]> git.cameronkatri.com Git - mandoc.git/blobdiff - html.c
Two minor improvements:
[mandoc.git] / html.c
diff --git a/html.c b/html.c
index da381a478f14b1158fa47a7eae4296f10ac15791..71c9c711e825851b2b52bec5a6d58d5904dcde11 100644 (file)
--- a/html.c
+++ b/html.c
@@ -1,7 +1,7 @@
-/*     $Id: html.c,v 1.246 2018/12/15 19:30:26 schwarze Exp $ */
+/* $Id: html.c,v 1.275 2021/09/09 14:47:24 schwarze Exp $ */
 /*
  * Copyright (c) 2008-2011, 2014 Kristaps Dzonsons <kristaps@bsd.lv>
- * Copyright (c) 2011-2015, 2017, 2018 Ingo Schwarze <schwarze@openbsd.org>
+ * 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
@@ -14,6 +14,9 @@
  * 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.
+ *
+ * Common functions for mandoc(1) HTML formatters.
+ * For use by individual formatters and by the main program.
  */
 #include "config.h"
 
 struct htmldata {
        const char       *name;
        int               flags;
-#define        HTML_NOSTACK     (1 << 0)
-#define        HTML_AUTOCLOSE   (1 << 1)
-#define        HTML_NLBEFORE    (1 << 2)
-#define        HTML_NLBEGIN     (1 << 3)
-#define        HTML_NLEND       (1 << 4)
-#define        HTML_NLAFTER     (1 << 5)
+#define        HTML_INPHRASE    (1 << 0)  /* Can appear in phrasing context. */
+#define        HTML_TOPHRASE    (1 << 1)  /* Establishes phrasing context. */
+#define        HTML_NOSTACK     (1 << 2)  /* Does not have an end tag. */
+#define        HTML_NLBEFORE    (1 << 3)  /* Output line break before opening. */
+#define        HTML_NLBEGIN     (1 << 4)  /* Output line break after opening. */
+#define        HTML_NLEND       (1 << 5)  /* Output line break before closing. */
+#define        HTML_NLAFTER     (1 << 6)  /* Output line break after closing. */
 #define        HTML_NLAROUND    (HTML_NLBEFORE | HTML_NLAFTER)
 #define        HTML_NLINSIDE    (HTML_NLBEGIN | HTML_NLEND)
 #define        HTML_NLALL       (HTML_NLAROUND | HTML_NLINSIDE)
-#define        HTML_INDENT      (1 << 6)
-#define        HTML_NOINDENT    (1 << 7)
+#define        HTML_INDENT      (1 << 7)  /* Indent content by two spaces. */
+#define        HTML_NOINDENT    (1 << 8)  /* Exception: never indent content. */
 };
 
