Flashing is not working on SAMR21 any more

[OlegHahm]

As stated in #2408 (comment) make flash (using the unified OpenOCD script from #2408) is currently not working for the SAMR21 - at least on ArchLinux. (Neither with the AUR version of OpenOCD nor with a fresh compiled from current git HEAD.) Putting an exit at the end of the program command or adding a shutdown command at the end helps, but leads to 1 as an exit code indicating an error - but I don't know why, because flashing seems to work correctly.

[OlegHahm]

@thomaseichinger, the full output with -d3 is very long. The last lines look like

User : 29351 21254 command.c:763 jim_echo(): ** Verified OK **
Debug: 29352 21254 command.c:145 script_debug(): command - ocd_command ocd_command type ocd_shutdown
Debug: 29353 21255 command.c:145 script_debug(): command - shutdown ocd_shutdown
Debug: 29354 21255 adi_v5_cmsis_dap.c:129 cmsis_dap_queue_ap_read(): reg = 16
Debug: 29355 21256 adi_v5_cmsis_dap.c:161 cmsis_dap_run():  
Debug: 29356 21256 adi_v5_cmsis_dap.c:83 cmsis_dap_queue_dp_read(): reg = 4
User : 29358 21256 server.c:588 handle_shutdown_command(): shutdown command invoked
.-(~/git/RIOT/tests/bloom_bytes)-------------------------------------------------------------------------------(oleg@zoidberg)-
`--> echo $?
1      

[haukepetersen]

to resume the discussion from #2408: i removed the -c shutdown after -c reset run, as it lead to duplicate shutdown on most boards. As example the stm32f4discovery with -c shutdown:

...
** Verified OK **
shutdown command invoked
shutdown command invoked

without -cshutdown`:

...
** Verified OK **
shutdown command invoked

@thomaseichinger testsed for the samr21 - does openOcd maybe behave different between Linux and OSX?

[OlegHahm]

Does it hurt to shutdown twice?

[LudwigKnuepfer]

Can't we pass arguments/additional commands nowadays?

[haukepetersen]

Does it hurt to shutdown twice?

It's does not really that I know of, but its not clean either...

By the way, this is the output when flashing the samr21-xpro without the -c shutdown on my computer:

...
** Programming Finished **
** Verify Started **
verified 49832 bytes in 4.127607s (11.790 KiB/s)
** Verified OK **
shutdown command invoked

So it seems it has something to do with different OpenOCD versions? My version is Open On-Chip Debugger 0.9.0-dev-00186-g30203b3 (2014-11-18-16:43) build from their sources.

[haukepetersen]

just updated openOCD -> Open On-Chip Debugger 0.9.0-dev-00301-g986102f (2015-02-19-16:37)

now I have the same behavior as @OlegHahm. Also flashing the stm32f4discovery seems to be broken: Error: jtag status contains invalid mode value - communication failure after successful verification...

[OlegHahm]

and via versa: reverted to OpenOCD version 0.9.0-dev-00186-g30203b3 and it works like charm again.

[LudwigKnuepfer]

🍌 software ;)

[haukepetersen]

ok, my fault. Seems like OpenOCD changed something. I guess #2471 should fix it for everyone?

[jnohlgard]

@LudwigOrtmann

Can't we pass arguments/additional commands nowadays?

Only extra initialization since the OPENOCD_EXTRA_INIT argument is put before all of the commands on the command line.

[haukepetersen]

and the forget about the error for the stm board stated above, that was because I forgot to add the backslash after the -c reset halt command...

[jnohlgard]

@LudwigOrtmann @haukepetersen How about we add argument handling to the script?
That will allow us to properly quote the arguments which is not possible via environment variables unless we use some eval function or similar on the variables.

[LudwigKnuepfer]

+1 for argument handling

[PeterKietzmann]

@gebart is this issue fixed in #2487? Can we close it?

[jnohlgard]

@PeterKietzmann I don't have a SAMR21 board to test with.
@LudwigOrtmann : Can you verify this is working now?

[PeterKietzmann]

I do have one but didn't want to go into detail here. AFAIK just flashing works, will verify soon. What should I especially have a look at/test?

[OlegHahm]

Before #2471 or #2472 are merged the problem isn't solved.

[PeterKietzmann]

Ok, good that these PRs are referenced now so we may not forget to close the issue after merging