7 +273,7 @@ term_flushln(struct termp *p)
        }
 
        if (TERMP_HANG & p->flags) {
-               p->overstep = (int)(vis - maxvis +
+               p->overstep += (int)(p->offset + vis - p->rmargin +
                    p->trailspace * (*p->width)(p, ' '));
 
                /*
@@ -329,6 +329,7 @@ term_vspace(struct termp *p)
                (*p->endline)(p);
 }
 
+/* Swap current and previous font; for \fP and .ft P */
 void
 term_fontlast(struct termp *p)
 {
@@ -339,6 +340,7 @@ term_fontlast(struct termp *p)
        p->fontq[p->fonti] = f;
 }
 
+/* Set font, save current, discard previous; for \f, .ft, .B etc. */
 void
 term_fontrepl(struct termp *p, enum termfont f)
 {
@@ -347,38 +349,39 @@ term_fontrepl(struct termp *p, enum termfont f)
        p->fontq[p->fonti] = f;
 }
 
+/* Set font, save previous. */
 void
 term_fontpush(struct termp *p, enum termfont f)
 {
 
-       assert(p->fonti + 1 < 10);
        p->fontl = p->fontq[p->fonti];
-       p->fontq[++p->fonti] = f;
+       if (++p->fonti == p->fontsz) {
+               p->fontsz += 8;
+               p->fontq = mandoc_reallocarray(p->fontq,
+                   p->fontsz, sizeof(enum termfont *));
+       }
+       p->fontq[p->fonti] = f;
 }
 
-const void *
+/* Retrieve pointer to current font. */
+const enum termfont *
 term_fontq(struct termp *p)
 {
 
        return(&p->fontq[p->fonti]);
 }
 
-enum termfont
-term_fonttop(struct termp *p)
-{
-
-       return(p->fontq[p->fonti]);
-}
-
+/* Flush to make the saved pointer current again. */
 void
-term_fontpopq(struct termp *p, const void *key)
+term_fontpopq(struct termp *p, const enum termfont *key)
 {
 
-       while (p->fonti >= 0 && key < (void *)(p->fontq + p->fonti))
+       while (p->fonti >= 0 && key < p->fontq + p->fonti)
                p->fonti--;
        assert(p->fonti >= 0);
 }
 
+/* Pop one font off the stack. */
 void
 term_fontpop(struct termp *p)
 {
@@ -397,7 +400,6 @@ term_word(struct termp *p, const char *word)
 {
        const char       nbrsp[2] = { ASCII_NBRSP, 0 };
        const char      *seq, *cp;
-       char             c;
        int              sz, uc;
        size_t           ssz;
        enum mandoc_esc  esc;
@@ -418,7 +420,7 @@ term_word(struct termp *p, const char *word)
        else
                p->flags |= TERMP_NOSPACE;
 
-       p->flags &= ~TERMP_SENTENCE;
+       p->flags &= ~(TERMP_SENTENCE | TERMP_NONEWLINE);
 
        while ('\0' != *word) {
                if ('\\' != *word) {
@@ -444,70 +446,72 @@ term_word(struct termp *p, const char *word)
                word++;
                esc = mandoc_escape(&word, &seq, &sz);
                if (ESCAPE_ERROR == esc)
-                       break;
-
-               if (TERMENC_ASCII != p->enc)
-                       switch (esc) {
-                       case ESCAPE_UNICODE:
-                               uc = mchars_num2uc(seq + 1, sz - 1);
-                               if ('\0' == uc)
-                                       break;
-                               encode1(p, uc);
-                               continue;
-                       case ESCAPE_SPECIAL:
-                               uc = mchars_spec2cp(p->symtab, seq, sz);
-                               if (uc <= 0)
-                                       break;
-                               encode1(p, uc);
-                               continue;
-                       default:
-                               break;
-                       }
+                       continue;
 
