Merge branch 'user-preferences-layout-option' of https://github.com/gopeter/gitlabhq into gopeter-user-preferences-layout-option
Signed-off-by: Dmitriy Zaporozhets <dmitriy.zaporozhets@gmail.com>

Refactor Ci::Commit and Ci::Build to have all builds for same :sha on single page

This makes Ci::Commit to have only :sha and simplifies routing to have only :sha in path. The :ref and :push_data is now parameter of Ci::Build.

All commit related data (git author, message and .gitlab-ci.yml) is read directly from repository.

All code related for creating builds is moved to CreateBuildsService.

Status deduction is rewritten to make if more efficient and easier to integrate with Commit Status API.

This is partially working, tests are not yet touched.

This slightly changes view of Commit:
![Screen_Shot_2015-10-02_at_15.21.47](https://gitlab.com/gitlab-org/gitlab-ce/uploads/ad3f1ccdcc87659ea437d8db6c5b9f94/Screen_Shot_2015-10-02_at_15.21.47.png)

@dzaporozhets What do you think?


See merge request !1502

Fix referals for :back and relative URL installs.

`app/views/layouts/_head.html.haml` currently sets the referrer policy to "origin". This disables the use of `{ redirect_to :back }`, which is used frequently throughout GitLab, and also breaks relative URL installs for browsers that obey this policy (e.g., Chrome).

For example, using the "origin" policy the "Save Changes" button on the profile screen will (a) dump you at the GitLab home screen (which I can confirm on GitLab.com) or (b) dump you out of GitLab entirely for a relative URL install (which I can confirm on my local install).

The right policy is "origin-when-cross-origin", which disables HTTP_REFERER when coming from other sites but enables it internally. (https://w3c.github.io/webappsec-referrer-policy/) I can confirm that this fixes the problems with ":back" on my install.

See merge request !1505

Document known issue with Reply by email and multiple application servers.

See merge request !1499

Revert "Merge branch 'projects' into 'master'"

This reverts commit 2b493695, reversing
changes made to b5c12f74.

This reverts !1487 MR because it cause bugs on mobile devices 


cc @skyruler 

See merge request !1509

This ensures that blocks defines using "benchmark_subject" have access
to methods defined using let/subject & friends.

Basic RSpec/benchmark-ips powered benchmark suite

Corresponding issue: #2909, see the commit messages for more details.

A few things to note:

1. The current use of `subject` isn't exactly easy on the eyes due to them having to return a Proc, I'm not sure yet how (and if) we can work around this.
2. The maximum amount of iterations in the current `User.by_login` benchmark is arbitrary, we might have to adjust it once said method's performance has been improved.
3. Benchmarks currently take 2 seconds to warm up and 5 seconds to run (benchmark-ips defaults). 
4. The custom RSpec matcher file (`benchmark_matchers.rb`) is a bit messy, any feedback on this would be appreciated

Any comments/feedback on this would be greatly appreciated.

See merge request !1503

Fix anchors to comments in diffs

https://gitlab.com/gitlab-org/gitlab-ce/issues/2218

See merge request !1508

Golang download instructions assume amd64



See merge request !1488

This reverts commit 2b493695, reversing
changes made to b5c12f74.

This will be disallowed again once the existing benchmarks pass (which
relies on #2341).

Signed-off-by: Dmitriy Zaporozhets <dmitriy.zaporozhets@gmail.com>

This class method can be used in "describe" blocks to specify the
subject of a benchmark. This lets you write:

    benchmark_subject { Foo }

instead of:

    benchmark_subject { -> { Foo } }

Signed-off-by: Dmitriy Zaporozhets <dmitriy.zaporozhets@gmail.com>

change build status image to svg

Fix "useless assignment" Rubocop warnings