Bug 24181 – GTK+ crashes in fast/loader/empty-embed-src-attribute.html

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKitGTK

[Gustavo Noronha (kov)]

This is because a load is tried for a null request, which ends up crashing. This is most probably caused by the load code paths not doing proper checks or conversions on the URL or the request. The crash doesn't happen in other ports because the URL gets set to something by the port's platform code.

See https://bugs.webkit.org/show_bug.cgi?id=23806 for more information.

[Gustavo Noronha (kov)]

Created attachment 28005 [details]
proposed fix

As I noted on the bug report I mentioned when reporting this one, this doesn't seem to be a GTK+-specific problem, and this is probably not the definitive fix, but since this test was added with a work-around fix for other ports, I believe this work-around for GTK+ is in order.

[Gustavo Noronha (kov)]

Created attachment 28012 [details]
a different approach that still loads

This is a better approach that still causes the load events to happen, as intended.

[Holger Freyther]

That looks so ugly. Give me some hours to find something better.

[Holger Freyther]

19:04 < dglazkov> zecke: I can't see this problem. My brain has completely blacklisted anything that may even potentially be related to this 
                  issue
19:04 < dglazkov> :)
19:05 < dglazkov> zecke: but somewhere deep in the back, it's still thinking about it!
19:05 < dglazkov> I think it's in medula oblongata
19:05 < zecke> dglazkov: hehe, the question is what is chromium or the mac loading for an empty url?
19:07 < dglazkov> the policy checker issues PolicyIgnore for empty URLs
19:07 < zecke> thanks! :)
19:07 < zecke> _manyoso: you want to see the above as well
19:07 < dglazkov> zecke, _manyoso: but that's the wrong fix.
19:08 < dglazkov> we should really just fix it in the loader
19:08 < dglazkov> I mean, it's an okay short-term fix

so PolicyIgnore would be another option?

[Gustavo Noronha (kov)]

Created attachment 28051 [details]
yet another approach

OK, this works too =).

[Dave Hyatt]

Comment on attachment 28051 [details]
yet another approach

r=me

[Gustavo Noronha (kov)]

Landed in r41313.