On Thursday 24 September 2009, Wookey wrote: > So, if anyone has any bright ideas for fixing the underlying xsvf > problem so we can move to newer code or why it should be different on > i386.
Actually, I do have an idea about the XSVF thing; but not why 32-bit loses for you. Short version of the story ... look at the XSVF spec for XSTATE; compare to the implementation. That's not what it's doing! The fix there will be more or less to call the SVF code for any multiple-TCK transition (other than to RESET). The particular paths to use are absolutely-the-shortest. What it does now involves using the "new" (not so much, any more) "short" (not really) tables. Not what SVF does. ISTR you had another problem too, but that much seems clearly wrong and potentially related. Secondary issue, the SVF code doesn't quite do what it ought to do either. That's more easily fixed. Working on patches... - Dave _______________________________________________ Openocd-development mailing list Openocd-development@lists.berlios.de https://lists.berlios.de/mailman/listinfo/openocd-development