 static const struct htmldata htmltags[TAG_MAX] = {
        {"html",        HTML_NLALL},
        {"head",        HTML_NLALL | HTML_INDENT},
-       {"body",        HTML_NLALL},
-       {"meta",        HTML_NOSTACK | HTML_AUTOCLOSE | HTML_NLALL},
+       {"meta",        HTML_NOSTACK | HTML_NLALL},
+       {"link",        HTML_NOSTACK | HTML_NLALL},
+       {"style",       HTML_NLALL | HTML_INDENT},
        {"title",       HTML_NLAROUND},
+       {"body",        HTML_NLALL},
        {"div",         HTML_NLAROUND},
-       {"div",         0},
-       {"h1",          HTML_NLAROUND},
-       {"h2",          HTML_NLAROUND},
-       {"span",        0},
-       {"link",        HTML_NOSTACK | HTML_AUTOCLOSE | HTML_NLALL},
-       {"br",          HTML_NOSTACK | HTML_AUTOCLOSE | HTML_NLALL},
-       {"a",           0},
+       {"section",     HTML_NLALL},
        {"table",       HTML_NLALL | HTML_INDENT},
        {"tr",          HTML_NLALL | HTML_INDENT},
        {"td",          HTML_NLAROUND},
@@ -78,15 +78,22 @@ static      const struct htmldata htmltags[TAG_MAX] = {
        {"dl",          HTML_NLALL | HTML_INDENT},
        {"dt",          HTML_NLAROUND},
        {"dd",          HTML_NLAROUND | HTML_INDENT},
-       {"pre",         HTML_NLALL | HTML_NOINDENT},
-       {"var",         0},
-       {"cite",        0},
-       {"b",           0},
-       {"i",           0},
-       {"code",        0},
-       {"small",       0},
-       {"style",       HTML_NLALL | HTML_INDENT},
-       {"math",        HTML_NLALL | HTML_INDENT},
+       {"h1",          HTML_TOPHRASE | HTML_NLAROUND},
+       {"h2",          HTML_TOPHRASE | HTML_NLAROUND},
+       {"p",           HTML_TOPHRASE | HTML_NLAROUND | HTML_INDENT},
+       {"pre",         HTML_TOPHRASE | HTML_NLAROUND | HTML_NOINDENT},
+       {"a",           HTML_INPHRASE | HTML_TOPHRASE},
+       {"b",           HTML_INPHRASE | HTML_TOPHRASE},
+       {"cite",        HTML_INPHRASE | HTML_TOPHRASE},
+       {"code",        HTML_INPHRASE | HTML_TOPHRASE},
+       {"i",           HTML_INPHRASE | HTML_TOPHRASE},
+       {"small",       HTML_INPHRASE | HTML_TOPHRASE},
+       {"span",        HTML_INPHRASE | HTML_TOPHRASE},
+       {"var",         HTML_INPHRASE | HTML_TOPHRASE},
+       {"br",          HTML_INPHRASE | HTML_NOSTACK | HTML_NLALL},
+       {"hr",          HTML_INPHRASE | HTML_NOSTACK},
+       {"mark",        HTML_INPHRASE },
+       {"math",        HTML_INPHRASE | HTML_NLALL | HTML_INDENT},
        {"mrow",        0},
        {"mi",          0},
        {"mn",          0},
@@ -106,8 +113,14 @@ static     const struct htmldata htmltags[TAG_MAX] = {
 };
 
 /* Avoid duplicate HTML id= attributes. */
+
+struct id_entry {
+       int      ord;   /* Ordinal number of the latest occurrence. */
+       char     id[];  /* The id= attribute without any ordinal suffix. */
+};
 static struct ohash     id_unique;
 
+static void     html_reset_internal(struct html *);
 static void     print_byte(struct html *, char);
 static void     print_endword(struct html *);
 static void     print_indent(struct html *);
@@ -117,7 +130,7 @@ static      void     print_ctag(struct html *, struct tag *);
 static int      print_escape(struct html *, char);
 static int      print_encode(struct html *, const char *, const char *, int);
 static void     print_href(struct html *, const char *, const char *, int);
-static void     print_metaf(struct html *, enum mandoc_esc);
+static void     print_metaf(struct html *);
 
 
 void *
@@ -128,6 +141,7 @@ html_alloc(const struct manoutput *outopts)
        h = mandoc_calloc(1, sizeof(struct html));
 
        h->tag = NULL;
+       h->metac = h->metal = ESCAPE_FONTROMAN;
        h->style = outopts->style;
        if ((h->base_man1 = outopts->man) == NULL)
                h->base_man2 = NULL;
@@ -139,40 +153,52 @@ html_alloc(const struct manoutput *outopts)
        if (outopts->toc)
                h->oflags |= HTML_TOC;
 
-       mandoc_ohash_init(&id_unique, 4, 0);
+       mandoc_ohash_init(&id_unique, 4, offsetof(struct id_entry, id));
 
        return h;
 }
 
-void
-html_free(void *p)
+static void
+html_reset_internal(struct html *h)
 {
        struct tag      *tag;
-       struct html     *h;
-       char            *cp;
+       struct id_entry *entry;
        unsigned int     slot;
 
-       h = (struct html *)p;
        while ((tag = h->tag) != NULL) {
                h->tag = tag->next;
                free(tag);
        }
-       free(h);
-
-       cp = ohash_first(&id_unique, &slot);
-       while (cp != NULL) {
-               free(cp);
-               cp = ohash_next(&id_unique, &slot);
+       entry = ohash_first(&id_unique, &slot);
+       while (entry != NULL) {
+               free(entry);
+               entry = ohash_next(&id_unique, &slot);
        }
        ohash_delete(&id_unique);
 }
 
