git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: git stash data loss

From
Jeff King <peff@peff.net>
Date
Jul 27, 2012, 13:29 UTC
Message-ID
<20120727132953.GA8263@sigill.intra.peff.net>
In-Reply-To
<55831343358368@web1e.yandex.ru>
On Fri, Jul 27, 2012 at 07:06:08AM +0400, Aleksandr Pryimak wrote:
Show 10 quoted lines
> I also recreated it
> 
> aleksandr@beast:/tmp/test$ git init
> Initialized empty Git repository in /tmp/test/.git/
> aleksandr@beast:/tmp/test$ touch x
> aleksandr@beast:/tmp/test$ git add x
> aleksandr@beast:/tmp/test$ git commit -m "Initial"
> [master (root-commit) d3569a0] Initial
>  0 files changed, 0 insertions(+), 0 deletions(-)
>  create mode 100644 x
OK, so we have "x" as a tracked file.
> aleksandr@beast:/tmp/test$ rm x
> aleksandr@beast:/tmp/test$ mkdir x/
> aleksandr@beast:/tmp/test$ ls
> x

And then we remove it in favor of a directory. Note that git does not track directories directly, only files inside them. So from git's perspective, the working tree has no content in it at all.

> aleksandr@beast:/tmp/test$ git stash
> Saved working directory and index state WIP on master: d3569a0 Initial
> HEAD is now at d3569a0 Initial

And now we stash it. That will stash the working tree removal of x. It will not stash anything about the new directory x, because it has no content inside it.

Show 14 quoted lines
> aleksandr@beast:/tmp/test$ ls
> x
> aleksandr@beast:/tmp/test$ git stash pop
> Removing x
> # On branch master
> # Changes not staged for commit:
> #   (use "git add/rm <file>..." to update what will be committed)
> #   (use "git checkout -- <file>..." to discard changes in working directory)
> #
> #	deleted:    x
> #
> no changes added to commit (use "git add" and/or "git commit -a")
> Dropped refs/stash@{0} (c500443ae16cf0d846b195cb97eb388dec5f440e)
> aleksandr@beast:/tmp/test$ ls

And your stash pop restores the deletion of "x". It does _not_ reinstate the directory "x", because as I stated above, that is not part of the stash.

So what is the data loss? That the "mkdir x" was lost? That has nothing
to do with stash, but rather the fact that git does not track empty
directories. You could see the same thing with:
 
  mkdir x && git commit -a

which will not record your mkdir at all. In other words, this is by design.

If we put actual files inside "x", which git does track, then they would be part of the stash, and should be properly retained. But they're not:

  $ rm x && mkdir x && echo foo >x/file

Now we have some precious contents in the form of "x/file". They are untracked by git, but git should be careful about removing them.

  $ git stash
  Saved working directory and index state WIP on master: 2d32d3a initial
  HEAD is now at 2d32d3a initial
  $ ls -l x
  -rw-r--r-- 1 peff peff 0 Jul 27 09:19 x
  $ git stash show --raw
  :100644 000000 e69de29... 0000000... D  x

Now this _is_ data loss. Stash blows away untracked files inside the directory, but does not record them in the resulting stash. And that should be fixed.

With "-u", I'd expect stash to include the untracked file x/foo in the stash. Without "-u", I'd expect stash to fail, since it would be deleting untracked files.

-Peff
Previous: Aleksandr PryimakNext: Jeff King
Message 2 of 7 in “git stash data loss”
  1. Aleksandr PryimakJul 27, 2012
  2. Jeff KingJul 27, 2012
  3. Jeff KingJul 27, 2012
  4. Junio C HamanoAug 17, 2012
  5. Jeff KingAug 21, 2012
  6. Junio C HamanoSep 14, 2012
  7. Aleksandr PriymakJul 27, 2012

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.