===================================================================
RCS file: /cvs/docbook2mdoc/docbook2mdoc.c,v
retrieving revision 1.1.1.1
retrieving revision 1.12
diff -u -p -r1.1.1.1 -r1.12
--- docbook2mdoc/docbook2mdoc.c	2014/03/28 02:04:47	1.1.1.1
+++ docbook2mdoc/docbook2mdoc.c	2014/03/29 22:44:06	1.12
@@ -1,4 +1,4 @@
-/*	$Id: docbook2mdoc.c,v 1.1.1.1 2014/03/28 02:04:47 kristaps Exp $ */
+/*	$Id: docbook2mdoc.c,v 1.12 2014/03/29 22:44:06 kristaps Exp $ */
 /*
  * Copyright (c) 2014 Kristaps Dzonsons <kristaps@bsd.lv>
  *
@@ -24,6 +24,7 @@
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
+#include <unistd.h>
 
 /*
  * All recognised node types.
@@ -31,12 +32,21 @@
 enum	nodeid {
 	NODE_ROOT = 0, /* Must comes first. */
 	/* Alpha-ordered hereafter. */
+	NODE_ARG,
 	NODE_CITEREFENTRY,
+	NODE_CMDSYNOPSIS,
 	NODE_CODE,
+	NODE_COMMAND,
+	NODE_FUNCDEF,
+	NODE_FUNCPROTOTYPE,
 	NODE_FUNCSYNOPSIS,
 	NODE_FUNCSYNOPSISINFO,
+	NODE_FUNCTION,
 	NODE_MANVOLNUM, 
+	NODE_OPTION,
 	NODE_PARA,
+	NODE_PARAMDEF,
+	NODE_PARAMETER,
 	NODE_PROGRAMLISTING,
 	NODE_REFCLASS,
 	NODE_REFDESCRIPTOR,
@@ -49,6 +59,7 @@ enum	nodeid {
 	NODE_REFPURPOSE,
 	NODE_REFSECT1,
 	NODE_REFSYNOPSISDIV,
+	NODE_STRUCTNAME,
 	NODE_SYNOPSIS,
 	NODE_TEXT, 
 	NODE_TITLE,
@@ -56,44 +67,104 @@ enum	nodeid {
 };
 
 /*
+ * All recognised attribute keys.
+ */
+enum	attrkey {
+	/* Alpha-order... */
+	ATTRKEY_CHOICE = 0,
+	ATTRKEY_ID,
+	ATTRKEY_REP,
+	ATTRKEY__MAX
+};
+
+/*
+ * All [explicitly] recognised attribute values.
+ * If an attribute has ATTRVAL__MAX, it could be a free-form.
+ */
+enum	attrval {
+	/* Alpha-order... */
+	ATTRVAL_NOREPEAT,
+	ATTRVAL_OPT,
+	ATTRVAL_PLAIN,
+	ATTRVAL_REPEAT,
+	ATTRVAL_REQ,
+	ATTRVAL__MAX
+};
+
+/*
  * Global parse state.
  * Keep this as simple and small as possible.
  */
 struct	parse {
+	XML_Parser	 xml;
 	enum nodeid	 node; /* current (NODE_ROOT if pre-tree) */
+	const char	*fname; /* filename */
 	int		 stop; /* should we stop now? */
 	struct pnode	*root; /* root of parse tree */
 	struct pnode	*cur; /* current node in tree */
-	char		*b;
-	size_t		 bsz;
-	size_t		 mbsz;
+	char		*b; /* nil-terminated buffer for pre-print */
+	size_t		 bsz; /* current length of b */
+	size_t		 mbsz; /* max bsz allocation */
+	int		 newln; /* output: are we on a fresh line */
 };
 
 struct	node {
-	const char	*name;
+	const char	*name; /* docbook element name */
 	unsigned int	 flags;
 #define	NODE_IGNTEXT	 1 /* ignore all contained text */
 };
 
 TAILQ_HEAD(pnodeq, pnode);
+TAILQ_HEAD(pattrq, pattr);
 
+struct	pattr {
+	enum attrkey	 key;
+	enum attrval	 val;
+	char		*rawval;
+	TAILQ_ENTRY(pattr) child;
+};
+
 struct	pnode {
 	enum nodeid	 node; /* node type */
 	char		*b; /* binary data buffer */
 	size_t		 bsz; /* data buffer size */
 	struct pnode	*parent; /* parent (or NULL if top) */
 	struct pnodeq	 childq; /* queue of children */
+	struct pattrq	 attrq; /* attributes of node */
 	TAILQ_ENTRY(pnode) child;
 };
 