+void
+html_reset(void *p)
+{
+       html_reset_internal(p);
+       mandoc_ohash_init(&id_unique, 4, offsetof(struct id_entry, id));
+}
+
+void
+html_free(void *p)
+{
+       html_reset_internal(p);
+       free(p);
+}
+
 void
 print_gen_head(struct html *h)
 {
        struct tag      *t;
 
        print_otag(h, TAG_META, "?", "charset", "utf-8");
+       print_otag(h, TAG_META, "??", "name", "viewport",
+           "content", "width=device-width, initial-scale=1.0");
        if (h->style != NULL) {
                print_otag(h, TAG_LINK, "?h??", "rel", "stylesheet",
                    h->style, "type", "text/css", "media", "all");
@@ -190,109 +216,193 @@ print_gen_head(struct html *h)
        print_endline(h);
        print_text(h, "td.head-vol { text-align: center; }");
        print_endline(h);
-       print_text(h, "div.Pp { margin: 1ex 0ex; }");
-       print_endline(h);
-       print_text(h, "div.Nd, div.Bf, div.Op { display: inline; }");
+       print_text(h, ".Nd, .Bf, .Op { display: inline; }");
        print_endline(h);
-       print_text(h, "span.Pa, span.Ad { font-style: italic; }");
+       print_text(h, ".Pa, .Ad { font-style: italic; }");
        print_endline(h);
-       print_text(h, "span.Ms { font-weight: bold; }");
+       print_text(h, ".Ms { font-weight: bold; }");
        print_endline(h);
-       print_text(h, "dl.Bl-diag ");
+       print_text(h, ".Bl-diag ");
        print_byte(h, '>');
        print_text(h, " dt { font-weight: bold; }");
        print_endline(h);
-       print_text(h, "code.Nm, code.Fl, code.Cm, code.Ic, "
-           "code.In, code.Fd, code.Fn,");
-       print_endline(h);
-       print_text(h, "code.Cd { font-weight: bold; "
-           "font-family: inherit; }");
+       print_text(h, "code.Nm, .Fl, .Cm, .Ic, code.In, .Fd, .Fn, .Cd "
+           "{ font-weight: bold; font-family: inherit; }");
        print_tagq(h, t);
 }
 
-static void
-print_metaf(struct html *h, enum mandoc_esc deco)
+int
+html_setfont(struct html *h, enum mandoc_esc font)
 {
-       enum htmlfont    font;
-
-       switch (deco) {
+       switch (font) {
        case ESCAPE_FONTPREV:
                font = h->metal;
                break;
        case ESCAPE_FONTITALIC:
-               font = HTMLFONT_ITALIC;
-               break;
        case ESCAPE_FONTBOLD:
-               font = HTMLFONT_BOLD;
-               break;
        case ESCAPE_FONTBI:
-               font = HTMLFONT_BI;
-               break;
-       case ESCAPE_FONTCW:
-               font = HTMLFONT_CW;
+       case ESCAPE_FONTROMAN:
+       case ESCAPE_FONTCR:
+       case ESCAPE_FONTCB:
+       case ESCAPE_FONTCI:
                break;
        case ESCAPE_FONT:
-       case ESCAPE_FONTROMAN:
-               font = HTMLFONT_NONE;
+               font = ESCAPE_FONTROMAN;
                break;
        default:
-               abort();
+               return 0;
        }
+       h->metal = h->metac;
+       h->metac = font;
+       return 1;
+}
 
+static void
+print_metaf(struct html *h)
+{
        if (h->metaf) {
                print_tagq(h, h->metaf);
                h->metaf = NULL;
        }
-
-       h->metal = h->metac;
-       h->metac = font;
-
-       switch (font) {
-       case HTMLFONT_ITALIC:
+       switch (h->metac) {
+       case ESCAPE_FONTITALIC:
                h->metaf = print_otag(h, TAG_I, "");
                break;
-       case HTMLFONT_BOLD:
+       case ESCAPE_FONTBOLD:
                h->metaf = print_otag(h, TAG_B, "");
                break;
-       case HTMLFONT_BI:
+       case ESCAPE_FONTBI:
                h->metaf = print_otag(h, TAG_B, "");
                print_otag(h, TAG_I, "");
                break;
-       case HTMLFONT_CW:
+       case ESCAPE_FONTCR:
+               h->metaf = print_otag(h, TAG_SPAN, "c", "Li");
+               break;
+       case ESCAPE_FONTCB:
+               h->metaf = print_otag(h, TAG_SPAN, "c", "Li");
+               print_otag(h, TAG_B, "");
+               break;
+       case ESCAPE_FONTCI:
                h->metaf = print_otag(h, TAG_SPAN, "c", "Li");
+               print_otag(h, TAG_I, "");
                break;
        default:
                break;
        }
 }
 
