- riscos/gui.c(path_to_url): escape the characters which need to be escaped when converting the host path to file: URL.

- utils/{url.c,url.h}(url_escape):
  * added parameter 'toskip' to specify number of input characters which need to be skipped in the escape process. This avoids extra malloc buffer juggling.
  * added parameter 'escexceptions' to specify the characters which need to be excluded from the escape process.

Solves SF tracker ID 1910169.
Note that when discname in path contains '/' characters (case: "file:///Sunfish#192.168.0.50::/home/joty.$/jo.html") or there is no discname specified at all (case "file:///HostFS:$/jo.htm"), you 
need an UnixLib fix as in http://www.riscos.info/websvn/listing.php?repname=gccsdk&path=%2Ftrunk%2Fgcc4%2F&rev=3395&sc=1

svn path=/trunk/netsurf/; revision=4069

render/form.c

@@ -564,7 +564,7 @@ char *form_url_encode(struct form *form,
 	s[0] = 0;
 
 	for (; control; control = control->next) {
-		url_err = url_escape(control->name, true, &name);
+		url_err = url_escape(control->name, 0, true, NULL, &name);
 		if (url_err == URL_FUNC_NOMEM) {
 			free(s);
 			return 0;
@@ -572,7 +572,7 @@ char *form_url_encode(struct form *form,
 
 		assert(url_err == URL_FUNC_OK);
 
-		url_err = url_escape(control->value, true, &value);
+		url_err = url_escape(control->value, 0, true, NULL, &value);
 		if (url_err == URL_FUNC_NOMEM) {
 			free(name);
 			free(s);

riscos/gui.c

@@ -1965,16 +1965,16 @@ void ro_msg_window_info(wimp_message *message)
 char *path_to_url(const char *path)
 {
 	int spare;
-	char *buffer = 0;
-	char *url = 0;
+	char *buffer, *url, *escurl;
 	os_error *error;
+	url_func_result url_err;
 
 	error = xosfscontrol_canonicalise_path(path, 0, 0, 0, 0, &spare);
 	if (error) {
 		LOG(("xosfscontrol_canonicalise_path failed: 0x%x: %s",
 				error->errnum, error->errmess));
 		warn_user("PathToURL", error->errmess);
-		return 0;
+		return NULL;
 	}
 
 	buffer = malloc(1 - spare);
@@ -1984,7 +1984,7 @@ char *path_to_url(const char *path)
 		warn_user("NoMemory", 0);
 		free(buffer);
 		free(url);
-		return 0;
+		return NULL;
 	}
 
 	error = xosfscontrol_canonicalise_path(path, buffer, 0, 0, 1 - spare,
@@ -1995,14 +1995,30 @@ char *path_to_url(const char *path)
 		warn_user("PathToURL", error->errmess);
 		free(buffer);
 		free(url);
-		return 0;
+		return NULL;
 	}
 
-	strcpy(url, "file://");
-	__unixify(buffer, __RISCOSIFY_NO_REVERSE_SUFFIX, url + 7,
-			1 - spare + 3 /* 10 - "file://" */, 0);
-	free(buffer);
-	return url;
+	memcpy(url, "file://", sizeof("file://")-1);
+	if (__unixify(buffer, __RISCOSIFY_NO_REVERSE_SUFFIX,
+			url + sizeof("file://")-1,
+			1 - spare + 10 - (sizeof("file://")-1),
+			0) == NULL) {
+		LOG(("__unixify failed: %s", buffer));
+		free(buffer);
+		free(url);
+		return NULL;
+	}
+	free(buffer); buffer = NULL;
+
+	/* We don't want '/' to be escaped.  */
+	url_err = url_escape(url, sizeof("file://")-1, false, "/", &escurl);
+	free(url); url = NULL;
+	if (url_err != URL_FUNC_OK) {
+		LOG(("url_escape failed: %s", url));
+		return NULL;
+	}
+
+	return escurl;
 }
 
