===================================================================
RCS file: /cvs/pod2mdoc/pod2mdoc.c,v
retrieving revision 1.18
retrieving revision 1.28
diff -u -p -r1.18 -r1.28
--- pod2mdoc/pod2mdoc.c	2014/04/02 22:36:56	1.18
+++ pod2mdoc/pod2mdoc.c	2014/07/11 10:31:28	1.28
@@ -1,4 +1,4 @@
-/*	$Id: pod2mdoc.c,v 1.18 2014/04/02 22:36:56 kristaps Exp $ */
+/*	$Id: pod2mdoc.c,v 1.28 2014/07/11 10:31:28 schwarze Exp $ */
 /*
  * Copyright (c) 2014 Kristaps Dzonsons <kristaps@bsd.lv>
  *
@@ -27,7 +27,9 @@
 #include <unistd.h>
 
 /*
- * In what section can we find Perl manuals?
+ * In what section can we find Perl module manuals?
+ * Sometimes (Mac OS X) it's 3pm, sometimes (OpenBSD, etc.) 3p.
+ * XXX IF YOU CHANGE THIS, CHANGE POD2MDOC.1 AS WELL.
  */
 #define	PERL_SECTION	"3p"
 
@@ -170,85 +172,134 @@ formatescape(const char *buf, size_t *start, size_t en
 
 /*
  * Run some heuristics to intuit a link format.
- * I recognise L<foo::bar> as a Perl manpage, printing it in section 3p;
- * or a general UNIX foo(5) manpage.
- * If I recognise one, I set "start" to be the end of the sequence so
+ * I set "start" to be the end of the sequence (last right-carrot) so
  * that the caller can safely just continue processing.
- * Otherwise, I don't touch "start".
+ * If this is just an empty tag, I'll return 0.
  */
 static int
 trylink(const char *buf, size_t *start, size_t end, size_t dsz)
 {
-	size_t		 sv, nstart, nend, i, j, textsz;
-	const char	*text;
-	int		 hasdouble;
+	size_t		 linkstart, realend, linkend, 
+			 i, j, textsz, stack;
 
 	/* 
 	 * Scan to the start of the terminus. 
 	 * This function is more or less replicated in the formatcode()
 	 * for null or index formatting codes.
+	 * However, we're slightly different because we might have
+	 * nested escapes we need to ignore.
 	 */
-	for (sv = nstart = *start; nstart < end; nstart++) {
-		if ('>' != buf[nstart])
+	stack = 0;
+	for (linkstart = realend = *start; realend < end; realend++) {
+		if ('<' == buf[realend])
+			stack++;
+		if ('>' != buf[realend])
 			continue;
-		else if (dsz == 1)
+		else if (stack-- > 0)
+			continue;
+		if (dsz == 1)
 			break;
-		assert(nstart > 0);
-		if (' ' != buf[nstart - 1])
+		assert(realend > 0);
+		if (' ' != buf[realend - 1])
 			continue;
-		for (i = nstart, j = 0; i < end && j < dsz; j++) 
+		for (i = realend, j = 0; i < end && j < dsz; j++) 
 			if ('>' != buf[i++])
 				break;
 		if (dsz == j) 
 			break;
 	}
-	
-	/* We don't care about stubs. */
-	if (nstart == end || nstart == *start)
+
+	/* Ignore stubs. */
+	if (realend == end || realend == *start)
 		return(0);
 
-	/* Set nend to the end of content. */
-	nend = nstart;
-	if (dsz > 1)
-		nend--;
+	/* Set linkend to the end of content. */
+	linkend = dsz > 1 ? realend - 1 : realend;
 
-	/* Re-scan to see if we have a title. */
-	text = &buf[*start];
-	for (textsz = *start; textsz < nend; textsz++)
-		if ('|' == buf[textsz])
+	/* Re-scan to see if we have a title or section. */
+	for (textsz = *start; textsz < linkend; textsz++)
+		if ('|' == buf[textsz] || '/' == buf[textsz])
 			break;
 