+void
+html_close_paragraph(struct html *h)
+{
+       struct tag      *this, *next;
+       int              flags;
+
+       this = h->tag;
+       for (;;) {
+               next = this->next;
+               flags = htmltags[this->tag].flags;
+               if (flags & (HTML_INPHRASE | HTML_TOPHRASE))
+                       print_ctag(h, this);
+               if ((flags & HTML_INPHRASE) == 0)
+                       break;
+               this = next;
+       }
+}
+
+/*
+ * ROFF_nf switches to no-fill mode, ROFF_fi to fill mode.
+ * TOKEN_NONE does not switch.  The old mode is returned.
+ */
+enum roff_tok
+html_fillmode(struct html *h, enum roff_tok want)
+{
+       struct tag      *t;
+       enum roff_tok    had;
+
+       for (t = h->tag; t != NULL; t = t->next)
+               if (t->tag == TAG_PRE)
+                       break;
+
+       had = t == NULL ? ROFF_fi : ROFF_nf;
+
+       if (want != had) {
+               switch (want) {
+               case ROFF_fi:
+                       print_tagq(h, t);
+                       break;
+               case ROFF_nf:
+                       html_close_paragraph(h);
+                       print_otag(h, TAG_PRE, "");
+                       break;
+               case TOKEN_NONE:
+                       break;
+               default:
+                       abort();
+               }
+       }
+       return had;
+}
+
+/*
+ * Allocate a string to be used for the "id=" attribute of an HTML
+ * element and/or as a segment identifier for a URI in an <a> element.
+ * The function may fail and return NULL if the node lacks text data
+ * to create the attribute from.
+ * The caller is responsible for free(3)ing the returned string.
+ *
+ * If the "unique" argument is non-zero, the "id_unique" ohash table
+ * is used for de-duplication.  If the "unique" argument is 1,
+ * it is the first time the function is called for this tag and
+ * location, so if an ordinal suffix is needed, it is incremented.
+ * If the "unique" argument is 2, it is the second time the function
+ * is called for this tag and location, so the ordinal suffix
+ * remains unchanged.
+ */
 char *
 html_make_id(const struct roff_node *n, int unique)
 {
        const struct roff_node  *nch;
-       char                    *buf, *bufs, *cp;
+       struct id_entry         *entry;
+       char                    *buf, *cp;
+       size_t                   len;
        unsigned int             slot;
-       int                      suffix;
-
-       for (nch = n->child; nch != NULL; nch = nch->next)
-               if (nch->type != ROFFT_TEXT)
-                       return NULL;
 
-       buf = NULL;
-       deroff(&buf, n);
-       if (buf == NULL)
-               return NULL;
+       if (n->tag != NULL)
+               buf = mandoc_strdup(n->tag);
+       else {
+               switch (n->tok) {
+               case MDOC_Sh:
+               case MDOC_Ss:
+               case MDOC_Sx:
+               case MAN_SH:
+               case MAN_SS:
+                       for (nch = n->child; nch != NULL; nch = nch->next)
+                               if (nch->type != ROFFT_TEXT)
+                                       return NULL;
+                       buf = NULL;
+                       deroff(&buf, n);
+                       if (buf == NULL)
+                               return NULL;
+                       break;
+               default:
+                       if (n->child == NULL || n->child->type != ROFFT_TEXT)
+                               return NULL;
+                       buf = mandoc_strdup(n->child->string);
+                       break;
+               }
+       }
 
        /*
         * In ID attributes, only use ASCII characters that are
         * permitted in URL-fragment strings according to the
         * explicit list at:
         * https://url.spec.whatwg.org/#url-fragment-string
+        * In addition, reserve '~' for ordinal suffixes.
         */
 
        for (cp = buf; *cp != '\0'; cp++)
                if (isalnum((unsigned char)*cp) == 0 &&
-                   strchr("!$&'()*+,-./:;=?@_~", *cp) == NULL)
+                   strchr("!$&'()*+,-./:;=?@_", *cp) == NULL)
                        *cp = '_';
 
        if (unique == 0)
@@ -300,25 +410,21 @@ html_make_id(const struct roff_node *n, int unique)
 
