On 12 September 2013 09:03, Claudio Fontana <claudio.font...@huawei.com> wrote: > On 10.09.2013 10:45, Peter Maydell wrote: >> On 10 September 2013 09:27, Claudio Fontana <claudio.font...@huawei.com> >> wrote: >>> On another side, I end up having to manually revert some parts >>> of these which you put as prerequisites, during bisection when >>> landing after them, which is a huge time drain when tracking >>> regressions introduced in the later part of the series. >> >> I don't understand this; can you explain? If these early >> refactoring patches have bugs then we should just identify >> them and fix them. If they don't have bugs why would you >> need to manually revert parts of them?
> To revert the next patches which do introduce bugs. Huh? The next patches would apply on top of the refactoring patches, so you don't need to remove the refactoring to revert the functional changes. (On the other hand if we did things the way round you're suggesting with the functional changes first then we would need to revert or manually undo the refactoring parts in order to revert the functional change patches.) Personally I think that "first refactor/clean up, then add new features/improvements" is a fairly standard order to do things. -- PMM
