Merge lp:~ivaldi/midori/fix-1208995 into lp:midori

[André Stösel]

Fix segfault if url contains " %00"

[gue5t gue5t (gue5t)]

This fixes the crash but causes urls like "http://www.google.com/%21%00ok" to not decode any characters, whereas the url "http://www.google.com/%21ok" gets normalized to "http://www.google.com/!ok" for display. Presumably we should ensure that 0x00 characters are not urldecoded while still decoding the others.

We also shouldn't ever decode %25 (encoded '%'), because doing so can cause hitting enter in the url bar to change the actual url, e.g. a page which is locally on disk as "get%100.html" would be urlencoded as "get%25100.html" but would request "get%100.html" when the user hits enter in the urlbar after visiting the page.

review: Needs Fixing

[André Stösel (ivaldi)]

I agree, we should review the url en/decoding.
However, this isn't related with the segfault -> I opened a new bug #1209167 for this.

[gue5t gue5t (gue5t)]

I agree, we can fix the handling later. This does fix the segfault, no regression, so r+.

review: Approve

=== modified file 'katze/midori-uri.vala'
--- katze/midori-uri.vala	2012-11-07 00:21:15 +0000
+++ katze/midori-uri.vala	2013-08-06 21:56:22 +0000
@@ -84,10 +84,12 @@
                 else if (!unescaped.validate ())
                     return uri;
                 string path;
-                string hostname = parse_hostname (unescaped, out path);
-                string decoded = hostname_to_unicode (hostname);
-                if (decoded != null)
-                    return "http://" + decoded + path;
+                string? hostname = parse_hostname (unescaped, out path);
+                if (hostname != null) {
+                    string decoded = hostname_to_unicode (hostname);
+                    if (decoded != null)
+                        return "http://" + decoded + path;
+                }
                 return unescaped;
             }
             return uri;