I think this is happening during teardown in some protocol tests, though I can't exactly pinpoint the conditions for reproducing the crash. It has not been appearing on the bots, but I have run into it hard in my current work on <https://webkit.org/b/147934> NOTES: The WKTR harness automatically calls WKInspectorClose at the end of a test run, but the real inspector may not have been opened at that point. Regardless, it sends a WebInspector::Close message on the frontend connection, which may have not been initialized. I think this can be solved with an early exit in WebInspector::closeFrontend().
<rdar://problem/18545278>
Created attachment 258846 [details] Proposed Fix
Adding this as a blocking bug, since it seems to be triggered by the refactoring.
Comment on attachment 258846 [details] Proposed Fix r=me
Comment on attachment 258846 [details] Proposed Fix Clearing flags on attachment: 258846 Committed r188362: <http://trac.webkit.org/changeset/188362>
All reviewed patches have been landed. Closing bug.