Make WebSocketHandleCFNet respect proxy auto-configuration files via CFProxySupport

• Component: Platform
• Status: RESOLVED FIXED

[Alexey Proskuryakov]

Patch forthcoming.

We'll also need custom code for Tiger, which didn't have CFProxySupport.

[Alexey Proskuryakov]

Created attachment 43617 [details]
proposed patch

[Oliver Hunt]

Comment on attachment 43617 [details]
proposed patch

I don't get the change in SocketStreamHandle -- is the new code tiger safe?  You seem to have removed the tiger guard...

Nothing else seems to stand out as being wrong, but this code is to far from anything i really understand to be able to meaningfully review it sorry :-(

[Alexey Proskuryakov]

> You seem to have removed the tiger guard...

Not removed, just moved.

[Brady Eidson]

Comment on attachment 43617 [details]
proposed patch

We can't layout test PAC stuff, can we?

r=me but...

> -struct MainThreadEventCallbackInfo {
> -    MainThreadEventCallbackInfo(CFStreamEventType type, SocketStreamHandle* handle) : type(type), handle(handle) { }
> -    CFStreamEventType type;
> -    SocketStreamHandle* handle;
> -};
> +struct MainThreadEventCallbackInfo {
> +    MainThreadEventCallbackInfo(CFStreamEventType type, SocketStreamHandle* handle) : type(type), handle(handle) { }
> +    CFStreamEventType type;
> +    SocketStreamHandle* handle;
> +};

Maybe I'm blind, I don't know why this diff exists.

>  #if PLATFORM(WIN)
> -void SocketStreamHandle::readStreamCallbackMainThread(void* invocation)
> -{
> -    MainThreadEventCallbackInfo* info = static_cast<MainThreadEventCallbackInfo*>(invocation);
> -    info->handle->readStreamCallback(info->type);
> -}
> -
> -void SocketStreamHandle::writeStreamCallbackMainThread(void* invocation)
> -{
> -    MainThreadEventCallbackInfo* info = static_cast<MainThreadEventCallbackInfo*>(invocation);
> -    info->handle->writeStreamCallback(info->type);
> -}
> +void SocketStreamHandle::readStreamCallbackMainThread(void* invocation)
> +{
> +    MainThreadEventCallbackInfo* info = static_cast<MainThreadEventCallbackInfo*>(invocation);
> +    info->handle->readStreamCallback(info->type);
> +}
> +
> +void SocketStreamHandle::writeStreamCallbackMainThread(void* invocation)
> +{
> +    MainThreadEventCallbackInfo* info = static_cast<MainThreadEventCallbackInfo*>(invocation);
> +    info->handle->writeStreamCallback(info->type);
> +}

Or this one.

[Alexey Proskuryakov]

>We can't layout test PAC stuff, can we?

We don't have the infrastructure to test proxies at all.

>Maybe I'm blind, I don't know why this diff exists.

Inconsistent line endings somehow got into svn repository (even though svn usually prevents this). Xcode automatically cured the problem.

Committed <http://trac.webkit.org/changeset/51314>.