+static	const char *attrkeys[ATTRKEY__MAX] = {
+	"choice",
+	"id",
+	"rep"
+};
+
+static	const char *attrvals[ATTRVAL__MAX] = {
+	"norepeat",
+	"opt",
+	"plain",
+	"repeat",
+	"req"
+};
+
 static	const struct node nodes[NODE__MAX] = {
 	{ NULL, 0 }, 
+	{ "arg", 0 }, 
 	{ "citerefentry", NODE_IGNTEXT }, 
+	{ "cmdsynopsis", NODE_IGNTEXT }, 
 	{ "code", 0 }, 
+	{ "command", 0 }, 
+	{ "funcdef", 0 },
+	{ "funcprototype", NODE_IGNTEXT }, 
 	{ "funcsynopsis", NODE_IGNTEXT }, 
 	{ "funcsynopsisinfo", 0 }, 
+	{ "function", 0 }, 
 	{ "manvolnum", 0 }, 
+	{ "option", 0 }, 
 	{ "para", 0 }, 
+	{ "paramdef", 0 },
+	{ "parameter", 0 }, 
 	{ "programlisting", 0 }, 
 	{ "refclass", NODE_IGNTEXT }, 
 	{ "refdescriptor", NODE_IGNTEXT }, 
@@ -106,13 +177,57 @@ static	const struct node nodes[NODE__MAX] = {
 	{ "refpurpose", 0 }, 
 	{ "refsect1", 0 }, 
 	{ "refsynopsisdiv", NODE_IGNTEXT }, 
+	{ "structname", 0 }, 
 	{ "synopsis", 0 }, 
 	{ NULL, 0 }, 
 	{ "title", 0 }, 
 };
 
+static void
+pnode_print(struct parse *p, struct pnode *pn);
+
+static int
+isattrkey(enum nodeid node, enum attrkey key)
+{
+
+	switch (key) {
+	case (ATTRKEY_CHOICE):
+		return(node == NODE_ARG);
+	case (ATTRKEY_ID):
+		/* Common to all. */
+		return(1);
+	case (ATTRKEY_REP):
+		return(node == NODE_ARG);
+	default:
+		break;
+	}
+	abort();
+	return(0);
+}
+
+static int
+isattrval(enum attrkey key, enum attrval val)
+{
+
+	switch (val) {
+	case (ATTRVAL_OPT):
+	case (ATTRVAL_PLAIN):
+	case (ATTRVAL_REQ):
+		return(key == ATTRKEY_CHOICE);
+	case (ATTRVAL_REPEAT):
+	case (ATTRVAL_NOREPEAT):
+		return(key == ATTRKEY_REP);
+	default:
+		break;
+	}
+	abort();
+	return(0);
+}
+
 /*
  * Look up whether "parent" is a valid parent for "node".
+ * This is sucked directly from the DocBook specification: look at the
+ * "children" and "parent" sections of each node.
  */
 static int
 isparent(enum nodeid node, enum nodeid parent)
@@ -121,6 +236,15 @@ isparent(enum nodeid node, enum nodeid parent)
 	switch (node) {
 	case (NODE_ROOT):
 		return(0);
+	case (NODE_ARG):
+		switch (parent) {
+		case (NODE_ARG):
+		case (NODE_CMDSYNOPSIS):
+			return(1);
+		default:
+			break;
+		}
+		return(0);
 	case (NODE_CITEREFENTRY):
 		switch (parent) {
 		case (NODE_FUNCSYNOPSISINFO):
@@ -137,6 +261,16 @@ isparent(enum nodeid node, enum nodeid parent)
 			break;
 		}
 		return(0);
+	case (NODE_CMDSYNOPSIS):
+		switch (parent) {
+		case (NODE_PARA):
+		case (NODE_REFSECT1):
+		case (NODE_REFSYNOPSISDIV):
+			return(1);
+		default:
+			break;
+		}
+		return(0);
 	case (NODE_CODE):
 		switch (parent) {
 		case (NODE_FUNCSYNOPSISINFO):
@@ -153,15 +287,27 @@ isparent(enum nodeid node, enum nodeid parent)
 			break;
 		}
 		return(0);
-	case (NODE_MANVOLNUM):
+	case (NODE_COMMAND):
 		switch (parent) {
-		case (NODE_CITEREFENTRY):
-		case (NODE_REFMETA):
+		case (NODE_CMDSYNOPSIS):
+		case (NODE_FUNCSYNOPSISINFO):
+		case (NODE_PARA):
+		case (NODE_PROGRAMLISTING):
+		case (NODE_REFDESCRIPTOR):
+		case (NODE_REFENTRYTITLE):
+		case (NODE_REFNAME):
+		case (NODE_REFPURPOSE):
+		case (NODE_SYNOPSIS):
+		case (NODE_TITLE):
 			return(1);
 		default:
 			break;
 		}
 		return(0);
+	case (NODE_FUNCDEF):
+		return(NODE_FUNCPROTOTYPE == parent);
+	case (NODE_FUNCPROTOTYPE):
+		return(NODE_FUNCSYNOPSIS == parent);
 	case (NODE_FUNCSYNOPSIS):
 		switch (parent) {
 		case (NODE_PARA):
@@ -174,6 +320,50 @@ isparent(enum nodeid node, enum nodeid parent)
 		return(0);
 	case (NODE_FUNCSYNOPSISINFO):
 		return(NODE_FUNCSYNOPSIS == parent);
+	case (NODE_FUNCTION):
+		switch (parent) {
+		case (NODE_CODE):
+		case (NODE_FUNCDEF):
+		case (NODE_FUNCSYNOPSISINFO):
+		case (NODE_PARA):
+		case (NODE_PROGRAMLISTING):
+		case (NODE_REFDESCRIPTOR):
+		case (NODE_REFENTRYTITLE):
+		case (NODE_REFNAME):
+		case (NODE_REFPURPOSE):
+		case (NODE_SYNOPSIS):
+		case (NODE_TITLE):
+			return(1);
+		default:
+			break;
+		}
+		return(0);
+	case (NODE_MANVOLNUM):
+		switch (parent) {
+		case (NODE_CITEREFENTRY):
+		case (NODE_REFMETA):
+			return(1);
+		default:
+			break;
+		}
+		return(0);
+	case (NODE_OPTION):
+		switch (parent) {
+		case (NODE_ARG):
+		case (NODE_FUNCSYNOPSISINFO):
+		case (NODE_PARA):
+		case (NODE_PROGRAMLISTING):
+		case (NODE_REFDESCRIPTOR):
+		case (NODE_REFENTRYTITLE):
+		case (NODE_REFNAME):
+		case (NODE_REFPURPOSE):
+		case (NODE_SYNOPSIS):
+		case (NODE_TITLE):
+			return(1);
+		default:
+			break;
+		}
+		return(0);
 	case (NODE_PARA):
 		switch (parent) {
 		case (NODE_REFSECT1):
@@ -183,6 +373,26 @@ isparent(enum nodeid node, enum nodeid parent)
 			break;
 		}
 		return(0);
+	case (NODE_PARAMDEF):
+		return(NODE_FUNCPROTOTYPE == parent);
+	case (NODE_PARAMETER):
+		switch (parent) {
+		case (NODE_CODE):
+		case (NODE_FUNCSYNOPSISINFO):
+		case (NODE_PARA):
+		case (NODE_PARAMDEF):
+		case (NODE_PROGRAMLISTING):
+		case (NODE_REFDESCRIPTOR):
+		case (NODE_REFENTRYTITLE):
+		case (NODE_REFNAME):
+		case (NODE_REFPURPOSE):
+		case (NODE_SYNOPSIS):
+		case (NODE_TITLE):
+			return(1);
+		default:
+			break;
+		}
+		return(0);
 	case (NODE_PROGRAMLISTING):
 		switch (parent) {
 		case (NODE_PARA):
@@ -221,6 +431,26 @@ isparent(enum nodeid node, enum nodeid parent)
 		return(parent == NODE_REFENTRY);
 	case (NODE_REFSYNOPSISDIV):
 		return(parent == NODE_REFENTRY);
+	case (NODE_STRUCTNAME):
+		switch (parent) {
+		case (NODE_CODE):
+		case (NODE_FUNCSYNOPSISINFO):
+		case (NODE_FUNCTION):
+		case (NODE_OPTION):
+		case (NODE_PARA):
+		case (NODE_PARAMETER):
+		case (NODE_PROGRAMLISTING):
+		case (NODE_REFDESCRIPTOR):
+		case (NODE_REFENTRYTITLE):
+		case (NODE_REFNAME):
+		case (NODE_REFPURPOSE):
+		case (NODE_SYNOPSIS):
+		case (NODE_TITLE):
+			return(1);
+		default:
+			break;
+		}
+		return(0);
 	case (NODE_SYNOPSIS):
 		switch (parent) {
 		case (NODE_REFSYNOPSISDIV):
@@ -249,11 +479,18 @@ isparent(enum nodeid node, enum nodeid parent)
 	return(0);
 }
 
