08:23 <PaulFertser> maribu[m]: hm, do you probably have some bootloader mode on that target?

08:23 <PaulFertser> Sometimes it's easier to switch to bootloader mode and attach from there.

08:24 <PaulFertser> Do you actually connect system reset line?

08:28 <maribu[m]> No, I have not flashed any bootloader or changed the bootloader mode. I am using the CC2650-Launchpad with an integrated XDS110 debugger and a CC2650STK ("Sensor Tag") with the XDS110 "daughter board". The reset line is therefore wired internally for the launchpad, and connected via the same connector as JTAG for the sensor tag. There is little room for wiring error in this setup.

08:29 <maribu[m]> Note that the CC2650 has this ICEPick JTAG router that is also reset when the reset signal is applied. I think that after each reset the ICEPick JTAG router needs to be re-configured and the switch from cJTAG to JTAG needs to be reperformed.

08:32 <maribu[m]> I get JTAG protocol errors (invalid ACKs) as symptom if I change the reset_config from none to anything else. With none, it does work, but the debugger fails to attach to the MCU once I have flashed a RIOT based firmware (that does make use of power saving modes). It is a known issue that due to power saving modes reliable attaching the debugger only works when connecting under reset with RIOT.

08:34 <maribu[m]> The "solution" other RIOT users use is to pull all the jumpers connecting the CC2650 to the XDS110, connect a J-Link via the 10 pin 1.27 mm standard JTAG connector, and use JLinkExe to connect under reset...

08:35 <PaulFertser> maribu[m]: I mean MCUs usually have some ROM bootloader that can be activated with external straps.

08:37 <PaulFertser> Probably icepick can hold the target in reset but I do not see it mentioned in existing configs.

08:38 <maribu[m]> OK, I see. This would help as I could then attach without the reset signal. However, one of my use cases is to automatically iterate of ~100 firmwares containing integration tests of RIOT. This requires a fully automated flashing procedure.

08:53 <PaulFertser> maribu[m]: understandable. Probably debug logs from jlink software would tell more about how exactly it's doing that.

09:01 <karlp> this sounds familiar, but I can't remember if there was a resolution or not last time either :|

09:06 <PaulFertser> Probably someone needs to dig icepick docs.

09:06 <karlp> we don't have an awful lot of ti users though.