[Webkit-unassigned] [Bug 67277] REGRESSION (r94132): broke fast/loader/location-port.html on GTK

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=67277





--- Comment #9 from Philippe Normand <pnormand at igalia.com>  2011-09-08 00:24:59 PST ---
(In reply to comment #8)
> pnormand: Keep in mind that it passes for Safari, which uses KURL unmodified. Also, mrobinson and I ran it on our local machines and saw the tests pass, too.
> 

Did you unskip it before running it? Because I noticed that if a test is skipped and I try to run it directly with run-webkit-tests it will just be ignored. This is a new behavior compared to old-run-webkit-tests, I think.

> ::parse treats all consecutive digits in the port field as port number, so setting to 0 should work fine. (And setPort doesn't have any exception for port 0)
> 
> Could you elaborate why you think the test shouldn't be able to pass at all?

the bug is that setting port to 0 on a file url, hostname becomes empty string instead of ":0". Location::host() returns:

return url.hasPort() ? url.host() + ":" + String::number(url.port()) : url.host();

In the location-port.html test, when port is set to 0, it is expected for host to become ":0":

in checkTest3(): internalFrame.contentWindow.location.port = 0;
in checkTest4(): shouldBe('internalFrame.contentWindow.location.host', ':0');

But this is not what I observe locally :/
Oh, well. I guess I'm just becoming crazy ;-)

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.