+/*
+ * Process a stream of characters.
+ * We store text as nodes in and of themselves.
+ * If a text node is already open, append to it.
+ * If it's not open, open one under the current context.
+ */
 static void
 xml_char(void *arg, const XML_Char *p, int sz)
 {
 	struct parse	*ps = arg;
 	struct pnode	*dat;
+	int		 i;
 
 	/* Stopped or no tree yet. */
 	if (ps->stop || NODE_ROOT == ps->node)
@@ -268,8 +505,17 @@ xml_char(void *arg, const XML_Char *p, int sz)
 	 * Are we in the midst of processing text?
 	 * If we're not processing text right now, then create a text
 	 * node for doing so.
+	 * However, don't do so unless we have some non-whitespace to
+	 * process: strip out all leading whitespace to be sure.
 	 */
 	if (NODE_TEXT != ps->node) {
+		for (i = 0; i < sz; i++)
+			if ( ! isspace((int)p[i]))
+				break;
+		if (i == sz)
+			return;
+		p += i;
+		sz -= i;
 		dat = calloc(1, sizeof(struct pnode));
 		if (NULL == dat) {
 			perror(NULL);
@@ -279,10 +525,10 @@ xml_char(void *arg, const XML_Char *p, int sz)
 		dat->node = ps->node = NODE_TEXT;
 		dat->parent = ps->cur;
 		TAILQ_INIT(&dat->childq);
+		TAILQ_INIT(&dat->attrq);
 		TAILQ_INSERT_TAIL(&ps->cur->childq, dat, child);
 		ps->cur = dat;
 		assert(NULL != ps->root);
-
 	}
 
 	/* Append to current buffer. */
@@ -297,23 +543,40 @@ xml_char(void *arg, const XML_Char *p, int sz)
 	ps->cur->bsz += (size_t)sz;
 }
 
