There has been some speculation that I have some evil plans with the recent changes. Not so.
I just made a few wrong assumptions about how this would pan out, planned badly and made a few crucial mistakes in execution. Poof! Again: sorry for the mess. Damage done, I'm trying best as I can to clean up as quickly as possible. There are no known regressions in svn head as of writing(except that Magnus reported an assert that he can't reproduce yet). Since I want to give the patient a bit of rest before I start committing performance fixes, I'm currently profiling & staying ready to jump in and address any problems you guys are suffering due to my latest brutal experimentation on svn head. The ZY1000 went from 407 to 307kBytes/s upload on ARM7, so rest assured that I've got no evil motives in bringing down the performance of USB vs. embedded host and that I'll attend to the performance problems as soon as the patient is well enough to be operated on. Profiling meanwhile... :-) I started this in_handler stuff not because I have some ulterior motive, but because I had idle time on my hands this week (not anymore!), and then I started on this pet project to clean things up once and for all. We have a major new testing round planned(we've got lots of target boards and several interfaces that we are going to test on here). Thanks to everybody who has contributed targets! I'm postponing that large regression testing round(it takes time!) until Zach has finished his pile of patches and things seem generally quiet. I really do believe that the OpenOCD svn head will calm down *a lot* once this huge round of changes & cleanup is done. Also, I believe that OpenOCD is only it is small beginnings and that it is important that we take this unique oportunity to clean things up now. Once the user crowd grows, this will be harder and harder. I'm looking forward to seing what the result will be of Zach's solitiation of what policies to use for OpenOCD. If there is a single positive aspect of this whole in_handler mess, it's that an OpenOCD policy probably will be set that is much more orderly than what we have today. Also I believe that the group will be much more appreciative of the extra time & patience such a policy requires. Again: sorry for the mess. I'm now trying to clean it all up and things are looking much better than they did just hours ago. -- Øyvind Harboe Embedded software and hardware consulting services http://consulting.zylin.com _______________________________________________ Openocd-development mailing list [email protected] https://lists.berlios.de/mailman/listinfo/openocd-development
