Bug 26925 – URL fragment breaks application cache loads

• Status: RESOLVED
• Resolution: FIXED
• Component: New Bugs

[Andrew Grieve]

The URL Hash is store in the AppCache database and it does not allow an appcached page to be loaded from appcache with a different fragment from the one it was initially loaded with.

Steps to reproduce:

1. Create a Hello World style webpage and a manifest that has no entries in it.
2. Open the page and change the URL hash to #foo
2. Do a browser refresh (to put it into AppCache)
3. Go offline
4. Do a browser refresh and ensure the page refreshes from AppCache.
5. Change the URL hash to #bar
6. Do a browser refresh and notice that it fails to load.

[Mark Rowe (bdash)]

<rdar://problem/7029995>

[Andrew Grieve]

I also opened a radar bug for this (sorry for not mentioning it before):

<rdar://problem/7027850>

Also, I've found an existing hello world style page to test against:

http://www.thecssninja.com/demo/offline_webapp/index.html

[Alexey Proskuryakov]

I think that the behavior is correct per the current spec - it never says that fragments should be ignored when loading resources, or when storing implicit resources. It only says that fragments should be ignored on URLs in manifests, as far as I can see.

But I also think that the desired behavior is to ignore them.

[Michael Nordman]

I think the spec has a bug if that's what it says case :)

[Ian 'Hixie' Hickson]

http://www.whatwg.org/specs/web-apps/current-work/#parsing-cache-manifests

The spec says to drop the fragment identifier part of any URLs parsed in the manifest.

[Ian 'Hixie' Hickson]

Nevermind, I understand now. Yeah, this is wrong in the spec.

[Alexey Proskuryakov]

Created attachment 32687 [details]
proposed fix

[Darin Adler]

Comment on attachment 32687 [details]
proposed fix

r=me

Should we use KURL more and String less?

[Alexey Proskuryakov]

Committed <http://trac.webkit.org/changeset/45848>.

(In reply to comment #8)
> Should we use KURL more and String less?

Probably not worth it in most cases here, as they only check hasRef() in debug builds.

[Ian 'Hixie' Hickson]

spec is fixed

[Alexey Proskuryakov]

Oops, forgot to mark this bug as fixed: <http://trac.webkit.org/projects/webkit/changeset/45848>.