-	if (textsz < nend) {
-		sv = textsz + 1;
+	if (textsz < linkend && '|' == buf[textsz]) {
+		/* With title: set start, then end at section. */
+		linkstart = textsz + 1;
 		textsz = textsz - *start;
-	} else 
+		for (i = linkstart; i < linkend; i++)
+			if ('/' == buf[i])
+				break;
+		if (i < linkend)
+			linkend = i;
+	} else if (textsz < linkend && '/' == buf[textsz]) {
+		/* With section: set end at section. */
+		linkend = textsz;
 		textsz = 0;
+	} else
+		/* No title, no section. */
+		textsz = 0;
 
-	/* Now see if we're a Perl manual. */
-	for (hasdouble = 0, i = sv + 1; i < end; i++) 
-		if (':' == buf[i] && ':' == buf[i - 1]) {
-			hasdouble = 1;
+	*start = realend;
+	j = linkend - linkstart;
+
+	/* Do we have only subsection material? */
+	if (0 == j && '/' == buf[linkend]) {
+		linkstart = linkend + 1;
+		linkend = dsz > 1 ? realend - 1 : realend;
+		if (0 == (j = linkend - linkstart))
+			return(0);
+		printf("Sx %.*s", (int)j, &buf[linkstart]);
+		return(1);
+	} else if (0 == j)
+		return(0);
+
+	/* See if we qualify as being a link or not. */
+	if ((j > 4 && 0 == memcmp("http:", &buf[linkstart], j)) ||
+		(j > 5 && 0 == memcmp("https:", &buf[linkstart], j)) ||
+		(j > 3 && 0 == memcmp("ftp:", &buf[linkstart], j)) ||
+		(j > 4 && 0 == memcmp("sftp:", &buf[linkstart], j)) ||
+		(j > 3 && 0 == memcmp("smb:", &buf[linkstart], j)) ||
+		(j > 3 && 0 == memcmp("afs:", &buf[linkstart], j))) {
+		/* Gross. */
+		printf("Lk %.*s", (int)((dsz > 1 ? realend - 1 : 
+			realend) - linkstart), &buf[linkstart]);
+		return(1);
+	} 
+	
+	/* See if we qualify as a mailto. */
+	if (j > 6 && 0 == memcmp("mailto:", &buf[linkstart], j)) {
+		printf("Mt %.*s", (int)j, &buf[linkstart]);
+		return(1);
+	}
+
+	/* See if we're a foo(5), foo(5x), or foo(5xx) manpage. */
+	if ((j > 3 && ')' == buf[linkend - 1]) && 
+		('(' == buf[linkend - 3])) {
+		printf("Xr %.*s %c", (int)(j - 3), 
+			&buf[linkstart], buf[linkend - 2]);
+		return(1);
+	} else if ((j > 4 && ')' == buf[linkend - 1]) &&
+		('(' == buf[linkend - 4])) {
+		printf("Xr %.*s %.*s", (int)(j - 4), 
+			&buf[linkstart], 2, &buf[linkend - 3]);
+		return(1);
+	} else if ((j > 5 && ')' == buf[linkend - 1]) &&
+		('(' == buf[linkend - 5])) {
+		printf("Xr %.*s %.*s", (int)(j - 5), 
+			&buf[linkstart], 3, &buf[linkend - 4]);
+		return(1);
+	}
+
+	/* Last try: do we have a double-colon? */
+	for (i = linkstart + 1; i < linkend; i++)
+		if (':' == buf[i] && ':' == buf[i - 1])
 			break;
-		}
 
-	/*
-	 * Provide for some common invocations of the link primitive.
-	 * First, allow us to link to other Perl manuals.
-	 */
-	if (hasdouble)
+	if (i < linkend)
 		printf("Xr %.*s " PERL_SECTION, 
-			(int)(nend - sv), &buf[sv]);
-	else if (nend - sv > 3 && isalnum(buf[sv]) && 
-			')' == buf[nend - 1] && 
-			isdigit((int)buf[nend - 2]) &&
-			'(' == buf[nend - 3]) 
-		printf("Xr %.*s %c", 
-			(int)(nend - 3 - sv),
-			&buf[sv], buf[nend - 2]);
+			(int)j, &buf[linkstart]);
 	else
-		return(0);
+		printf("Xr %.*s 1", (int)j, &buf[linkstart]);
 
