Hi Junio, On Fri, 29 May 2020, Junio C Hamano wrote: > Johannes Schindelin <Johannes.Schindelin@xxxxxx> writes: > > > Dear Git users, > > > > On Tue, 26 May 2020, Johannes Schindelin wrote: > > > >> I hereby announce that Git for Windows 2.27.0-rc2 is available from: > >> > >> https://github.com/git-for-windows/git/releases/tag/v2.27.0-rc2.windows.1 > > > > Git for Windows v2.27.0 will be out on Tuesday (or even on Monday if Git > > v2.27.0 is released early enough). It has a (last-minute) bug fix where > > the length of symbolic links that is recorded in the Git index is the > > _actual_ length of the link target, so that `git status` agrees between > > Git for Windows and WSL Git. > > > > Unfortunately, that means that every existing checkout that contains > > symbolic links tracked by Git will report them as changed until `git add > > <path-to-symlink>` (or `git add -u`) is called. > > So, this is not something "git update-index --refresh" would be able > to fix? I just tested this, and no, it does not fix the issue. The bug in question records the length of _any_ symbolic link as 4096 in the index, and an `update-index --refresh` with a fixed version of Git apparently does not update that (the contents did _not_ change, after all). Ciao, Dscho > > > > > In cases where users use both Git Bash and another version of Git, this > > might cause the problem where one of them will always report the symbolic > > links as modified in `git status`, but not in `git diff`. > > > > Sorry about the late heads-up, I only just realized that this might pose a > > problem. Granted, symbolic links are not all _that_ common on Windows. > > > > Ciao, > > Johannes >
