see patch
Created attachment 65144 [details] Patch
Comment on attachment 65144 [details] Patch > - client()->didReceiveData(this, buffer, buffers.dwBufferLength, 0); > + > + if (ResourceHandleClient* resourceHandleClient = client()) > + resourceHandleClient->didReceiveData(this, buffer, buffers.dwBufferLength, 0); > buffers.dwBufferLength = bufferSize; > } > > @@ -272,8 +293,24 @@ void ResourceHandle::onRequestComplete(L > InternetCloseHandle(d->m_secondaryHandle); > InternetCloseHandle(d->m_resourceHandle); > > - client()->didFinishLoading(this); > - delete this; > + if (ResourceHandleClient* resourceHandleClient = client()) > + resourceHandleClient->didFinishLoading(this); > + > + deref(); // balances ref in start > +} These changes seem completely separate from what the bug claims to be fixing. I think you should split them into a separate bug/patch. > +String ResourceHandle::queryHTTPHeader(DWORD infoLevel) > +{ > + DWORD bufferSize = 0; > + HttpQueryInfoW(d->m_requestHandle, infoLevel, 0, &bufferSize, 0); > + > + UChar* characters; > + String result = String::createUninitialized(bufferSize / sizeof(UChar), characters); > + > + if (HttpQueryInfoW(d->m_requestHandle, infoLevel, characters, &bufferSize, 0)) > + return result.left(result.length() - 1); // Remove NullTermination. I think using a Vector<UChar> and String::adopt would be more efficient, since you won't have to copy the whole string at the end. It would be clearer to have an early return if the second call to HttpQueryInfoW fails. It might be nicer for this function to be a non-member function with internal linkage.
Created attachment 65150 [details] Patch
Comment on attachment 65150 [details] Patch r=me
Comment on attachment 65150 [details] Patch Clearing flags on attachment: 65150 Committed r65859: <http://trac.webkit.org/changeset/65859>
All reviewed patches have been landed. Closing bug.