I’ve managed to get OpenOCD 0.1.0 working on Windows, arm-elf-dgb 6.8, Turtelizer2 and AT91SAM7X256 board, but it is unstable. For example after stopping on hardware breakpoint I am able to make only couple of steps through program and then OpenOCD crashes with segmentation fault. It is when I use it in remote TCP target mode. When I try to use it in pipe mode calling starting it by
target remote | openocd -f file.cfg --pipe
it fails to allocate buffer on Windows heap and again crashes.
Any suggestions what can be wrong? Are older versions of OpenOCD more stable?
In general openocd tends to become more stable with time, and 0.1.0 was created to avoid the most experimental features.
It is often really usable but the 0.1.0 release number says “no guarantees whatsoever”.
It should not segfault when stepping, but there are known problems when single stepping code, especially in Thumb mode that has been corrected since 0.1.0 release. There is a release 0.2.0 coming out anytime now, it will be more stable but surely with uresolved problems.
I would not suggest using the pipe mode under windows if it does not work at first try. This type of tricks are very UNIX based, it is possible but tricky and requires a lot of testing to do the same under windows
I’ve tried last night to use OpenOCD build from SVN 1690 revision. It proved to be much more stable. I mean there was no segmentation fault and it was doing what it was supposed to do. Of course the testing lasted only approx. 1 hour, but as I mentioned no problems found so far.
It seems that the bug I encountered before was fixed in newer revision. So I wonder if you still need a debug log. Please let me know and if it is still needed, I will revert openocd and provide you the log.