+static void
+pnode_trim(struct pnode *pn)
+{
+
+	assert(NODE_TEXT == pn->node);
+	for ( ; pn->bsz > 0; pn->bsz--)
+		if ( ! isspace((int)pn->b[pn->bsz - 1]))
+			break;
+}
+
 /*
  * Begin an element.
  * First, look for the element.
  * If we don't find it and we're not parsing, keep going.
- * If we don't find it (and we're parsing), puke and exit.
+ * If we don't find it and we're parsing, puke and exit.
  * If we find it but we're not parsing yet (i.e., it's not a refentry
  * and thus out of context), keep going.
- * If we're at the root and already have a tree, puke and exit.
+ * If we find it and we're at the root and already have a tree, puke and
+ * exit (FIXME: I don't think this is right?).
+ * If we find it but we're parsing a text node, close out the text node,
+ * return to its parent, and keep going.
  * Make sure that the element is in the right context.
  * Lastly, put the node onto our parse tree and continue.
  */
 static void
 xml_elem_start(void *arg, const XML_Char *name, const XML_Char **atts)
 {
-	struct parse	*ps = arg;
-	enum nodeid	 node;
-	struct pnode	*dat;
+	struct parse	 *ps = arg;
+	enum nodeid	  node;
+	enum attrkey	  key;
+	enum attrval	  val;
+	struct pnode	 *dat;
+	struct pattr	 *pattr;
+	const XML_Char	**att;
 
 	if (ps->stop)
 		return;
@@ -321,6 +584,7 @@ xml_elem_start(void *arg, const XML_Char *name, const 
 	/* Close out text node, if applicable... */
 	if (NODE_TEXT == ps->node) {
 		assert(NULL != ps->cur);
+		pnode_trim(ps->cur);
 		ps->cur = ps->cur->parent;
 		assert(NULL != ps->cur);
 		ps->node = ps->cur->node;
@@ -333,21 +597,27 @@ xml_elem_start(void *arg, const XML_Char *name, const 
 			break;
 
 	if (NODE__MAX == node && NODE_ROOT == ps->node) {
-		fprintf(stderr, "%s: ignoring node\n", name);
 		return;
 	} else if (NODE__MAX == node) {
-		fprintf(stderr, "%s: unknown node\n", name);
+		fprintf(stderr, "%s:%zu:%zu: unknown node \"%s\"\n", 
+			ps->fname, XML_GetCurrentLineNumber(ps->xml),
+			XML_GetCurrentColumnNumber(ps->xml), name);
 		ps->stop = 1;
 		return;
 	} else if (NODE_ROOT == ps->node && NULL != ps->root) {
-		fprintf(stderr, "%s: reentering?\n", name);
+		fprintf(stderr, "%s:%zu:%zu: multiple refentries\n",
+			ps->fname, XML_GetCurrentLineNumber(ps->xml),
+			XML_GetCurrentColumnNumber(ps->xml));
 		ps->stop = 1;
 		return;
 	} else if (NODE_ROOT == ps->node && NODE_REFENTRY != node) {
-		fprintf(stderr, "%s: known node w/o context\n", name);
 		return;
 	} else if ( ! isparent(node, ps->node)) {
-		fprintf(stderr, "%s: bad parent\n", name);
+		fprintf(stderr, "%s:%zu:%zu: bad parent \"%s\"\n",
+			ps->fname, XML_GetCurrentLineNumber(ps->xml),
+			XML_GetCurrentColumnNumber(ps->xml),
+			NULL == nodes[ps->node].name ?
+			"(none)" : nodes[ps->node].name);
 		ps->stop = 1;
 		return;
 	}
@@ -360,6 +630,7 @@ xml_elem_start(void *arg, const XML_Char *name, const 
 	dat->node = ps->node = node;
 	dat->parent = ps->cur;
 	TAILQ_INIT(&dat->childq);
+	TAILQ_INIT(&dat->attrq);
 
 	if (NULL != ps->cur)
 		TAILQ_INSERT_TAIL(&ps->cur->childq, dat, child);
@@ -367,11 +638,53 @@ xml_elem_start(void *arg, const XML_Char *name, const 
 	ps->cur = dat;
 	if (NULL == ps->root)
 		ps->root = dat;
+
+	/*
+	 * Process attributes.
+	 */
+	for (att = atts; NULL != *att; att += 2) {
+		for (key = 0; key < ATTRKEY__MAX; key++)
+			if (0 == strcmp(*att, attrkeys[key]))
+				break;
+		if (ATTRKEY__MAX == key) {
+			fprintf(stderr, "%s:%zu:%zu: unknown "
+				"attribute \"%s\"\n", ps->fname, 
+				XML_GetCurrentLineNumber(ps->xml),
+				XML_GetCurrentColumnNumber(ps->xml), 
+				*att);
+			continue;
+		} else if ( ! isattrkey(node, key)) {
+			fprintf(stderr, "%s:%zu:%zu: bad "
+				"attribute \"%s\"\n", ps->fname, 
+				XML_GetCurrentLineNumber(ps->xml),
+				XML_GetCurrentColumnNumber(ps->xml), 
+				*att);
+			continue;
+		}
+		for (val = 0; val < ATTRVAL__MAX; val++)
+			if (0 == strcmp(*(att + 1), attrvals[val]))
+				break;
+		if (ATTRVAL__MAX != val && ! isattrval(key, val)) {
+			fprintf(stderr, "%s:%zu:%zu: bad "
+				"value \"%s\"\n", ps->fname, 
+				XML_GetCurrentLineNumber(ps->xml),
+				XML_GetCurrentColumnNumber(ps->xml), 
+				*(att + 1));
+			continue;
+		}
+		pattr = calloc(1, sizeof(struct pattr));
+		pattr->key = key;
+		pattr->val = val;
+		if (ATTRVAL__MAX == val)
+			pattr->rawval = strdup(*(att + 1));
+		TAILQ_INSERT_TAIL(&dat->attrq, pattr, child);
+	}
+
 }
 