-	*start = nstart;
 	return(1);
 }
 
@@ -407,6 +458,13 @@ formatcode(struct state *st, const char *buf, size_t *
 			(*start) += dsz;
 			break;
 		}
+		if (*start < end) {
+			assert('>' == buf[*start]);
+			(*start)++;
+		}
+		if (isspace(last))
+			while (*start < end && isspace((int)buf[*start]))
+				(*start)++;
 		return(0);
 	}
 
@@ -451,12 +509,18 @@ formatcode(struct state *st, const char *buf, size_t *
 					printf("Ar ");
 				break;
 			} 
-			printf("Sy ");
+			if (0 == strncmp(buf + *start, "NULL", 4) &&
+			    ('=' == buf[*start + 4] ||
+			     '>' == buf[*start + 4]))
+				printf("Dv ");
+			else
+				printf("Sy ");
 			break;
 		case (FMT_CODE):
 			printf("Qo Li ");
 			break;
 		case (FMT_LINK):
+			/* Try to link; use "No" if it's empty. */
 			if ( ! trylink(buf, start, end, dsz))
 				printf("No ");
 			break;
@@ -802,10 +866,47 @@ static void
 verbatim(struct state *st, const char *buf, size_t start, size_t end)
 {
 	int		 last;
+	size_t		 i;
 
 	if ( ! st->parsing || st->paused)
 		return;
-
+again:
+	/* 
+	 * If we're in the SYNOPSIS, see if we're an #include block.
+	 * If we are, then print the "In" macro and re-loop.
+	 * This handles any number of inclusions, but only when they
+	 * come before the remaining parts...
+	 */
+	if (SECT_SYNOPSIS == st->sect) {
+		i = start;
+		for (i = start; i < end && ' ' == buf[i]; i++)
+			/* Spin. */ ;
+		if (i == end)
+			return;
+		/* We're an include block! */
+		if (end - i > 10 && 
+			0 == memcmp(&buf[i], "#include <", 10)) {
+			start = i + 10;
+			while (start < end && ' ' == buf[start])
+				start++;
+			fputs(".In ", stdout);
+			/* Stop til the '>' marker or we hit eoln. */
+			while (start < end && 
+				'>' != buf[start] && '\n' != buf[start])
+				putchar(buf[start++]);
+			putchar('\n');
+			if (start < end && '>' == buf[start])
+				start++;
+			if (start < end && '\n' == buf[start])
+				start++;
+			if (start < end) 
+				goto again;
+			return;
+		}
+	}
+	
+	if (start == end)
+		return;
 	puts(".Bd -literal");
 	for (last = ' '; start < end; start++) {
 		/*
@@ -1172,8 +1273,6 @@ readfile(const struct args *args, const char *fname)
 	time_t		 ttm;
 	struct stat 	 st;
 
-	assert(NULL != fname);
-
 	fd = 0 != strcmp("-", fname) ? 
 		open(fname, O_RDONLY, 0) : STDIN_FILENO;
 
@@ -1279,8 +1378,8 @@ main(int argc, char *argv[])
 
 	/* Accept only a single input file. */
 
-	if (argc > 2)
-		return(EXIT_FAILURE);
+	if (argc > 1)
+		goto usage;
 	else if (1 == argc)
 		fname = *argv;
 
@@ -1289,7 +1388,7 @@ main(int argc, char *argv[])
 
 usage:
 	fprintf(stderr, "usage: %s [-d date] " 
-		"[-n title] [-s section]\n", name);
+	    "[-n title] [-s section] [file]\n", name);
 
 	return(EXIT_FAILURE);
 }