Dustin J. Mitchell wrote on 19.07.2010 03:00: > Any results? I'd like to get this committed shortly so that it can > get into the 3.1.2 release.
Sorry for not posting any results, I was out of the office a few days. The change you suggested doesn't fix the problem, but when I reduce the split_diskbuffer to 1 GB, it works. I think, the problem might be the component 'gsize slice_remaining' of 'struct slab_source_state'. As I'm running a 32-bit system, gsize is a 32-bit value as far as I know. I'm currently running a test with slice_remaining defined as guint64 but it will take a few hours until the first tape change occurs and I see any results. Jens
