On 05/30/2012 06:48 PM, Lucas Meneghel Rodrigues wrote: > On Wed, 2012-05-30 at 14:41 -0700, Nishanth Aravamudan wrote: >> On 30.05.2012 [18:26:58 -0300], Lucas Meneghel Rodrigues wrote: >>> On Wed, 2012-05-30 at 14:23 -0700, Nishanth Aravamudan wrote: >>>> On 26.05.2012 [18:03:31 -0300], Lucas Meneghel Rodrigues wrote: >>>>> On Fri, 2012-05-25 at 16:31 -0700, Nishanth Aravamudan wrote: >>>>>> Sometimes there are firewalls between test machines and the greater >>>>>> Internet, so it seems unwise to depend upon external network access in >>>>>> the boottool/grubby building code. But some sites won't have those >>>>>> restrictions. Add the ability to configure a local mirror for the grubby >>>>>> tarball in the CLIENT section, but default to the external location. >>>>> The problem with this patch is that makes boottool dependent on autotest >>>>> libraries, when the script itself is sometimes used in a stand alone >>>>> fashion. Therefore, I can't accept this as is. >>>> Ah makes sense. I didn't realize boottool was used stand-alone, sorry. >>>> >>>>> Cleber, I believe we should try to locate and download boottool from the >>>>> copy present in the autotest tree, before trying to reach out to github. >>>>> What do you say? >>>> Ah I didn't even realize there was one in client/deps/grubby -- so, >>>> would we try and push it out with the rest of autotest? I'm not sure >>>> pulling will work, as the client/deps/grubby path isn't guaranteed (nor >>>> is it setup to be, afaict) part of the web-exposed path. >>> This is what I'd like to do, find a way to ensure the grubby tarball >>> gets copied when the client is installed. This way we wouldn't ever have >>> to resort to an external copy.
Well, did you guys miss the response I posted a couple of days ago? Copying it again: --- On client mode that is definitely the best thing to do. But that would fail on server mode. I suggest that boottool looks for the grubby tarball on those locations: 1) current directory (would solve server mode if we also send the tarball to the client) 2) autotest source tree 3) remote github uri How does that sound? --- So, adding that to the rsync'd path list sounds like implementing #2. #1 is still needed, and number #3 is a fallback that may be skipped. Are we all on the same page here? Cheers, CR. >> Sounds like a plan -- in theory could we simply add it to the rsync'd >> path(s)? >> >> And then update the grubby code to look locally only? > Yes, and yes! > >> This also makes it easier to keep things in-sync -- rather than relying >> on either a local or remote mirror to have the appropriate version of >> the tarball, we can ensure version updates and the tarball updates >> happen in the same commit. > I agree, it'll make our lives easier. > >> Thanks, >> Nish >> > _______________________________________________ Autotest mailing list Autotest@test.kernel.org http://test.kernel.org/cgi-bin/mailman/listinfo/autotest