 /*
  * Roll up the parse tree.
- * Does nothing else special.
+ * If we're at a text node, roll that one up first.
  * If we hit the root, then assign ourselves as the NODE_ROOT.
  */
 static void
@@ -385,6 +698,7 @@ xml_elem_end(void *arg, const XML_Char *name)
 	/* Close out text node, if applicable... */
 	if (NODE_TEXT == ps->node) {
 		assert(NULL != ps->cur);
+		pnode_trim(ps->cur);
 		ps->cur = ps->cur->parent;
 		assert(NULL != ps->cur);
 		ps->node = ps->cur->node;
@@ -396,10 +710,14 @@ xml_elem_end(void *arg, const XML_Char *name)
 		ps->node = ps->cur->node;
 }
 
+/*
+ * Recursively free a node (NULL is ok).
+ */
 static void
 pnode_free(struct pnode *pn)
 {
 	struct pnode	*pp;
+	struct pattr	*ap;
 
 	if (NULL == pn)
 		return;
@@ -409,10 +727,19 @@ pnode_free(struct pnode *pn)
 		pnode_free(pp);
 	}
 
+	while (NULL != (ap = TAILQ_FIRST(&pn->attrq))) {
+		TAILQ_REMOVE(&pn->attrq, ap, child);
+		free(ap->rawval);
+		free(ap);
+	}
+
 	free(pn->b);
 	free(pn);
 }
 
+/*
+ * Unlink a node from its parent and pnode_free() it.
+ */
 static void
 pnode_unlink(struct pnode *pn)
 {
@@ -422,13 +749,32 @@ pnode_unlink(struct pnode *pn)
 	pnode_free(pn);
 }
 
+/*
+ * Unlink all children of a node and pnode_free() them.
+ */
 static void
+pnode_unlinksub(struct pnode *pn)
+{
+
+	while ( ! TAILQ_EMPTY(&pn->childq))
+		pnode_unlink(TAILQ_FIRST(&pn->childq));
+}
+
+/*
+ * Reset the lookaside buffer.
+ */
+static void
 bufclear(struct parse *p)
 {
 
 	p->b[p->bsz = 0] = '\0';
 }
 
+/*
+ * Append NODE_TEXT contents to the current buffer, reallocating its
+ * size if necessary.
+ * The buffer is ALWAYS nil-terminated.
+ */
 static void
 bufappend(struct parse *p, struct pnode *pn)
 {
@@ -447,22 +793,35 @@ bufappend(struct parse *p, struct pnode *pn)
 }
 
 /*
- * Print text presumably on a macro line.
- * Ignore any child macros.
- * Convert all whitespace to regular spaces.
+ * Recursively append all NODE_TEXT nodes to the buffer.
+ * This descends into non-text nodes, but doesn't do anything beyond
+ * them.
+ * In other words, this is a recursive text grok.
  */
 static void
-pnode_printmacrolinepart(struct parse *p, struct pnode *pn)
+bufappend_r(struct parse *p, struct pnode *pn)
 {
 	struct pnode	*pp;
+
+	if (NODE_TEXT == pn->node)
+		bufappend(p, pn);
+	TAILQ_FOREACH(pp, &pn->childq, child)
+		bufappend_r(p, pp);
+}
+
+#define MACROLINE_NORM	0
+#define MACROLINE_UPPER	1
+/*
+ * Recursively print text presumably on a macro line.
+ * Convert all whitespace to regular spaces.
+ */
+static void
+pnode_printmacrolinetext(struct parse *p, struct pnode *pn, int fl)
+{
 	char		*cp;
 
 	bufclear(p);
-	while (NULL != (pp = TAILQ_FIRST(&pn->childq))) {
-		if (NODE_TEXT == pp->node)
-			bufappend(p, pp);
-		pnode_unlink(pp);
-	}
+	bufappend_r(p, pn);
 
