WebURLSchemeHandlerProxy::loadSynchronously crash with sync request

[Jin]

The problem code is the following:

void WebURLSchemeHandlerProxy::loadSynchronously(ResourceLoadIdentifier loadIdentifier, const ResourceRequest& request, ResourceResponse& response, ResourceError& error, Vector<char>& data)
{
    IPC::DataReference dataReference;
    if (!m_webPage.sendSync(Messages::WebPageProxy::LoadSynchronousURLSchemeTask(URLSchemeTaskParameters { m_identifier, loadIdentifier, request }), Messages::WebPageProxy::LoadSynchronousURLSchemeTask::Reply(response, error, dataReference))) {
        error = failedCustomProtocolSyncLoad(request);
        return;
    }
    
    data.resize(dataReference.size());
    memcpy(data.data(), dataReference.data(), dataReference.size());
}


The “IPC::DataReference dataReference;” just "reference" data from sendSync but not "copy" it. 
When the sendSync function pop stack, the data will be dealloc.  
So the dataReference get a  wild pointer.


Steps:

1、 Setup custom Scheme
    UCWKURLSchemeHandler *handler = [[UCWKURLSchemeHandler alloc] init];   
    [configuration setURLSchemeHandler:handler forURLScheme:@"uc"];

2、enable "Malloc Scribble"  debug Environment

3、set breakpoint at WebURLSchemeHandlerProxy::loadSynchronously  at the line  “ data.resize(dataReference.size());”

4、launch the  iOS Simultaor 

5、Send a  *Sync* XMLHttpRequest with the custom scheme

Result：
check the value of “dataReference”

You can see my Screen Shot.


This is the same result at MacOS Version

[Jin]

Created attachment 372546 [details]
Screen Shot

[Radar WebKit Bug Importer]

<rdar://problem/51934147>

[Brady Eidson]

Created attachment 372582 [details]
Patch

[Geoffrey Garen]

Comment on attachment 372582 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=372582&action=review

r=me

> Source/WebKit/WebProcess/WebPage/WebURLSchemeHandlerProxy.cpp:67
> +    data.shrink(0);

Why shrink here?

[WebKit Commit Bot]

Comment on attachment 372582 [details]
Patch

Clearing flags on attachment: 372582

Committed r246660: <https://trac.webkit.org/changeset/246660>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Brady Eidson]

(In reply to Geoffrey Garen from comment #4)
> Comment on attachment 372582 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=372582&action=review
> 
> r=me
> 
> > Source/WebKit/WebProcess/WebPage/WebURLSchemeHandlerProxy.cpp:67
> > +    data.shrink(0);
> 
> Why shrink here?

Sorry, saw the r=me in the email but missed the question.

Agreed it's unnecessary.
"Bad™" habit sanitizing out parameters gained for early CS profs.