Move 'Commits' and 'Releases' tabs down to body.
This patch also reduces page load time for pages that do not need
to use commits count anywhere. Get commits count can hurt
performance badly for huge repositories that has tens of thousands
commits like Linux Kernel.
Previously, we load all releases of a repository which could hurt
performance when the repository has a lot of releases.
Now we're able to only load releases in current page view we need
to show by matching with 'tag_name'.
Sometimes user could simply request wrong URL, but if that wrong URL
has nothing related Git HTTP operations, HTTP Basic Authentication
should not prompted. Instead, clean 404 page should be presented.
The patch also supports Git HTTP operations without '.git' suffix
to the repository name, which addresses #4226 and #4189.
Was only possible to correctly redirect to pull request page within
same repository. And didn't take care of case when upstream has
disabled pull request.
Also add a new method 'PullRequestURL' to unify the code.
1. Fix sample content didn't show up when hook doesn't exist.
2. Fix CSS 'word-break' to use 'break-word' not 'break-all'.
3. Fix single quote in 'locale_en-US.ini' file.
Usage: <url>?page={page}&pageSize={pageSize}
Also avoid/removed getting total commits count for pagination,
users are only allowed navigation by 'newer' and 'older'.
We used to handle SSH and HTTP push separately which produces
duplicated code, but now with post-receive hook, the process
is unified to one single place and much cleaner.
Thus, UpdateTask struct is removed.
Narrow down the range of Git HTTP routes to reduce condufsing
HTTP Basic Authentication window popup on browser.
By detecting <old-commit, new-commit, ref-name> inside post-receive
hook, Git HTTP doesn't need to read the whole content body anymore,
which completely solve the RAM problem reported in #636.