 	/* Convert all space to spaces. */
 	for (cp = p->b; '\0' != *cp; cp++)
@@ -470,8 +829,7 @@ pnode_printmacrolinepart(struct parse *p, struct pnode
 			*cp = ' ';
 
 	for (cp = p->b; isspace((int)*cp); cp++)
-		/* Spin. */ ;
-
+		/* Spin past whitespace (XXX: necessary?) */ ;
 	for ( ; '\0' != *cp; cp++) {
 		/* Escape us if we look like a macro. */
 		if ((cp == p->b || ' ' == *(cp - 1)) &&
@@ -484,13 +842,23 @@ pnode_printmacrolinepart(struct parse *p, struct pnode
 			  ('\0' == *(cp + 3) || 
 			   ' ' == *(cp + 3)))))
 			fputs("\\&", stdout);
-		putchar(*cp);
+		if (MACROLINE_UPPER & fl)
+			putchar(toupper((int)*cp));
+		else
+			putchar((int)*cp);
 		/* If we're a character escape, escape us. */
 		if ('\\' == *cp)
 			putchar('e');
 	}
 }
 
+static void
+pnode_printmacrolinepart(struct parse *p, struct pnode *pn)
+{
+
+	pnode_printmacrolinetext(p, pn, 0);
+}
+
 /*
  * Just pnode_printmacrolinepart() but with a newline.
  * If no text, just the newline.
@@ -499,11 +867,49 @@ static void
 pnode_printmacroline(struct parse *p, struct pnode *pn)
 {
 
-	pnode_printmacrolinepart(p, pn);
+	pnode_printmacrolinetext(p, pn, 0);
 	putchar('\n');
 }
 
 static void
+pnode_printmopen(struct parse *p)
+{
+	if (p->newln) {
+		putchar('.');
+		p->newln = 0;
+	} else
+		putchar(' ');
+}
+
+static void
+pnode_printmclose(struct parse *p, int sv)
+{
+
+	if (sv && ! p->newln) {
+		putchar('\n');
+		p->newln = 1;
+	}
+}
+
+/*
+ * If the SYNOPSIS macro has a superfluous title, kill it.
+ */
+static void
+pnode_printrefsynopsisdiv(struct parse *p, struct pnode *pn)
+{
+	struct pnode	*pp;
+
+	TAILQ_FOREACH(pp, &pn->childq, child) 
+		if (NODE_TITLE == pp->node) {
+			pnode_unlink(pp);
+			return;
+		}
+}
+
+/*
+ * Start a hopefully-named `Sh' section.
+ */
+static void
 pnode_printrefsect(struct parse *p, struct pnode *pn)
 {
 	struct pnode	*pp;
@@ -512,14 +918,18 @@ pnode_printrefsect(struct parse *p, struct pnode *pn)
 		if (NODE_TITLE == pp->node)
 			break;
 
+	fputs(".Sh ", stdout);
+
 	if (NULL != pp) {
-		fputs(".Sh ", stdout);
 		pnode_printmacroline(p, pp);
 		pnode_unlink(pp);
 	} else
-		puts(".Sh UNKNOWN");
+		puts("UNKNOWN");
 }
 
+/*
+ * Start a reference, extracting the title and volume.
+ */
 static void
 pnode_printciterefentry(struct parse *p, struct pnode *pn)
 {
@@ -533,16 +943,16 @@ pnode_printciterefentry(struct parse *p, struct pnode 
 			title = pp;
 
 	fputs(".Xr ", stdout);
+
 	if (NULL != title) {
 		pnode_printmacrolinepart(p, title);
-		pnode_unlink(title);
+		putchar(' ');
 	} else
-		fputs("unknown", stdout);
-	putchar(' ');
-	if (NULL != manvol) {
+		fputs("unknown ", stdout);
+
+	if (NULL != manvol)
 		pnode_printmacroline(p, manvol);
-		pnode_unlink(manvol);
-	} else
+	else
 		puts("1");
 }
 
@@ -558,24 +968,172 @@ pnode_printrefmeta(struct parse *p, struct pnode *pn)
 		else if (NODE_REFENTRYTITLE == pp->node)
 			title = pp;
 
-	puts(".Dd $Mdocdate: March 28 2014 $");
+	puts(".Dd $Mdocdate" "$");
 	fputs(".Dt ", stdout);
 
 	if (NULL != title) {
-		pnode_printmacrolinepart(p, title);
-		pnode_unlink(title);
+		/* FIXME: uppercase. */
+		pnode_printmacrolinetext(p, title, MACROLINE_UPPER);
+		putchar(' ');
 	} else
-		fputs("UNKNOWN", stdout);
-	putchar(' ');
-	if (NULL != manvol) {
+		fputs("UNKNOWN ", stdout);
+
+	if (NULL != manvol)
 		pnode_printmacroline(p, manvol);
-		pnode_unlink(manvol);
-	} else
+	else
 		puts("1");
 
