Support remote links in XPS documents.

Update xps path handling to cope with URLs.

Fix premature freeing of links.

Spot remote URLs and use appropriate link type.

8 files changed, 127 insertions, 46 deletions

diff --git a/apps/pdfapp.c b/apps/pdfapp.c
index aa529220..20e67c62 100644
--- a/apps/pdfapp.c
+++ b/apps/pdfapp.c
@@ -375,6 +375,7 @@ static void pdfapp_loadpage_xps(pdfapp_t *app)
 	fz_free_device(mdev);
 
 	app->page_links = page->links;
+	page->links = NULL;
 
 	xps_free_page(app->xps, page);
 }
diff --git a/xps/muxps.h b/xps/muxps.h
index 2f87edc6..2e5c4206 100644
--- a/xps/muxps.h
+++ b/xps/muxps.h
@@ -32,7 +32,8 @@ typedef struct xps_document_s xps_document;
  */
 
 int xps_strcasecmp(char *a, char *b);
-void xps_absolute_path(char *output, char *base_uri, char *path, int output_size);
+void xps_resolve_url(char *output, char *base_uri, char *path, int output_size);
+int xps_url_is_remote(char *path);
 
 /*
  * XML document model
diff --git a/xps/xps_doc.c b/xps/xps_doc.c
index d1943fc0..0bc7e443 100644
--- a/xps/xps_doc.c
+++ b/xps/xps_doc.c
@@ -90,32 +90,42 @@ xps_add_link(xps_document *doc, fz_rect area, char *base_uri, char *target_uri)
 		len = 2 + (base_uri ? strlen(base_uri) : 0) +
 			(target_uri ? strlen(target_uri) : 0);
 		buffer = fz_malloc(doc->ctx, len);
-		xps_absolute_path(buffer, base_uri, target_uri, len);
-		uri = buffer;
-
-		/* FIXME: This won't work for remote docs */
-		/* Skip until we find the fragment marker */
-		while (*uri && *uri != '#')
-			uri++;
-		if (*uri == '#')
-			uri++;
-
-		for (target = doc->target; target; target = target->next)
-			if (!strcmp(target->name, uri))
-				break;
+		xps_resolve_url(buffer, base_uri, target_uri, len);
+		if (xps_url_is_remote(buffer))
+		{
+			dest.kind = FZ_LINK_URI;
+			dest.ld.uri.is_map = 0;
+			dest.ld.uri.uri = buffer;
+			buffer = NULL;
+		}
+		else
+		{
+			uri = buffer;
 
-		if (target == NULL)
-			break;
+			/* FIXME: This won't work for remote docs */
+			/* Skip until we find the fragment marker */
+			while (*uri && *uri != '#')
+				uri++;
+			if (*uri == '#')
+				uri++;
 
-		dest.kind = FZ_LINK_GOTO;
-		dest.ld.gotor.flags = fz_link_flag_l_valid | fz_link_flag_t_valid | fz_link_flag_r_valid | fz_link_flag_b_valid;
-		dest.ld.gotor.lt.x = area.x0;
-		dest.ld.gotor.lt.y = area.y0;
-		dest.ld.gotor.rb.x = area.x1;
-		dest.ld.gotor.rb.y = area.y1;
-		dest.ld.gotor.page = target->page;
-		dest.ld.gotor.file_spec = NULL;
-		dest.ld.gotor.new_window = 0;
+			for (target = doc->target; target; target = target->next)
+				if (!strcmp(target->name, uri))
+					break;
+
+			if (target == NULL)
+				break;
+
+			dest.kind = FZ_LINK_GOTO;
+			dest.ld.gotor.flags = 0;
+			dest.ld.gotor.lt.x = 0;
+			dest.ld.gotor.lt.y = 0;
+			dest.ld.gotor.rb.x = 0;
+			dest.ld.gotor.rb.y = 0;
+			dest.ld.gotor.page = target->page;
+			dest.ld.gotor.file_spec = NULL;
+			dest.ld.gotor.new_window = 0;
+		}
 
