[ Sent upstream: https://gitlab.com/gitlab-org/gitlab-shell/merge_requests/40

  This was filed upstream but decided the better way to go is to
  configure default git templates to be just empty.

  Until we implement that, we continue to carry this patch as NXD ]

gitlab-shell uses completely own whole set of git hooks and just
symlinks repo.git/hooks to path/to/gitlab-shell/hooks/.

Original hooks that git added there - when initializing repository -
from its template, and which are by default deactivated, are currently
moved to repos.git/hooks.old.<timestamp>, e.g.

$ ls -F nexedi/helloweb.git/
    config
    description
    HEAD
    hooks@                  -->  ...path/to/gitlab-shell/hooks/
    hooks.old.1446564288/
    info/
    packed-refs
    refs/

$ ls -F nexedi/helloweb.git/hooks.old.1446564288/
    applypatch-msg.sample*
    post-update.sample*
    pre-commit.sample*
    pre-push.sample*
    update.sample*
    commit-msg.sample*
    pre-applypatch.sample*
    prepare-commit-msg.sample*
    pre-rebase.sample*

On the other hand, hooks.old.<timestamp> are not used anywhere and are
left there "just in case". But what is the point to keep those pristine
samples git creates by default? And we are sure those files are always
results of gitlab-shell initializing a repository, as it is
gitlab-shell's job to create/remove/update/etc repositories.

So do not keep old hooks, remove them.

P.S. The patch that introduced whole-hooks-dir-symlinking is
    db81327b (Symlink the whole hooks directory).

/cc @dzaporozhets, @jacobvosmaer

Pass relevant git environment variables while calling `/allowed`

1. Starting version 2.11, git changed the way the pre-receive flow works.
  - Previously, the new potential objects would be added to the main repo. If the pre-receive passes, the new objects stay in the repo but are linked up. If the pre-receive fails, the new objects stay orphaned in the repo, and are cleaned up during the next `git gc`.
  - In 2.11, the new potential objects are added to a temporary "alternate object directory", that git creates for this purpose. If the pre-receive passes, the objects from the alternate object directory are migrated to the main repo. If the pre-receive fails the alternate object directory is simply deleted.
2. In our workflow, the pre-recieve script calls the `/allowed` endpoint on the
   rails server. This `/allowed` endpoint calls out directly to git to perform
   various checks. These direct calls to git do _not_ have the necessary
   environment variables set which allow access to the "alternate object
   directory" (explained above). Therefore these calls to git are not able to
   access any of the new potential objects to be added during this push.

3. We fix this by passing the relevant environment variables
   (`GIT_ALTERNATE_OBJECT_DIRECTORIES`, `GIT_OBJECT_DIRECTORY`, and
   `GIT_QUARANTINE_PATH`) to the `/allowed` endpoint, which will then include
   these environment variables while calling out to git.

---

- Related to gitlab-org/gitlab-ce#25301. 
- Corresponding backend MR: gitlab-org/gitlab-ce!7967
- Corresponding EE MR: gitlab-org/gitlab-ee!964

See merge request !112

1. Starting version 2.11, git changed the way the pre-receive flow works.

  - Previously, the new potential objects would be added to the main repo. If the
    pre-receive passes, the new objects stay in the repo but are linked up. If
    the pre-receive fails, the new objects stay orphaned in the repo, and are
    cleaned up during the next `git gc`.

  - In 2.11, the new potential objects are added to a temporary "alternate object
    directory", that git creates for this purpose. If the pre-receive passes, the
    objects from the alternate object directory are migrated to the main repo. If
    the pre-receive fails the alternate object directory is simply deleted.

2. In our workflow, the pre-recieve script calls the `/allowed` endpoint on the
   rails server. This `/allowed` endpoint calls out directly to git to perform
   various checks. These direct calls to git do _not_ have the necessary
   environment variables set which allow access to the "alternate object
   directory" (explained above). Therefore these calls to git are not able to
   access any of the new potential objects to be added during this push.

3. We fix this by passing the relevant environment variables
   (GIT_ALTERNATE_OBJECT_DIRECTORIES, GIT_OBJECT_DIRECTORY, and
   GIT_QUARANTINE_PATH) to the `/allowed` endpoint, which will then include
   these environment variables while calling out to git.

Make custom hook dir configurable

Add a new configuration option, custom_hook_dir. When this is set, we
will look for global custom hooks in: `<custom_hook_dir>/{pre-receive,update,post-receive}.d/*`

When this is not set, default to `<ROOT_PATH>/hooks`.

Relates to https://gitlab.com/gitlab-org/omnibus-gitlab/issues/1754 and
gitlab-org/gitlab-ce!8040.

See merge request !113

Add a new configuration option, custom_hooks_dir. When this is set, we
will look for global custom hooks in:
    <custom_hooks_dir>/{pre-receive,update,post-receive}.d/*

When this is not set, default to <REPO_PATH>/hooks.

Make merge request text after push clearer

The message to create a merge request after pushing a new branch was not clear. It is now clear that it is optional to create a merge request for the pushed branch.

Part of [#21451](https://gitlab.com/gitlab-org/gitlab-ce/issues/21451)

See merge request !109

Add changelog entry

Editing changelog entry version

Chained global hooks

Closes #32. Docs MR: https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/6721

See merge request !111

this fixes problem that tests succeeded locally but failed in ci
where parent dirs were missing

update hooks lookup to use <hook>.d/* from repository hooks dir

the order would be:

1. <repository>.git/custom_hooks/<hook_name> - per project hook
2. <repository>.git/custom_hooks/<hook_name>.d/* - per project hooks
3. <repository>.git/hooks/<hook_name>.d/* - global hooks

only executable files are matched and backup files excluded (*~)
and the resulting list is sorted per each lookup

This commit adds the option of having another set of global custom hooks
along with the already supported repository local custom hooks.

The repository local custom hook is executed first (if available). If
successful, execution continues with the global custom hook (if available).
This way, local custom hooks get priority over global custom hooks.

Global custom hooks can be enabled by placing an executable file into the
"custom_hooks" directory within gitlab-shell (create if it does not exist,
yet).

This commit takes the GitlabCustomHook a bit clother to the other hook
handling classes by receiving the repo_path as argument to initialize()
instead of passing it to each method.

This commit fixes an issue where an existing but unexecutable hook would
cause an uncaught execption.

Fetch repositories with --prune option by default.

This will allow us to not recreate branches that has been deleted on the upstream and the mirror.

Related issue: https://gitlab.com/gitlab-org/gitlab-ee/issues/1277

I've reviewed the code for updating local and remote mirrors also  I've manually tested these features and all looks good.

Here is the description of the `--prune` option:


> --prune
>
> Before fetching, remove any remote-tracking references that no longer exist on the remote. Tags are not subject to pruning if they are fetched only because of the default tag auto-following or due to a --tags option. However, if tags are fetched due to an explicit refspec (either on the command line or in the remote configuration, for example if the remote was cloned with the --mirror option), then they are also subject to 


@brodock do you see any issues when fetching repos for GitLab Geo?

See merge request !110

This will allow us to not recreate branches that has been deleted on the upstream and the mirror.

Fix gitlab_custom_hook dependencies

See merge request !107

Add instrumentation to push hooks

See merge request !106