 	puts(".Os");
 }
 
+static void
+pnode_printfuncdef(struct parse *p, struct pnode *pn)
+{
+	struct pnode	*pp, *ftype, *func;
+
+	ftype = func = NULL;
+	TAILQ_FOREACH(pp, &pn->childq, child)
+		if (NODE_TEXT == pp->node)
+			ftype = pp;
+		else if (NODE_FUNCTION == pp->node)
+			func = pp;
+
+	if (NULL != ftype) {
+		fputs(".Ft ", stdout);
+		pnode_printmacroline(p, ftype);
+	}
+
+	if (NULL != func) {
+		fputs(".Fo ", stdout);
+		pnode_printmacroline(p, func);
+	} else 
+		puts(".Fo UNKNOWN");
+}
+
+static void
+pnode_printparamdef(struct parse *p, struct pnode *pn)
+{
+	struct pnode	*pp, *ptype, *param;
+
+	ptype = param = NULL;
+	TAILQ_FOREACH(pp, &pn->childq, child)
+		if (NODE_TEXT == pp->node)
+			ptype = pp;
+		else if (NODE_PARAMETER == pp->node)
+			param = pp;
+
+	fputs(".Fa \"", stdout);
+	if (NULL != ptype) {
+		pnode_printmacrolinepart(p, ptype);
+		putchar(' ');
+	}
+
+	if (NULL != param)
+		pnode_printmacrolinepart(p, param);
+
+	puts("\"");
+}
+
+static void
+pnode_printfuncprototype(struct parse *p, struct pnode *pn)
+{
+	struct pnode	*pp, *fdef;
+
+	TAILQ_FOREACH(fdef, &pn->childq, child)
+		if (NODE_FUNCDEF == fdef->node) 
+			break;
+
+	if (NULL != fdef)
+		pnode_printfuncdef(p, fdef);
+	else
+		puts(".Fo UNKNOWN");
+
+	TAILQ_FOREACH(pp, &pn->childq, child) 
+		if (NODE_PARAMDEF == pp->node)
+			pnode_printparamdef(p, pp);
+
+	puts(".Fc");
+}
+
+/* 
+ * The <arg> element is more complicated than it should be because text
+ * nodes are treated like ".Ar foo", but non-text nodes need to be
+ * re-sent into the printer (i.e., without the preceding ".Ar").
+ * This also handles the case of "repetition" (or in other words, the
+ * ellipsis following an argument) and optionality.
+ */
+static void
+pnode_printarg(struct parse *p, struct pnode *pn)
+{
+	struct pnode	*pp;
+	struct pattr	*ap;
+	int		 isop, isrep;
+
+	isop = 1;
+	isrep = 0;
+	TAILQ_FOREACH(ap, &pn->attrq, child) 
+		if (ATTRKEY_CHOICE == ap->key &&
+			(ATTRVAL_PLAIN == ap->val ||
+			 ATTRVAL_REQ == ap->val)) 
+			isop = 0;
+		else if (ATTRKEY_REP == ap->key &&
+			(ATTRVAL_REPEAT == ap->val))
+			isrep = 1;
+
+	if (isop) {
+		pnode_printmopen(p);
+		fputs("Op ", stdout);
+	}
+
+	TAILQ_FOREACH(pp, &pn->childq, child) {
+		if (NODE_TEXT == pp->node) {
+			pnode_printmopen(p);
+			fputs("Ar ", stdout);
+		} 
+		pnode_print(p, pp);
+		if (NODE_TEXT == pp->node && isrep) 
+			fputs("...", stdout);
+	}
+}
+
+/* 
+ * Recursively search and return the first instance of "node".
+ */
+static struct pnode *
+pnode_findfirst(struct pnode *pn, enum nodeid node)
+{
+	struct pnode	*pp, *res;
+
+	res = NULL;
+	TAILQ_FOREACH(pp, &pn->childq, child) {
+		res = pp->node == node ? pp :
+			pnode_findfirst(pp, node);
+		if (NULL != res)
+			break;
+	}
+
+	return(res);
+}
+
+static void
+pnode_printprologue(struct parse *p, struct pnode *pn)
+{
+	struct pnode	*pp;
+
+	pp = NULL == p->root ? NULL :
+		pnode_findfirst(p->root, NODE_REFMETA);
+
+	if (NULL != pp) {
+		pnode_printrefmeta(p, pp);
+		pnode_unlink(pp);
+	} else {
+		puts(".\\\" Supplying bogus prologue...");
+		puts(".Dd $Mdocdate" "$");
+		puts(".Dt UNKNOWN 1");
+		puts(".Os");
+	}
+}
+
 /*
  * Print a parsed node (or ignore it--whatever).
  * This is a recursive function.
@@ -586,53 +1144,95 @@ pnode_print(struct parse *p, struct pnode *pn)
 {
 	struct pnode	*pp;
 	char		*cp;
-	int		 last;
+	int		 last, sv;
 
 	if (NULL == pn)
 		return;
 
-	if (NODE_TEXT != pn->node && NODE_ROOT != pn->node)
-		printf(".\\\" %s\n", nodes[pn->node].name);
+	sv = p->newln;
 