        /* Avoid duplicate HTML id= attributes. */
 
-       bufs = NULL;
-       suffix = 1;
        slot = ohash_qlookup(&id_unique, buf);
-       cp = ohash_find(&id_unique, slot);
-       if (cp != NULL) {
-               while (cp != NULL) {
-                       free(bufs);
-                       if (++suffix > 127) {
-                               free(buf);
-                               return NULL;
-                       }
-                       mandoc_asprintf(&bufs, "%s_%d", buf, suffix);
-                       slot = ohash_qlookup(&id_unique, bufs);
-                       cp = ohash_find(&id_unique, slot);
-               }
-               free(buf);
-               buf = bufs;
+       if ((entry = ohash_find(&id_unique, slot)) == NULL) {
+               len = strlen(buf) + 1;
+               entry = mandoc_malloc(sizeof(*entry) + len);
+               entry->ord = 1;
+               memcpy(entry->id, buf, len);
+               ohash_insert(&id_unique, slot, entry);
+       } else if (unique == 1)
+               entry->ord++;
+
+       if (entry->ord > 1) {
+               cp = buf;
+               mandoc_asprintf(&buf, "%s~%d", cp, entry->ord);
+               free(cp);
        }
-       ohash_insert(&id_unique, slot, buf);
        return buf;
 }
 
@@ -408,11 +514,14 @@ print_encode(struct html *h, const char *p, const char *pend, int norecurse)
                case ESCAPE_FONTBOLD:
                case ESCAPE_FONTITALIC:
                case ESCAPE_FONTBI:
-               case ESCAPE_FONTCW:
                case ESCAPE_FONTROMAN:
+               case ESCAPE_FONTCR:
+               case ESCAPE_FONTCB:
+               case ESCAPE_FONTCI:
                        if (0 == norecurse) {
                                h->flags |= HTML_NOSPACE;
-                               print_metaf(h, esc);
+                               if (html_setfont(h, esc))
+                                       print_metaf(h);
                                h->flags &= ~HTML_NOSPACE;
                        }
                        continue;
@@ -526,12 +635,33 @@ print_otag(struct html *h, enum htmltag tag, const char *fmt, ...)
 
        tflags = htmltags[tag].flags;
 
+       /* Flow content is not allowed in phrasing context. */
+
+       if ((tflags & HTML_INPHRASE) == 0) {
+               for (t = h->tag; t != NULL; t = t->next) {
+                       if (t->closed)
+                               continue;
+                       assert((htmltags[t->tag].flags & HTML_TOPHRASE) == 0);
+                       break;
+               }
+
+       /*
+        * Always wrap phrasing elements in a paragraph
+        * unless already contained in some flow container;
+        * never put them directly into a section.
+        */
+
+       } else if (tflags & HTML_TOPHRASE && h->tag->tag == TAG_SECTION)
+               print_otag(h, TAG_P, "c", "Pp");
+
        /* Push this tag onto the stack of open scopes. */
 
        if ((tflags & HTML_NOSTACK) == 0) {
                t = mandoc_malloc(sizeof(struct tag));
                t->tag = tag;
                t->next = h->tag;
+               t->refcnt = 0;
+               t->closed = 0;
                h->tag = t;
        } else
                t = NULL;
@@ -610,12 +740,6 @@ print_otag(struct html *h, enum htmltag tag, const char *fmt, ...)
                        print_encode(h, arg1, NULL, 1);
                        fmt++;
                        break;
-               case 'T':
-                       print_encode(h, arg1, NULL, 1);
-                       print_word(h, "\" title=\"");
-                       print_encode(h, arg1, NULL, 1);
-                       fmt++;
-                       break;
                default:
                        print_encode(h, arg1, NULL, 1);
                        break;
@@ -647,7 +771,7 @@ print_otag(struct html *h, enum htmltag tag, const char *fmt, ...)
 
