http/tests/websocket/tests/reload-crash.html sometimes crashes on Windows. Here are some crash logs:
We're crashing beneath SocketStreamHandle::readStreamCallback. It looks like we trampoline this call over to the main thread; maybe the SocketStreamHandle has already been destroyed by the time the main thread runs the function?
Perhaps a duplicate of bug 44138?
Added to the Skipped file in r79865.
In this slightly different crash log: http://build.webkit.org/results/Windows%207%20Release%20(Tests)/r79864%20(9851)/http/tests/websocket/tests/reload-crash-crash-log.txt
...the crashing code is this:
523: length = CFReadStreamRead(m_readStream.get(), localBuffer, sizeof(localBuffer));
524: ptr = localBuffer;
> 527: m_client->didReceiveData(this, reinterpret_cast<const char*>(ptr), length);
531: case kCFStreamEventCanAcceptBytes:
*** This bug has been marked as a duplicate of bug 44138 ***