 		link = fz_new_link(doc->ctx, area, dest);
 		link->next = doc->current_page->links;
@@ -256,7 +266,7 @@ xps_parse_metadata_imp(xps_document *doc, xml_element *item, xps_fixdoc *fixdoc)
 			if (target && type)
 			{
 				char tgtbuf[1024];
-				xps_absolute_path(tgtbuf, doc->base_uri, target, sizeof tgtbuf);
+				xps_resolve_url(tgtbuf, doc->base_uri, target, sizeof tgtbuf);
 				if (!strcmp(type, REL_START_PART))
 					doc->start_part = fz_strdup(doc->ctx, tgtbuf);
 				if (!strcmp(type, REL_DOC_STRUCTURE) && fixdoc)
@@ -270,7 +280,7 @@ xps_parse_metadata_imp(xps_document *doc, xml_element *item, xps_fixdoc *fixdoc)
 			if (source)
 			{
 				char srcbuf[1024];
-				xps_absolute_path(srcbuf, doc->base_uri, source, sizeof srcbuf);
+				xps_resolve_url(srcbuf, doc->base_uri, source, sizeof srcbuf);
 				xps_add_fixed_document(doc, srcbuf);
 			}
 		}
@@ -285,7 +295,7 @@ xps_parse_metadata_imp(xps_document *doc, xml_element *item, xps_fixdoc *fixdoc)
 			if (source)
 			{
 				char srcbuf[1024];
-				xps_absolute_path(srcbuf, doc->base_uri, source, sizeof srcbuf);
+				xps_resolve_url(srcbuf, doc->base_uri, source, sizeof srcbuf);
 				xps_add_fixed_page(doc, srcbuf, width, height);
 			}
 		}
@@ -421,7 +431,6 @@ xps_load_page(xps_document *doc, int number)
 			doc->current_page = page;
 			if (!page->root)
 				xps_load_fixed_page(doc, page);
-			page->links_resolved = 1;
 			return page;
 		}
 		n ++;
diff --git a/xps/xps_glyphs.c b/xps/xps_glyphs.c
index 0cc4af5d..c394dd18 100644
--- a/xps/xps_glyphs.c
+++ b/xps/xps_glyphs.c
@@ -473,7 +473,7 @@ xps_parse_glyphs(xps_document *doc, fz_matrix ctm,
 	 * Find and load the font resource
 	 */
 
-	xps_absolute_path(partname, base_uri, font_uri_att, sizeof partname);
+	xps_resolve_url(partname, base_uri, font_uri_att, sizeof partname);
 	subfont = strrchr(partname, '#');
 	if (subfont)
 	{
diff --git a/xps/xps_image.c b/xps/xps_image.c
index 2c159d14..10d3baee 100644
--- a/xps/xps_image.c
+++ b/xps/xps_image.c
@@ -83,7 +83,7 @@ xps_find_image_brush_source_part(xps_document *doc, char *base_uri, xml_element
 	if (!image_name)
 		fz_throw(doc->ctx, "cannot find image source");
 
-	xps_absolute_path(partname, base_uri, image_name, sizeof partname);
+	xps_resolve_url(partname, base_uri, image_name, sizeof partname);
 
 	return xps_read_part(doc, partname);
 }
diff --git a/xps/xps_resource.c b/xps/xps_resource.c
index cfac562d..dcf3717d 100644
--- a/xps/xps_resource.c
+++ b/xps/xps_resource.c
@@ -63,7 +63,7 @@ xps_parse_remote_resource_dictionary(xps_document *doc, char *base_uri, char *so
 	char *s;
 
 	/* External resource dictionaries MUST NOT reference other resource dictionaries */
-	xps_absolute_path(part_name, base_uri, source_att, sizeof part_name);
+	xps_resolve_url(part_name, base_uri, source_att, sizeof part_name);
 	part = xps_read_part(doc, part_name);
 	xml = xml_parse_document(doc->ctx, part->data, part->size);
 	xps_free_part(doc, part);
diff --git a/xps/xps_tile.c b/xps/xps_tile.c
index fe4d2350..e87d76f7 100644
--- a/xps/xps_tile.c
+++ b/xps/xps_tile.c
@@ -375,4 +375,5 @@ xps_run_page(xps_document *doc, xps_page *page, fz_device *dev, fz_matrix ctm, f
 	xps_parse_fixed_page(doc, ctm, page);
 	doc->cookie = NULL;
 	doc->dev = NULL;
+	page->links_resolved = 1;
 }
diff --git a/xps/xps_util.c b/xps/xps_util.c
index 6af4da91..5a3b1fbf 100644
--- a/xps/xps_util.c
+++ b/xps/xps_util.c
@@ -20,15 +20,77 @@ xps_strcasecmp(char *a, char *b)
 	return xps_tolower(*a) - xps_tolower(*b);
 }
 
