Am 06/11/2011 09:55 AM, schrieb Freddie Chopin: > On 2011-06-10 22:06, Peter Stuge wrote: >> And that just because the code does not have information, it should >> not make guesses. > > Restoring initial state of FT2232 interface is just like "pulling a > plug" - disconnecting JTAG from board. I really see no danger in such > operation... Physical disconnection can be even more random (some > signals get disconnected sooner than other etc.) and noone sees that > as a hazardous situation. Um - you mean with power applied to the target?
That *is* hazardous. It may work most of the time, but it may also lead to latch-up, frying the JTAG dongle or the target. Interfaces that allow hot-plug make special considerations to make that safe - including longer pins for GND, and some kind of extra protection on the data pins. CMOS signals on a 20-pin IDC connector are not safe for hotplug. > Most probably some of you are trying to be uber-correct without any > reason. The target will handle such SINGLE transient state without any > problems - no risk of erasing flash or firing a nuke. It will most > probably start running after reset and that's what we want. If you > want it in some specific state, just leave OpenOCD running - what's > the problem with that approach? I could live with that (like with other minor quirks), but I don't see the advantage with leaving the target in an undefined, unknown state. Since it was argued that this is the only "correct" way to exit OpenOCD, I need to object. I try to first design stuff in a way that is correct, and only back off and implement shortcuts when there is some benefit of doing so. cu Michael _______________________________________________ Openocd-development mailing list [email protected] https://lists.berlios.de/mailman/listinfo/openocd-development
