attached a saved html on which i get repeatable crash with midori-git and webkit-svn. File is a wget'ed output of http://cgi.ebay.com/IBM-Lenovo-ThinkPad-X61-C2D-12-1-120GB-3G-HSPDA_W0QQitemZ270313235609QQcmdZViewItemQQptZLaptops_Nov05?hash=item270313235609&_trksid=p3286.c0.m14&_trkparms=72%3A1234%7C66%3A2%7C65%3A12%7C39%3A1%7C240%3A1318%7C301%3A1%7C293%3A1%7C294%3A50 webkit is built with ./configure --with-http-backend=soup --prefix=/usr/ --disable-svg --disable-geolocation --disable-video --enable-optimizations
Created attachment 25870 [details] test case
Created attachment 25959 [details] core from webkit-debug
Created attachment 25960 [details] gdb bt
Created attachment 25961 [details] gdb run
ok. stripped down html source and found where is an issue. Now, im getting stable crash on a simple html page like: <img src=http:///google.com> Three slashes in image url is a cause of a crash.
Created attachment 26234 [details] new simplified test
Created attachment 26238 [details] patch which fixes a crash
Comment on attachment 26238 [details] patch which fixes a crash Please see http://webkit.org/coding/contributing.html and attach a new version with a ChangeLog and an applyable patch. > --- ResourceHandleSoup.cpp 2008-12-24 13:16:59.000000000 -0400 > +++ ResourceHandleSoup.cpp.new 2008-12-24 13:18:07.000000000 -0400 > @@ -335,7 +335,8 @@ > > if (equalIgnoringCase(protocol, "data")) > return startData(urlString); > - else if (equalIgnoringCase(protocol, "http") || equalIgnoringCase(protocol, "https")) > + else if ( (equalIgnoringCase(protocol, "http") || equalIgnoringCase(protocol, "https") ) && > + SOUP_URI_VALID_FOR_HTTP(soup_uri_new(urlString.utf8().data()))) > return startHttp(urlString); why did you decide to do this check here and not within startHttp? Is there anyone else calling startHttp?
the point of the check here is that once url will not be valid we will fallback to last 'else' to call didFail() without a code duplication. I will update patch shortly to include changelog.
Created attachment 26243 [details] updated patch
(In reply to comment #10) > Created an attachment (id=26243) [review] Okay, I'm convinced that this is the right patch. There is one minor style issue thiugh. Would you mind uploading a new patch without the extra whitespace in the if? WebKit.org has a rather strict CodingStyle.
Created attachment 26269 [details] coding style fix
Created attachment 26270 [details] typo
Comment on attachment 26270 [details] typo thanks.
Landed in r39528.