Fixing missing icons.
Removing the breadcrumb from the box header, so it will only be on the context bar.

Addding context bar to more repo related pages.
Fixing the base template to work on non-repo related pages.

Adding the context bar too all pages related to a Repository.
Deleted top menu items that are on the context bar already.

Pass in old groups data to CanWriteToGroup validator for later skipping group checks.
This will be a part of refactoring done to do user permissions changes without messing with main
repo form data

Pass in old groups data to CanWriteToGroup validator for later skipping group checks.
This will be a part of refactoring done to do user permissions changes without messing with main
repo form data

configurable locking codes.
- this allows changing the default 423 LOCKED code to 2XX codes which don't break the transactions. Insted
it gives users a warning for push.

changelog: fix page state after clearing selection

It was a mess. Now it is less of a mess.

diff_block: fix link to file revisions

Pull requests would link to file revisions in the wrong repo. That was
obviously only visible when merging between different repos - but then it would
link to a non-existing revision.

diff_block is apparently used pull-request-style with the 'b' revision of the
diff shown first. It thus also has to point at the 'other' repo which is where
the other revision can be found.

pullrequest: just replace children of other_ref when other_repo changes

We were inserting a full select statement as innerHTML of a select and would
thus get nested select statements. IE did not render that correctly ... and
other browsers did strange things.

Instead we now create a in memory DOM object and move the optgroups. After
appending children to the selector we have to set selectedIndex on all
browsers.

vcs: fix typo in MercurialChangeset.committer

It has apparently never been used and could perhaps just be removed. Especially
because it also seems suspicious that it will do safe_unicode twice.

compare/pullrequest: introduce merge parameter

This is more correct and flexible than the old way of looking on same/different
repo.

pullrequest: pullrequest from changelog view

It seems like it didn't work ... perhaps because I broke it. But now we clean
it up and make it work:

Use rev_end as the revision to merge. We don't know where to merge ... but
start_rev cannot be used for that.

This might to some extent have been working before ... but we prefer to clean
it up and start over again when the normal merge workflow is fully working.

fixed issue with renaming repos group together with changing parents with multiple nested trees
added regresion tests for such cases

fixes #762, LDAP and container created users are now activated based on
the registration settings in permissions

moved out password reset tasks from celery, it doesn't make any sense to keep them there, additionally they are broken
in when executing _(), and url() calls. This fixes issue #572

changelog: "compare fork with parent" should only be shown when no changesets selected

It do not use the selected changeset - showing the option is confusing.

Greying out would perhaps be better ... but the button is already grey.

css: consistent spacing around :

sed -i 's/^\( [^: ]*\) *: */\1: /g' rhodecode/public/css/style.css

css: consistent space before {

sed -i 's,\([^ ]\){,\1 {,g' `find -name '*.css'`

fixed issue with renaming repos group together with changing parents with multiple nested trees
added regresion tests for such cases

better detection of deleting groups with subgroups inside.
Added less strict checks on delete group routing so we can delete zombie groups
(those that are not in filesystem but in DB)

compare: cleanup of as_form handling

as_form is only used when loaded 'partial' from pullrequest.html. Including it
in the 'Detailed compare view' url did thus not do any harm.

as_form is never used in pullrequest_show.html or compare_diff.html and it is
thus never used together with swap_url ... which is why the wrong
'as_form=None' or 'as_form=False' in urls didn't do any harm.

css: linear-gradient do not have 'top' but 'to bottom'

- according to https://developer.mozilla.org/en-US/docs/CSS/linear-gradient