+/* A URL is defined as consisting of a:
+ * SCHEME (e.g. http:)
+ * AUTHORITY (username, password, hostname, port, eg //test:passwd@mupdf.com:999)
+ * PATH (e.g. /download)
+ * QUERY (e.g. ?view=page)
+ * FRAGMENT (e.g. #fred) (not strictly part of the URL)
+ */
+static char *
+skip_scheme(char *path)
+{
+	char *p = path;
+
+	/* Skip over: alpha *(alpha | digit | "+" | "-" | ".") looking for : */
+	if (*p >= 'a' && *p <= 'z')
+	{}
+	else if (*p >= 'A' && *p <= 'Z')
+	{}
+	else
+		return path;
+
+	while (*++p)
+	{
+		if (*p >= 'a' && *p <= 'z')
+		{}
+		else if (*p >= 'A' && *p <= 'Z')
+		{}
+		else if (*p >= '0' && *p <= '9')
+		{}
+		else if (*p == '+')
+		{}
+		else if (*p == '-')
+		{}
+		else if (*p == '.')
+		{}
+		else if (*p == ':')
+			return p+1;
+		else
+			break;
+	}
+	return path;
+}
+
+static char *
+skip_authority(char *path)
+{
+	char *p = path;
+
+	/* Authority section must start with '//' */
+	if (p[0] != '/' || p[1] != '/')
+		return path;
+	p += 2;
+
+	/* Authority is terminated by end of URL, '/' or '?' */
+	while (*p && *p != '/' && *p != '?')
+		p++;
+
+	return p;
+}
+
+
 #define SEP(x)	((x)=='/' || (x) == 0)
 
 static char *
 xps_clean_path(char *name)
 {
-	char *p, *q, *dotdot;
+	char *p, *q, *dotdot, *start;
 	int rooted;
 
-	rooted = name[0] == '/';
+	start = skip_scheme(name);
+	start = skip_authority(start);
+	rooted = start[0] == '/';
 
 	/*
 	 * invariants:
@@ -37,7 +99,7 @@ xps_clean_path(char *name)
 	 *		dotdot points just past the point where .. cannot backtrack
 	 *				any further (no slash).
 	 */
-	p = q = dotdot = name + rooted;
+	p = q = dotdot = start + rooted;
 	while (*p)
 	{
 		if(p[0] == '/') /* null element */
@@ -54,7 +116,7 @@ xps_clean_path(char *name)
 			}
 			else if (!rooted) /* /.. is / but ./../ is .. */
 			{
-				if (q != name)
+				if (q != start)
 					*q++ = '/';
 				*q++ = '.';
 				*q++ = '.';
@@ -63,36 +125,43 @@ xps_clean_path(char *name)
 		}
 		else /* real path element */
 		{
-			if (q != name+rooted)
+			if (q != start+rooted)
 				*q++ = '/';
 			while ((*q = *p) != '/' && *q != 0)
 				p++, q++;
 		}
 	}
 
-	if (q == name) /* empty string is really "." */
+	if (q == start) /* empty string is really "." */
 		*q++ = '.';
 	*q = '\0';
 
 	return name;
 }
 
-/* RJW: Possible problems here:
- * "/foo" + "../../bar" = "/bar" not "../bar"
- * "http://something/" + ... doesn't work.
- */
 void
-xps_absolute_path(char *output, char *base_uri, char *path, int output_size)
+xps_resolve_url(char *output, char *base_uri, char *path, int output_size)
 {
-	if (path[0] == '/')
+	char *p = skip_authority(skip_scheme(path));
+
+	if (p != path || path[0] == '/')
 	{
 		fz_strlcpy(output, path, output_size);
 	}
 	else
 	{
-		fz_strlcpy(output, base_uri, output_size);
-		fz_strlcat(output, "/", output_size);
+		int len = fz_strlcpy(output, base_uri, output_size);
+		if (len == 0 || output[len-1] != '/')
+			fz_strlcat(output, "/", output_size);
 		fz_strlcat(output, path, output_size);
 	}
 	xps_clean_path(output);
 }
+
+int
+xps_url_is_remote(char *path)
+{
+	char *p = skip_authority(skip_scheme(path));
+
+	return p != path;
+}