 

utils/url.c

@@ -871,17 +871,19 @@ no_path:
 /**
  * Escape a string suitable for inclusion in an URL.
  *
- * \param  unescaped  the unescaped string
- * \param  sptoplus   true iff spaces should be converted to +
- * \param  result     pointer to pointer to buffer to hold escaped string
+ * \param  unescaped      the unescaped string
+ * \param  toskip         number of bytes to skip in unescaped string
+ * \param  sptoplus       true iff spaces should be converted to +
+ * \param  escexceptions  NULL or a string of characters excluded to be escaped
+ * \param  result         pointer to pointer to buffer to hold escaped string
  * \return  URL_FUNC_OK on success
  */
 
-url_func_result url_escape(const char *unescaped, bool sptoplus,
-		char **result)
+url_func_result url_escape(const char *unescaped, size_t toskip,
+		bool sptoplus, const char *escexceptions, char **result)
 {
-	int len;
-	char *escaped, *d;
+	size_t len;
+	char *escaped, *d, *tmpres;
 	const char *c;
 
 	if (!unescaped || !result)
@@ -890,21 +892,25 @@ url_func_result url_escape(const char *unescaped, bool sptoplus,
 	*result = NULL;
 
 	len = strlen(unescaped);
+	if (len < toskip)
+		return URL_FUNC_FAILED;
+	len -= toskip;
 
 	escaped = malloc(len * 3 + 1);
 	if (!escaped)
 		return URL_FUNC_NOMEM;
 
-	for (c = unescaped, d = escaped; *c; c++) {
+	for (c = unescaped + toskip, d = escaped; *c; c++) {
 		/* Check if we should escape this byte.
 		 * '~' is unreserved and should not be percent encoded, if
 		 * you believe the spec; however, leaving it unescaped
 		 * breaks a bunch of websites, so we escape it anyway. */
-		if (!isascii(*c) || strchr(":/?#[]@" /* gen-delims */
-					"!$&'()*+,;=" /* sub-delims */
-					"<>%\"{}|\\^`~", /* others */
-					*c) ||
-				*c <= 0x20 || *c == 0x7f) {
+		if (!isascii(*c)
+			|| (strchr(":/?#[]@" /* gen-delims */
+				  "!$&'()*+,;=" /* sub-delims */
+				  "<>%\"{}|\\^`~" /* others */,	*c)
+				&& (!escexceptions || !strchr(escexceptions, *c)))
+			|| *c <= 0x20 || *c == 0x7f) {
 			if (*c == 0x20 && sptoplus) {
 				*d++ = '+';
 			} else {
@@ -917,16 +923,17 @@ url_func_result url_escape(const char *unescaped, bool sptoplus,
 			*d++ = *c;
 		}
 	}
-
 	*d++ = '\0';
 
-	(*result) = malloc(d - escaped);
-	if (!(*result)) {
+	tmpres = malloc(d - escaped + toskip);
+	if (!tmpres) {
 		free(escaped);
 		return URL_FUNC_NOMEM;
 	}
 
-	memcpy((*result), escaped, d - escaped);
+	memcpy(tmpres, unescaped, toskip); 
+	memcpy(tmpres + toskip, escaped, d - escaped);
+	*result = tmpres;
 
 	free(escaped);
 

utils/url.h

@@ -47,8 +47,8 @@ url_func_result url_host(const char *url, char **result);
 url_func_result url_scheme(const char *url, char **result);
 url_func_result url_nice(const char *url, char **result,
 		bool remove_extensions);
-url_func_result url_escape(const char *unescaped, bool sptoplus,
-		char **result);
+url_func_result url_escape(const char *unescaped, size_t toskip,
+		bool sptoplus, const char *escexceptions, char **result);
 url_func_result url_canonical_root(const char *url, char **result);
 url_func_result url_parent(const char *url, char **result);
 url_func_result url_plq(const char *url, char **result);