        /* Accommodate for "well-formed" singleton escaping. */
 
-       if (HTML_AUTOCLOSE & htmltags[tag].flags)
+       if (htmltags[tag].flags & HTML_NOSTACK)
                print_byte(h, '/');
 
        print_byte(h, '>');
@@ -665,38 +789,80 @@ print_otag(struct html *h, enum htmltag tag, const char *fmt, ...)
        return t;
 }
 
+/*
+ * Print an element with an optional "id=" attribute.
+ * If the element has phrasing content and an "id=" attribute,
+ * also add a permalink: outside if it can be in phrasing context,
+ * inside otherwise.
+ */
+struct tag *
+print_otag_id(struct html *h, enum htmltag elemtype, const char *cattr,
+    struct roff_node *n)
+{
+       struct roff_node *nch;
+       struct tag      *ret, *t;
+       char            *id, *href;
+
+       ret = NULL;
+       id = href = NULL;
+       if (n->flags & NODE_ID)
+               id = html_make_id(n, 1);
+       if (n->flags & NODE_HREF)
+               href = id == NULL ? html_make_id(n, 2) : id;
+       if (href != NULL && htmltags[elemtype].flags & HTML_INPHRASE)
+               ret = print_otag(h, TAG_A, "chR", "permalink", href);
+       t = print_otag(h, elemtype, "ci", cattr, id);
+       if (ret == NULL) {
+               ret = t;
+               if (href != NULL && (nch = n->child) != NULL) {
+                       /* man(7) is safe, it tags phrasing content only. */
+                       if (n->tok > MDOC_MAX ||
+                           htmltags[elemtype].flags & HTML_TOPHRASE)
+                               nch = NULL;
+                       else  /* For mdoc(7), beware of nested blocks. */
+                               while (nch != NULL && nch->type == ROFFT_TEXT)
+                                       nch = nch->next;
+                       if (nch == NULL)
+                               print_otag(h, TAG_A, "chR", "permalink", href);
+               }
+       }
+       free(id);
+       if (id == NULL)
+               free(href);
+       return ret;
+}
+
 static void
 print_ctag(struct html *h, struct tag *tag)
 {
        int      tflags;
 
-       /*
-        * Remember to close out and nullify the current
-        * meta-font and table, if applicable.
-        */
-       if (tag == h->metaf)
-               h->metaf = NULL;
-       if (tag == h->tblt)
-               h->tblt = NULL;
-
-       tflags = htmltags[tag->tag].flags;
-
-       if (tflags & HTML_INDENT)
-               h->indent--;
-       if (tflags & HTML_NOINDENT)
-               h->noindent--;
-       if (tflags & HTML_NLEND)
-               print_endline(h);
-       print_indent(h);
-       print_byte(h, '<');
-       print_byte(h, '/');
-       print_word(h, htmltags[tag->tag].name);
-       print_byte(h, '>');
-       if (tflags & HTML_NLAFTER)
-               print_endline(h);
-
-       h->tag = tag->next;
-       free(tag);
+       if (tag->closed == 0) {
+               tag->closed = 1;
+               if (tag == h->metaf)
+                       h->metaf = NULL;
+               if (tag == h->tblt)
+                       h->tblt = NULL;
+
+               tflags = htmltags[tag->tag].flags;
+               if (tflags & HTML_INDENT)
+                       h->indent--;
+               if (tflags & HTML_NOINDENT)
+                       h->noindent--;
+               if (tflags & HTML_NLEND)
+                       print_endline(h);
+               print_indent(h);
+               print_byte(h, '<');
+               print_byte(h, '/');
+               print_word(h, htmltags[tag->tag].name);
+               print_byte(h, '>');
+               if (tflags & HTML_NLAFTER)
+                       print_endline(h);
+       }
+       if (tag->refcnt == 0) {
+               h->tag = tag->next;
+               free(tag);
+       }
 }
 
 void
@@ -735,6 +901,25 @@ print_gen_comment(struct html *h, struct roff_node *n)
 void
 print_text(struct html *h, const char *word)
 {
+       print_tagged_text(h, word, NULL);
+}
+
+void
+print_tagged_text(struct html *h, const char *word, struct roff_node *n)
+{
+       struct tag      *t;
+       char            *href;
+
+       /*
+        * Always wrap text in a paragraph unless already contained in
+        * some flow container; never put it directly into a section.
+        */
+
+       if (h->tag->tag == TAG_SECTION)
+               print_otag(h, TAG_P, "c", "Pp");
+
+       /* Output whitespace before this text? */
+
        if (h->col && (h->flags & HTML_NOSPACE) == 0) {
                if ( ! (HTML_KEEP & h->flags)) {
                        if (HTML_PREKEEP & h->flags)
@@ -744,27 +929,21 @@ print_text(struct html *h, const char *word)
                        print_word(h, "&#x00A0;");
        }
 
