Re: Git doesn't support symlinks
On 2020-03-25 20:37:30-0400, Daniel Malendez <dmalendez@googlemail.com> wrote:
Show 26 quoted lines
> Thanks for your feedback!
>
> > On Mar 25, 2020, at 8:23 PM, Bryan Turner <bturner@atlassian.com> wrote:
> >
> > This didn't add the same thing. What does a "git status" show at this
> > point? I'd expect it would show "Versions/Current/Headers/interior.h"
> > as added to the index, not "Headers/interior.h".
>
> $ git add .
> $ git status
> Changes to be committed:
> (use "git rm --cached <file>..." to unstage)
> new file: Foo.framework/Headers
> new file: Foo.framework/Versions/A/Headers/Interior.h
>
> > It doesn't necessarily seem like a bug, to me; more like a case where
> > Git could potentially be "smarter" to try and determine that, while
> > "Headers/interior.h" is beyond a symlink, the target of that symlink
> > is still in the repository. (Versions/Current _is_ in the same
> > repository, right?)
>
> Agree, makes sense!
>
> I think what happens here is that `git add .` adds the
> Foo.framework/Headers directory first, to circumvent this check
> failing here?
"Foo.framework/Headers" is a symlinks, and git simply adds that, without caring where it's linked to.
To git, there's nothing inside "Foo.framework/Headers", it's a blob, which happends to be type: symlink.
IOW, you can:
$ git add Foo.framework/Headers # without add the target
$ ln -s /usr/bin bin
$ git add bin # git won't write to index anything from /usr/bin
> https://github.com/git/git/blob/master/pathspec.c#L596
> Whereas a single “git add path/to/file” would fail here
--
Danh