On Wed, Dec 25, 2019 at 01:50:14PM -0500, Eric S. Raymond wrote: > Segher Boessenkool <seg...@kernel.crashing.org>: > > Or doing what everyone else does: put an empty .gitignore file in > > otherwise empty directories. > > That is an ugly kludge that I will have no part of whatsoever. > > Conversion artifacts like this
It's not a conversion artifact. It's what people do if for whatever reason they want to commit an empty directory. It works, it is simple, it doesn't conflict with other things, and above all, it is the common way of handling this. Of course since (as Joseph notes) we do not really care about having the empty directories here at all, it is moot anyway. > the goal of a repository converter The only goal *we* (GCC) have is *one* converted repository. > The ideal should be to produce a converted history that looks as much > as possible like it has always been under the target > system. I don't know anyone who wants that, either. Why would that be useful? > Developers should have no need to know or care that the > history used to be managed differently Wait until they find out about changelogs. A *much* bigger change, and it will happen *later*, some time after the conversion to Git. Since historical commit messages are pretty much non-existent (to say it nicely), and it takes some sleuthing to line that up with the ML discussions (which is the important part of archaeology!), the only really useful part of most historical commits is the actual file changes. Which we hopefully have perfectly fine already, in all candidate conversions. Segher