Bug 97415 – We should not expect percent-escaped hostname for invalid URLs

• Status: RESOLVED
• Resolution: FIXED
• Component: WebCore Misc.

[Benjamin Poulain]

That's a Chrome behavior, no other browser does that.

[Benjamin Poulain]

Created attachment 165298 [details]
Patch

[WebKit Review Bot]

Comment on attachment 165298 [details]
Patch

Attachment 165298 [details] did not pass chromium-ews (chromium-xvfb):
Output: http://queues.webkit.org/results/13989490

New failing tests:
fast/url/ipv4.html

[Benjamin Poulain]

> New failing tests:
> fast/url/ipv4.html

Interesting, the other two tests should have also failed.

I forgot to update the Chromium results, I'll do that tomorrow.

[Adam Barth]

Comment on attachment 165298 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=165298&action=review

> LayoutTests/ChangeLog:8
> +        The URL layout tests encode some Chromium specific behavior.

Yeah, the "PASS" results on these tests are just set to whatever behavior Chromium has.  The idea was to make it easy to see what the differences were between GURL and KURL by just reading off the PASS/FAIL results in the apple-mac port.  I think it's fine to convert these tests to have the PASS behavior be what we think the right behavior is, especially when there's a clear choice.

[Adam Barth]

annevk might be interested in these sorts of patches since he's working on the URL spec.

[Anne van Kesteren]

Thanks Adam. My plan was indeed for the URL standard to agree with other browsers here. In particular, if a URL is deemed invalid .href just returns the input string and all other components are their default value (typically empty string).

[Benjamin Poulain]

(In reply to comment #6)
> Thanks Adam. My plan was indeed for the URL standard to agree with other browsers here. In particular, if a URL is deemed invalid .href just returns the input string and all other components are their default value (typically empty string).

This is the behavior I want.

KURL use to partially parse invalid string like GoogleURL, but I fixed it a few months ago. In my opinion, invalid input should not be modified.

[Benjamin Poulain]

Created attachment 165445 [details]
Patch

[Benjamin Poulain]

Comment on attachment 165445 [details]
Patch

Clearing flags on attachment: 165445

Committed r129445: <http://trac.webkit.org/changeset/129445>

[Benjamin Poulain]

All reviewed patches have been landed.  Closing bug.