On Mon, 27 May 2019 at 21:54, Vladimir Sitnikov <[email protected]> wrote: > > sebb> Because it is harder to review. > > It looks like as if you are the only one who wants to review.
The email thread is still young; I suspect people have not yet considered what is involved. > It is good (you care) and it is sad (no-one else cares) at the same time :-/ > > I do not care much the way we get to Git+Gradle, so let's migrate to Git > first. Great. > Could you please review https://github.com/vlsi/jmeter-git-cleanup-result ? > If that looks OK to you I would ask INFRA to use that contents for > apache/jmeter.git repository. As explained in another thread, I don't think that is a valid approach. There is no way to map SVN revisions to Git commits, also the history of deletions is lost. There may be other issues; I've not done a detailed check. I think we need to get agreement from Infra on any conversion which does not use the standard process, in case the process does not meet the requirements for provenance etc. > sebb>As I recall, Gradle requires multiple changes to the current layout, > sebb>so doing Git+Gradle together would make it much harder to check that > sebb>the SVN-Git conversion has worked. > > In fact, Gradle patch is a couple of commits on top of "migration to Git". > I don't suggest to replace Ant with Gradle through all the existing commits. > In other words, even "Git+Gradle" repository would contain "Git+Ant" commit > with old file layout. > One can easily checkout that commit and play with Git+Ant+old layout. > > That is why I see no much value in doing things one by one, however I'm > just fine with implementing svn->git first if that saves time on > conversations. > > Vladimir
