Re: Sparse Checkout Trouble (2.5.0)

[Johannes Schindelin <Johannes.Schindelin@gmx.de>]

Hi,

On Tue, 28 Jan 2020, Derrick Stolee wrote:

> On 1/28/2020 12:21 AM, Elijah Newren wrote:
> > On Sun, Jan 26, 2020 at 6:09 AM Derrick Stolee <stolee@gmail.com> wrote:
> >>
> >> On 1/25/2020 3:59 PM, Elijah Newren wrote:
> >>> On Fri, Jan 24, 2020 at 7:41 AM Derrick Stolee <stolee@gmail.com> wrote:
> >>>> I'm CC'ing Elijah because he also made changes to dir.c, and
> >>>> perhaps he has some idea of what's going on.
> >>>
> >>> If you think it might be related to the dir.c changes, I can take a
> >>> look.  I don't have any immediate ideas off the top of my head.
> >>
> >> The only thing I can think of is that these paths are already
> >> marked as sparse, but something is requiring us to test if the
> >> path can be created with the filesystem. I'll try to debug
> >> more into exactly where that is. It's telling that this happens
> >> both in cone mode and without.
> >
> > Yeah, I'll take a look.  The exponentially slow 'status --ignored'
> > report is forcing me to look at dir.c again anyway, though it's also
> > delaying me from getting a chance to look at this particular report.
>
> I made some progress, at least, in root-causing the issue.
> The problem bisects down to 4dc42c6c1 (mingw: refuse paths
> containing reserved names, 2019-12-21) [1]. CC'ing Dscho.
>
> That commit updates is_valid_win32_path() to fail on these
> paths. We were _already_ calling this method even for paths
> outside the sparse cone, but the method didn't fail for these
> examples.
>
> This means the fix is probably even more complicated: we need
> to not call this method when traversing paths that have the
> skip-worktree bit enabled. This may lead to some tiny
> performance gains when hydrating a very small fraction of a
> very large index.

Hmm. I am actually not sure that this would be a fix. It is all too easy
for a skip-worktree entry to become checked out (think e.g. a merge
conflict in a cherry-pick, during a three-way merge of a file that is not
in the cone but still needs to be handled).

My preferred solution for this issue would be for the files/directories to
be renamed using `git -c core.protectntfs=false mv <reserved-name>
<non-reserved-name>`.

I think if we try to play extra games with the skip-worktree bit, we risk
opening a vulnerability again.

Ciao,
Dscho

>
> Thanks,
> -Stolee
>
> [1] http://github.com/git/git/commit/4dc42c6c1867a52e22f1f04a1a247b5a7538b8af
>