                switch (esc) {
                case ESCAPE_UNICODE:
-                       encode1(p, '?');
+                       uc = mchars_num2uc(seq + 1, sz - 1);
                        break;
                case ESCAPE_NUMBERED:
-                       c = mchars_num2char(seq, sz);
-                       if ('\0' != c)
-                               encode(p, &c, 1);
+                       uc = mchars_num2char(seq, sz);
+                       if (uc < 0)
+                               continue;
                        break;
                case ESCAPE_SPECIAL:
-                       cp = mchars_spec2str(p->symtab, seq, sz, &ssz);
-                       if (NULL != cp)
-                               encode(p, cp, ssz);
-                       else if (1 == ssz)
-                               encode(p, seq, sz);
-                       break;
+                       if (p->enc == TERMENC_ASCII) {
+                               cp = mchars_spec2str(p->symtab,
+                                   seq, sz, &ssz);
+                               if (cp != NULL)
+                                       encode(p, cp, ssz);
+                       } else {
+                               uc = mchars_spec2cp(p->symtab, seq, sz);
+                               if (uc > 0)
+                                       encode1(p, uc);
+                       }
+                       continue;
                case ESCAPE_FONTBOLD:
                        term_fontrepl(p, TERMFONT_BOLD);
-                       break;
+                       continue;
                case ESCAPE_FONTITALIC:
                        term_fontrepl(p, TERMFONT_UNDER);
-                       break;
+                       continue;
                case ESCAPE_FONTBI:
                        term_fontrepl(p, TERMFONT_BI);
-                       break;
+                       continue;
                case ESCAPE_FONT:
                        /* FALLTHROUGH */
                case ESCAPE_FONTROMAN:
                        term_fontrepl(p, TERMFONT_NONE);
-                       break;
+                       continue;
                case ESCAPE_FONTPREV:
                        term_fontlast(p);
-                       break;
+                       continue;
                case ESCAPE_NOSPACE:
                        if (TERMP_SKIPCHAR & p->flags)
                                p->flags &= ~TERMP_SKIPCHAR;
                        else if ('\0' == *word)
-                               p->flags |= TERMP_NOSPACE;
-                       break;
+                               p->flags |= (TERMP_NOSPACE | TERMP_NONEWLINE);
+                       continue;
                case ESCAPE_SKIPCHAR:
                        p->flags |= TERMP_SKIPCHAR;
-                       break;
+                       continue;
                default:
-                       break;
+                       continue;
+               }
+
+               /*
+                * Common handling for Unicode and numbered
+                * character escape sequences.
+                */
+
+               if (p->enc == TERMENC_ASCII) {
+                       cp = ascii_uc2str(uc);
+                       encode(p, cp, strlen(cp));
+               } else {
+                       if ((uc < 0x20 && uc != 0x09) ||
+                           (uc > 0x7E && uc < 0xA0))
+                               uc = 0xFFFD;
+                       encode1(p, uc);
                }
        }
        p->flags &= ~TERMP_NBRWORD;
@@ -553,7 +557,7 @@ encode1(struct termp *p, int c)
        if (p->col + 6 >= p->maxcols)
                adjbuf(p, p->col + 6);
 
-       f = term_fonttop(p);
+       f = *term_fontq(p);
 
        if (TERMFONT_UNDER == f || TERMFONT_BI == f) {
                p->buf[p->col++] = '_';
@@ -585,7 +589,7 @@ encode(struct termp *p, const char *word, size_t sz)
         * character by character.
         */
 
-       if (TERMFONT_NONE == term_fonttop(p)) {
+       if (*term_fontq(p) == TERMFONT_NONE) {
                if (p->col + sz >= p->maxcols)
                        adjbuf(p, p->col + sz);
                for (i = 0; i < sz; i++)
@@ -659,7 +663,7 @@ size_t
 term_strlen(const struct termp *p, const char *cp)
 {
        size_t           sz, rsz, i;
-       int              ssz, skip, c;
+       int              ssz, skip, uc;
        const char      *seq, *rhs;
        enum mandoc_esc  esc;
        static const char rej[] = { '\\', ASCII_NBRSP, ASCII_HYPH,
@@ -683,64 +687,67 @@ term_strlen(const struct termp *p, const char *cp)
                        cp++;
                        esc = mandoc_escape(&cp, &seq, &ssz);
                        if (ESCAPE_ERROR == esc)
-                               return(sz);
-
-                       if (TERMENC_ASCII != p->enc)
-                               switch (esc) {
-                               case ESCAPE_UNICODE:
-                                       c = mchars_num2uc(seq + 1,
-                                           ssz - 1);
-                                       if ('\0' == c)
-                                               break;
-                                       sz += cond_width(p, c, &skip);
-                                       continue;
-                               case ESCAPE_SPECIAL:
-                                       c = mchars_spec2cp(p->symtab,
-                                           seq, ssz);
-                                       if (c <= 0)
-                                               break;
-                                       sz += cond_width(p, c, &skip);
-                                       continue;
-                               default:
-                                       break;
-                               }
+                               continue;
 
