Bug 104285 – [GTK] Add authentication support to DRT and fix exposed issues in the libsoup backend

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

[Martin Robinson]

This bug covers adding authentication support to DumpRenderTree and fixing any issues in ResourceHandleSoup that this exposes.

[Martin Robinson]

Created attachment 178368 [details]
Patch

[Martin Robinson]

This also requires a bump to the libsoup version in the jhbuild modules to fix a libsoup bug. I'll handle that in a separate patch.

[Martin Robinson]

Created attachment 178386 [details]
Properly handle authentication restarts in restartedCallback

[Dan Winship]

Comment on attachment 178386 [details]
Properly handle authentication restarts in restartedCallback

View in context: https://bugs.webkit.org/attachment.cgi?id=178386&action=review

> Source/WebCore/platform/network/soup/ResourceRequestSoup.cpp:146
> +    // soup_uri_new will convert empty usernames and passwords into null. Some parts of
> +    // soup like the SoupAuthenticationManager will only be active when both the username
> +    // and password are non-null. When we have credentials, empty usernames and passwords
> +    // should be empty strings instead of null.

This didn't sound right to me, so I added a test, and it does seem to work: http://git.gnome.org/browse/libsoup/commit/?id=66650e5

So, not sure what was failing for you.

> Tools/DumpRenderTree/gtk/TestRunnerGtk.cpp:170
> +    // soup_uri_to_string does not insert the password into the string, so we need to create the
> +    // URI string and then reinsert any credentials that were present in the SoupURI. All tests that
> +    // use URL-embedded credentials use HTTP, so it's safe here.

you could add a soup_uri_to_string_with_password() (or something) if you want. Other people have wanted that before...

> LayoutTests/platform/gtk/TestExpectations:-879
> -# No authentication challenge handling

woooo

> LayoutTests/platform/gtk/http/tests/misc/401-alternative-content-expected.txt:-1
> -PASS

any reason we don't just print out the debug message to match the default expectation?

[Dan Winship]

(In reply to comment #4)
> > LayoutTests/platform/gtk/http/tests/misc/401-alternative-content-expected.txt:-1
> > -PASS
> 
> any reason we don't just print out the debug message to match the default expectation?

oh, oops, that's a removal not an addition

[Martin Robinson]

Created attachment 178401 [details]
New patch with better comment explaining the null password issue

[Gustavo Noronha (kov)]

Comment on attachment 178401 [details]
New patch with better comment explaining the null password issue

LGTM!

[Martin Robinson]

Committed r137487: <http://trac.webkit.org/changeset/137487>