Renee Danson Sommerfeld wrote: > On Thu, Nov 05, 2009 at 07:25:34PM +0000, Alan Maguire wrote: > >> Renee Danson Sommerfeld wrote: >> >>> On Thu, Nov 05, 2009 at 05:37:30PM +0000, Alan Maguire wrote: >>> >>> >>>> I still need to test 12493 but I wanted to get the webrev out >>>> for review ASAP: >>>> >>>> http://zhadum.east/export/ws/amaguire/nwam1-bugs/webrev >>>> >>>> >>> Fixes for 12462 and 12472 look good. >>> > > Not sure where you are with builds and testing, but I'd say these > two fixes are ready to go back as soon as you're ready. I'd rather > not have them waiting on resolution of the bfu questions. > > >>> Some questions about 12493: In addition to adding those services, we >>> also add dependencies on those services (nwam depends on netcfg, many >>> services depend on location). What happens to those dependencies when >>> we remove the services? >>> >>> I'm thinking the ones that depend on location might be okay, because >>> they're added as dependents of the location service, so removing that >>> service removes those dependents. But I think nwamd has the dependency >>> on netcfg, so just removing netcfg won't change that. >>> >>> >>> >> What about making the dependency nwam has on netcfg >> a dependent in netcfg.xml? >> > > I suppose that would be okay. > > >>> And for that matter, what happens to the nwam manifest? The phase 1 >>> manifest is quite different from the earlier one. >>> >>> >>> >> Good point - I'll check on that when my build completes. >> > > Keep me posted. But one caution: I don't think backwards-bfu'ing is > sufficiently important to merit a lot of your time. If there are > quick easy things you can do to make it work better, okay; but don't > slave away at this. > > okay - I've got this to a pretty good state. If you're running NWAM phase 1, you can back-bfu successfully to pre-phase 1 NWAM with nwamd running. We need to explicitly import the network-physical.xml manifest prior to reboot to ensure we don't have to wait for manifest-import (this prevents nwam from going into maintenance on first reboot post-bfu). The new codepaths added to bfu.sh are only activated in cases where we're backbfu'ing so I'd consider these changes low-risk. Webrev respun, and I'll push later if there's no objections. Thanks!
Alan
