Bug 234809 – Make sure secure websocket connections in service workers can trigger authentication challenge callbacks

• Status: RESOLVED
• Resolution: FIXED
• Component: Service Workers

[youenn fablet]

Make sure secure websocket connections in service workers can trigger authentication challenge callbacks

[youenn fablet]

<rdar://85888177>

[youenn fablet]

Created attachment 448217 [details]
Patch

[youenn fablet]

Created attachment 448245 [details]
Patch

[youenn fablet]

Created attachment 448250 [details]
Patch

[youenn fablet]

Created attachment 448272 [details]
Patch

[youenn fablet]

Created attachment 448281 [details]
Patch

[Chris Dumez]

Comment on attachment 448250 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=448250&action=review

r=me

> Source/WebKit/NetworkProcess/cocoa/WebSocketTaskCocoa.h:73
> +    const std::optional<WebCore::SecurityOriginData>& topOrigin() const { return m_topOrigin; }

SecurityOriginData has an "empty" state so we could use a SecurityOriginData as data members (instead of a std::optional<>) and have the call sites check SecurityOriginData::isEmpty().

> Source/WebKit/NetworkProcess/cocoa/WebSocketTaskCocoa.mm:51
> +    if (clientOrigin.topOrigin == clientOrigin.clientOrigin)

This could use a comment to explain why we only initialize m_topOrigin if topOrigin is the same as clientOrigin.

> Source/WebKit/WebProcess/Network/WebSocketChannel.cpp:121
> +    MessageSender::send(Messages::NetworkConnectionToWebProcess::CreateSocketChannel { *request, protocol, m_identifier, m_webPageProxyID, ClientOrigin { m_document->topOrigin().data(), m_document->securityOrigin().data() } });

I bet we have this logic in several places and it may be worth having a getter on Document that returns a ClientOrigin at this point.

[Alex Christensen]

Comment on attachment 448250 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=448250&action=review

> LayoutTests/platform/mac-bigsur-wk2/http/tests/workers/service/serviceworker-websocket.https-expected.txt:3
> +PASS Open a WebSocket in service worker

These files are identical to the platform independent version and unnecessary.

[youenn fablet]

(In reply to Alex Christensen from comment #8)
> Comment on attachment 448250 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=448250&action=review
> 
> > LayoutTests/platform/mac-bigsur-wk2/http/tests/workers/service/serviceworker-websocket.https-expected.txt:3
> > +PASS Open a WebSocket in service worker
> 
> These files are identical to the platform independent version and
> unnecessary.

Not really, they are missing the canAuthenticateAgainstProtectionSpace line.
I guess we could reverse the files but in the future, all cocoa ports will probably be aligned with the updated test expectation.

[Alex Christensen]

Comment on attachment 448250 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=448250&action=review

>>> LayoutTests/platform/mac-bigsur-wk2/http/tests/workers/service/serviceworker-websocket.https-expected.txt:3
>>> +PASS Open a WebSocket in service worker
>> 
>> These files are identical to the platform independent version and unnecessary.
> 
> Not really, they are missing the canAuthenticateAgainstProtectionSpace line.
> I guess we could reverse the files but in the future, all cocoa ports will probably be aligned with the updated test expectation.

oh, of course.  I was looking at it wrong.  This is fine.

[youenn fablet]

Created attachment 448365 [details]
Patch for landing

[EWS]

Committed r287611 (245738@main): <https://commits.webkit.org/245738@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 448365 [details].

[youenn fablet]

Took it all.

(In reply to Chris Dumez from comment #7)
> Comment on attachment 448250 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=448250&action=review
> 
> r=me
> 
> > Source/WebKit/NetworkProcess/cocoa/WebSocketTaskCocoa.h:73
> > +    const std::optional<WebCore::SecurityOriginData>& topOrigin() const { return m_topOrigin; }
> 
> SecurityOriginData has an "empty" state so we could use a SecurityOriginData
> as data members (instead of a std::optional<>) and have the call sites check
> SecurityOriginData::isEmpty().
> 
> > Source/WebKit/NetworkProcess/cocoa/WebSocketTaskCocoa.mm:51
> > +    if (clientOrigin.topOrigin == clientOrigin.clientOrigin)
> 
> This could use a comment to explain why we only initialize m_topOrigin if
> topOrigin is the same as clientOrigin.
> 
> > Source/WebKit/WebProcess/Network/WebSocketChannel.cpp:121
> > +    MessageSender::send(Messages::NetworkConnectionToWebProcess::CreateSocketChannel { *request, protocol, m_identifier, m_webPageProxyID, ClientOrigin { m_document->topOrigin().data(), m_document->securityOrigin().data() } });
> 
> I bet we have this logic in several places and it may be worth having a
> getter on Document that returns a ClientOrigin at this point.

[youenn fablet]

*** Bug 233665 has been marked as a duplicate of this bug. ***