Hello, I have been thinking about this yesterday night and I think the solution already was already suggested in the forum thread: http://forum.sparkfun.com/viewtopic.php?t=7841&postdays=0&postorder=asc&start=45&sid=45354d14976fd68c1fa088693f66ed7f
The fact is that openOCD tries to validate the JTAG chain without looking at the config file, only using the JTAG standards. This is done in the jtag_validate_chain function. The fact is that this function fails with the i.MX31 because the third device on the inner chain is not compliant with the JTAG specification. If we skip this validation stage, the further steps seem to work with i.MX31 (jtag_examine_chain succeeds afterwards, tested right now). There could be an easy solution, to give an option in the config file so that this step is skipped in the initialisation process. But this is maybe an overkill, since in our case, only the third device on the chain is not compliant. The second option, would be that for any device in the chain, we could give an option that this device should not be validated. The problem is that according to Dominic on the forum thread, the jtag_validate_chain does not look at the config file (it is not its task). But I cannot see another way to support, as Dominic says it, broken designs like i.MX31. Since this touches to deep openOCD internals, we should discuss on how to implement this on the list before we try to implement something. Feedback very appreciated. Regards Valentin _______________________________________________ Openocd-development mailing list Openocd-development@lists.berlios.de https://lists.berlios.de/mailman/listinfo/openocd-development