                        rhs = NULL;
 
                        switch (esc) {
                        case ESCAPE_UNICODE:
-                               sz += cond_width(p, '?', &skip);
+                               uc = mchars_num2uc(seq + 1, ssz - 1);
                                break;
                        case ESCAPE_NUMBERED:
-                               c = mchars_num2char(seq, ssz);
-                               if ('\0' != c)
-                                       sz += cond_width(p, c, &skip);
+                               uc = mchars_num2char(seq, ssz);
+                               if (uc < 0)
+                                       continue;
                                break;
                        case ESCAPE_SPECIAL:
-                               rhs = mchars_spec2str(p->symtab,
-                                   seq, ssz, &rsz);
-
-                               if (ssz != 1 || rhs)
-                                       break;
-
-                               rhs = seq;
-                               rsz = ssz;
-                               break;
+                               if (p->enc == TERMENC_ASCII) {
+                                       rhs = mchars_spec2str(p->symtab,
+                                           seq, ssz, &rsz);
+                                       if (rhs != NULL)
+                                               break;
+                               } else {
+                                       uc = mchars_spec2cp(p->symtab,
+                                           seq, ssz);
+                                       if (uc > 0)
+                                               sz += cond_width(p, uc, &skip);
+                               }
+                               continue;
                        case ESCAPE_SKIPCHAR:
                                skip = 1;
-                               break;
+                               continue;
                        default:
-                               break;
+                               continue;
                        }
 
-                       if (NULL == rhs)
-                               break;
+                       /*
+                        * Common handling for Unicode and numbered
+                        * character escape sequences.
+                        */
+
+                       if (rhs == NULL) {
+                               if (p->enc == TERMENC_ASCII) {
+                                       rhs = ascii_uc2str(uc);
+                                       rsz = strlen(rhs);
+                               } else {
+                                       if ((uc < 0x20 && uc != 0x09) ||
+                                           (uc > 0x7E && uc < 0xA0))
+                                               uc = 0xFFFD;
+                                       sz += cond_width(p, uc, &skip);
+                                       continue;
+                               }
+                       }
 
                        if (skip) {
                                skip = 0;
                                break;
                        }
 
+                       /*
+                        * Common handling for all escape sequences
+                        * printing more than one character.
+                        */
+
                        for (i = 0; i < rsz; i++)
                                sz += (*p->width)(p, *rhs++);
                        break;
@@ -762,47 +769,55 @@ term_strlen(const struct termp *p, const char *cp)
        return(sz);
 }
 
-size_t
+int
 term_vspan(const struct termp *p, const struct roffsu *su)
 {
        double           r;
+       int              ri;
 
        switch (su->unit) {
+       case SCALE_BU:
+               r = su->scale / 40.0;
+               break;
        case SCALE_CM:
-               r = su->scale * 2;
+               r = su->scale * 6.0 / 2.54;
+               break;
+       case SCALE_FS:
+               r = su->scale * 65536.0 / 40.0;
                break;
        case SCALE_IN:
-               r = su->scale * 6;
+               r = su->scale * 6.0;
+               break;
+       case SCALE_MM:
+               r = su->scale * 0.006;
                break;
        case SCALE_PC:
                r = su->scale;
                break;
        case SCALE_PT:
-               r = su->scale / 8;
+               r = su->scale / 12.0;
                break;
-       case SCALE_MM:
-               r = su->scale / 1000;
+       case SCALE_EN:
+               /* FALLTHROUGH */
+       case SCALE_EM:
+               r = su->scale * 0.6;
                break;
        case SCALE_VS:
                r = su->scale;
                break;
        default:
-               r = su->scale - 1;
-               break;
+               abort();
+               /* NOTREACHED */
        }
-
-       if (r < 0.0)
-               r = 0.0;
-       return((size_t)r);
+       ri = r > 0.0 ? r + 0.4995 : r - 0.4995;
+       return(ri < 66 ? ri : 1);
 }
 
-size_t
+int
 term_hspan(const struct termp *p, const struct roffsu *su)
 {
        double           v;
 
-       v = ((*p->hspan)(p, su));
-       if (v < 0.0)
-               v = 0.0;
-       return((size_t)v);
+       v = (*p->hspan)(p, su);
+       return(v > 0.0 ? v + 0.0005 : v - 0.0005);
 }