 	switch (pn->node) {
+	case (NODE_ARG):
+		pnode_printarg(p, pn);
+		pnode_unlinksub(pn);
+		break;
 	case (NODE_CITEREFENTRY):
+		assert(p->newln);
 		pnode_printciterefentry(p, pn);
+		pnode_unlinksub(pn);
 		break;
 	case (NODE_CODE):
-		fputs(".Li ", stdout);
-		pnode_printmacroline(p, pn);
+		pnode_printmopen(p);
+		fputs("Li ", stdout);
 		break;
+	case (NODE_COMMAND):
+		pnode_printmopen(p);
+		fputs("Nm ", stdout);
+		break;
+	case (NODE_FUNCTION):
+		pnode_printmopen(p);
+		fputs("Fn ", stdout);
+		break;
+	case (NODE_FUNCPROTOTYPE):
+		assert(p->newln);
+		pnode_printfuncprototype(p, pn);
+		pnode_unlinksub(pn);
+		break;
 	case (NODE_FUNCSYNOPSISINFO):
-		fputs(".Fd ", stdout);
-		pnode_printmacroline(p, pn);
+		pnode_printmopen(p);
+		fputs("Fd ", stdout);
 		break;
+	case (NODE_OPTION):
+		pnode_printmopen(p);
+		fputs("Fl ", stdout);
+		break;
 	case (NODE_PARA):
-		/* FIXME: not always. */
+		assert(p->newln);
 		puts(".Pp");
 		break;
+	case (NODE_PARAMETER):
+		/* Suppress non-text children... */
+		pnode_printmopen(p);
+		fputs("Fa \"", stdout);
+		pnode_printmacrolinepart(p, pn);
+		puts("\"");
+		pnode_unlinksub(pn);
+		break;
 	case (NODE_PROGRAMLISTING):
+		assert(p->newln);
 		puts(".Bd -literal");
 		break;
 	case (NODE_REFMETA):
-		pnode_printrefmeta(p, pn);
+		abort();
 		break;
 	case (NODE_REFNAME):
-		fputs(".Nm ", stdout);
-		pnode_printmacroline(p, pn);
-		return;
+		/* Suppress non-text children... */
+		pnode_printmopen(p);
+		fputs("Nm ", stdout);
+		pnode_printmacrolinepart(p, pn);
+		pnode_unlinksub(pn);
+		break;
 	case (NODE_REFNAMEDIV):
+		assert(p->newln);
 		puts(".Sh NAME");
 		break;
 	case (NODE_REFPURPOSE):
+		assert(p->newln);
 		fputs(".Nd ", stdout);
-		pnode_printmacroline(p, pn);
-		return;
+		break;
 	case (NODE_REFSYNOPSISDIV):
+		assert(p->newln);
+		pnode_printrefsynopsisdiv(p, pn);
 		puts(".Sh SYNOPSIS");
 		break;
 	case (NODE_REFSECT1):
+		assert(p->newln);
 		pnode_printrefsect(p, pn);
 		break;
+	case (NODE_STRUCTNAME):
+		pnode_printmopen(p);
+		fputs("Vt ", stdout);
+		break;
 	case (NODE_TEXT):
 		bufclear(p);
 		bufappend(p, pn);
@@ -642,6 +1242,7 @@ pnode_print(struct parse *p, struct pnode *pn)
 		 * XXX: all whitespace, including tabs (?).
 		 * Remember to escape control characters and escapes.
 		 */
+		assert(p->bsz);
 		for (last = '\n', cp = p->b; '\0' != *cp; ) {
 			if ('\n' == last) {
 				/* Consume all whitespace. */
@@ -657,8 +1258,7 @@ pnode_print(struct parse *p, struct pnode *pn)
 			if ('\\' == last)
 				putchar('e');
 		}
-		if ('\n' != last)
-			putchar('\n');
+		p->newln = 0;
 		break;
 	default:
 		break;
@@ -668,8 +1268,34 @@ pnode_print(struct parse *p, struct pnode *pn)
 		pnode_print(p, pp);
 
 	switch (pn->node) {
+	case (NODE_ARG):
+	case (NODE_CODE):
+	case (NODE_COMMAND):
+	case (NODE_FUNCTION):
+	case (NODE_FUNCSYNOPSISINFO):
+	case (NODE_OPTION):
+	case (NODE_PARAMETER):
+	case (NODE_STRUCTNAME):
+	case (NODE_TEXT):
+		pnode_printmclose(p, sv);
+		break;
+	case (NODE_REFNAME):
+		/*
+		 * If we're in the NAME macro and we have multiple
+		 * <refname> macros in sequence, then print out a
+		 * trailing comma before the newline.
+		 */
+		if (NULL != pn->parent && 
+			NODE_REFNAMEDIV == pn->parent->node &&
+			NULL != TAILQ_NEXT(pn, child) &&
+			NODE_REFNAME == TAILQ_NEXT(pn, child)->node) 
+			fputs(" ,", stdout);
+		pnode_printmclose(p, sv);
+		break;
 	case (NODE_PROGRAMLISTING):
+		assert(p->newln);
 		puts(".Ed");
+		p->newln = 1;
 		break;
 	default:
 		break;
@@ -691,6 +1317,8 @@ readfile(XML_Parser xp, int fd, 
 	memset(&p, 0, sizeof(struct parse));
 
 	p.b = malloc(p.bsz = p.mbsz = 1024);
+	p.fname = fn;
+	p.xml = xp;
 
 	XML_SetCharacterDataHandler(xp, xml_char);
 	XML_SetElementHandler(xp, xml_elem_start, xml_elem_end);
@@ -707,6 +1335,8 @@ readfile(XML_Parser xp, int fd, 
 		 * Exit when we've read all or errors have occured
 		 * during the parse sequence.
 		 */
+		p.newln = 1;
+		pnode_printprologue(&p, p.root);
 		pnode_print(&p, p.root);
 		pnode_free(p.root);
 		free(p.b);