-       assert(NULL == h->metaf);
-       switch (h->metac) {
-       case HTMLFONT_ITALIC:
-               h->metaf = print_otag(h, TAG_I, "");
-               break;
-       case HTMLFONT_BOLD:
-               h->metaf = print_otag(h, TAG_B, "");
-               break;
-       case HTMLFONT_BI:
-               h->metaf = print_otag(h, TAG_B, "");
-               print_otag(h, TAG_I, "");
-               break;
-       case HTMLFONT_CW:
-               h->metaf = print_otag(h, TAG_SPAN, "c", "Li");
-               break;
-       default:
-               print_indent(h);
-               break;
-       }
+       /*
+        * Optionally switch fonts, optionally write a permalink, then
+        * print the text, optionally surrounded by HTML whitespace.
+        */
+
+       assert(h->metaf == NULL);
+       print_metaf(h);
+       print_indent(h);
+
+       if (n != NULL && (href = html_make_id(n, 2)) != NULL) {
+               t = print_otag(h, TAG_A, "chR", "permalink", href);
+               free(href);
+       } else
+               t = NULL;
 
-       assert(word);
        if ( ! print_encode(h, word, NULL, 0)) {
                if ( ! (h->flags & HTML_NONOSPACE))
                        h->flags &= ~HTML_NOSPACE;
@@ -772,10 +951,11 @@ print_text(struct html *h, const char *word)
        } else
                h->flags |= HTML_NOSPACE | HTML_NONEWLINE;
 
-       if (h->metaf) {
+       if (h->metaf != NULL) {
                print_tagq(h, h->metaf);
                h->metaf = NULL;
-       }
+       } else if (t != NULL)
+               print_tagq(h, t);
 
        h->flags &= ~HTML_IGNDELIM;
 }
@@ -783,36 +963,33 @@ print_text(struct html *h, const char *word)
 void
 print_tagq(struct html *h, const struct tag *until)
 {
-       struct tag      *tag;
+       struct tag      *this, *next;
 
-       while ((tag = h->tag) != NULL) {
-               print_ctag(h, tag);
-               if (until && tag == until)
-                       return;
+       for (this = h->tag; this != NULL; this = next) {
+               next = this == until ? NULL : this->next;
+               print_ctag(h, this);
        }
 }
 
+/*
+ * Close out all open elements up to but excluding suntil.
+ * Note that a paragraph just inside stays open together with it
+ * because paragraphs include subsequent phrasing content.
+ */
 void
 print_stagq(struct html *h, const struct tag *suntil)
 {
-       struct tag      *tag;
+       struct tag      *this, *next;
 
-       while ((tag = h->tag) != NULL) {
-               if (suntil && tag == suntil)
-                       return;
-               print_ctag(h, tag);
+       for (this = h->tag; this != NULL; this = next) {
+               next = this->next;
+               if (this == suntil || (next == suntil &&
+                   (this->tag == TAG_P || this->tag == TAG_PRE)))
+                       break;
+               print_ctag(h, this);
        }
 }
 
-void
-print_paragraph(struct html *h)
-{
-       struct tag      *t;
-
-       t = print_otag(h, TAG_DIV, "c", "Pp");
-       print_tagq(h, t);
-}
-
 
 /***********************************************************************
  * Low level output functions.
@@ -905,15 +1082,12 @@ print_indent(struct html *h)
 {
        size_t   i;
 
-       if (h->col)
+       if (h->col || h->noindent)
                return;
 
-       if (h->noindent == 0) {
-               h->col = h->indent * 2;
-               for (i = 0; i < h->col; i++)
-                       putchar(' ');
-       }
-       h->flags &= ~HTML_NOSPACE;
+       h->col = h->indent * 2;
+       for (i = 0; i < h->col; i++)
+               putchar(' ');
 }
 
 /*