Hrvoje Niksic <hniksic(a)iskon.hr&gt; writes: Note that Zhengo's and David's are hitting different asserts. I am not sure whether that matters. Are you sure about that? To me it seems the crash happens in map_extents, before the function to be mapped is even called. My hunch would be that the hook to invalidate/adjust the bufpos->byteind cache is failing in some borderline cases. In David's case he is doing trickery with calling 'widen' in the after change functions, and modifying the buffer. . In Zhengo's case there is process data coming in (maybe while the after changes functions are being run?). Jan