2011/10/19 Henrik Nordström <henrik@xxxxxxxxxxxxxxxxxxx>: > tis 2011-10-18 klockan 22:09 -0500 skrev Dennis Gilmore: > >> So ive done some thinking, and talked with some of the stakeholders. >> Ive come to the conclusion that we should skip f16 and shoot straight >> for rawhide. > > +1 from me. It's already late to start on F16. F17 is a more realistic > goal. > > Related to this FTBFS issues needs to be tracked more closely. There > have been quite many mainline FTBFS issues seen in the F15 rebuild which > is not even fixed in rawhide. And when moving to rawhide a number of new > FTBFS issues will arise. Well the patches used for f-15 packages should have ultimately been commited to mainline F-15/16/rawhide as they were found so we wouldn't have this problem. I was doing that with all the fixes I made to F-14 so at least those will be mainline. > Just not sure if that jump from F15 to rawhide should be done by trying > moving forward in koji, or as a two stage rebuild like done for > armv5tel. Depends largely on how good the shadowing scripts are at > ordering build dependencies or how complex it is in koji to reshake > packages that need to be rebuilt because of build ordering issues > resulting in broken builds. koji-shadow is pretty good, but it can be slow (this might be incentive enough to improve it!) and we will likely end up with a chunk of f-15/16 built as part of it, but we'd end up with that anyway. F-15 was the last mass rebuild so technically there should be no need for anything before that. I think we start asap using the f-15 builds and actually get things moving forwards and fix as we go rather than messing about all over again with staging it. I have a rough build order I did in koji to get f-14 bootstrapped with F-13 and will gladly help out to push them through so we can get a decent build chain to kick of koji-shadow. > Even with the limited package churn we have had in armv7hl there have > been quite many packages with bad builds due to dependencies changing, > and quite likely is some left still. But as the mock building is all > "scratch" builds it's then only a matter or rescheduling the same builds > again. How is this handled when building in koji? koji-shadow will go down the dep chain and build deps, and dep's deps. >> I suspect there are some pieces of f16 we will need to build to get >> things right. and if we get caught up quickly and can follow the >> branching from rawhide to f17 there is no reason we cant back track >> around and do f16 i just feel that once we get f15 done we should get >> moving towards parity. and showing that arm works and can keep up and >> follow along. well if you look at f-17 its still composed of a lot of f16 and even f15 (f15 being the last mass rebuild) but the build chain is still not to far removed from F-15 in that gcc is still 4.6, python is still 2.7 etc. Basically I think we need to start with the F-15 we have and kick it off and run and fix. If we wait until its all perfect we'll never start. Peter > A double rebuild is a very powerful tool for shaking out any build > issues. You will be left pretty much with a set of good packages and a > set of FTBFS packages. And package maintainers have rason to look into > FTBFS issues. > > Regards > Henrik > > _______________________________________________ > arm mailing list > arm@xxxxxxxxxxxxxxxxxxxxxxx > https://admin.fedoraproject.org/mailman/listinfo/arm > _______________________________________________ arm mailing list arm@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/arm
