index : cgit

cgit - CGI for Git
==================

This is an attempt to create a fast web interface for the Git SCM, using a
built-in cache to decrease server I/O pressure.


Installation
------------

Building cgit involves building a proper version of Git. How to do this depends
on how you obtained the cgit sources:

a) If you're working in a cloned cgit repository, you first need to initialize
and update the Git submodule:

    $ git submodule init     # register the Git submodule in .git/config
    $ $EDITOR .git/config    # if you want to specify a different url for git
    $ git submodule update   # clone/fetch and checkout correct git version

b) If you're building from a cgit tarball, you can download a proper git version
like this:

    $ make get-git

When either a) or b) has been performed, you can build and install cgit like
this:

    $ make
    $ sudo make install

This will install `cgit.cgi` and `cgit.css` into `/var/www/htdocs/cgit`. You can
configure this location (and a few other things) by providing a `cgit.conf` file
(see the Makefile for details).

Lua is optional and only powers the lua: filter extensions (authentication,
email and commit-message filters in custom/extensions/). A plain build
auto-detects a Lua through pkg-config, preferring LuaJIT, and falls back to a
Lua-less binary when none is found. Acceptable values are generally "luajit",
"lua", "lua5.4", "lua5.3", "lua5.2" and "lua5.1".

To pin an implementation:

    $ make LUA_PKGCONFIG=lua5.4

To build without Lua, so the binary needs no Lua at runtime:

    $ make NO_LUA=1


Previewing locally
------------------

cgit is a CGI program, so trying it out normally means configuring a web server.
For development there is a small dependency-free preview server at
`tools/serve.py` that runs the built binary and serves the static assets. It
needs only Python 3.

    $ python3 tools/serve.py --config /path/to/cgitrc --port 8080

It is a development aid and is not meant to face the internet.


Tests
-----

The suite in tests/ is built on the test library in the bundled Git tree, the
same harness Git uses for its own suite, one self-contained tNNNN-*.sh script
per area. Run it from the top level, which builds cgit and the bundled Git
tree before anything runs.

    $ make test

Once those are built, a single script runs directly from tests/, where -v
shows each check and -i stops at the first failure. The full option list
lives in vendor/git/t/README, and options for a run through make go in
CGIT_TEST_OPTS, where --valgrind runs every cgit invocation under valgrind.

    $ cd tests && ./t0104-tree.sh -v

The numbering walks outward, t000x for the ground the suite stands on, t01xx
for page content, t02xx for features that cut across pages, t03xx for
security regressions, t04xx for the helper tools under tools/ and t05xx for
the extensions under custom/extensions, each pairing unit checks under a
standalone Lua with a run through cgit itself. A failing script leaves its
trash directory behind under tests/trash/ with the pages it was looking at, and
scripts skip rather than fail when a helper program or Lua module is missing,
saying so in their output.

A new script takes the next free number, needs the executable bit and is
picked up with no further wiring. Three traps are worth knowing. The cgitrc
that tests/setup.sh writes enables the cache, so a test rendering one URL
under two configs needs a config of its own with cache-size=0. A bare plus in
a cgit_url argument decodes to a space, so the foo+bar fixture repository is
written foo%2bbar. And a log page echoes the search query back inside an
input value, so an assertion that a commit is absent should match the subject
link, written as ">commit 2</a>", rather than the bare text.


Dependencies
------------

* zlib
* optional: luajit or lua, most reliably used when pkg-config is available

cgit builds without OpenSSL or libcurl, so no development packages for those are
needed.


Filter extensions
-----------------

The optional Lua filters in `custom/extensions/` need extra Lua modules. Each
script's header lists the exact install commands for its own dependencies.

* The auth filters (`auth-file.lua`, `auth-inline.lua`) need `luaossl` and
  `luaposix`.
* The email filters (`email-gravatar.lua`, `email-libravatar.lua`) need
  `luaossl`.
* The syntax highlighter (`syntax-highlight.lua`) needs `lpeg` and a Scintillua
  lexer set.
* The about-page renderer (`about-render.lua`) needs `lpeg` for markdown and
  man pages. Plain text needs only Lua.

These filters target Lua 5.1 through 5.4 and LuaJIT. `luaossl` has no Lua 5.5
build, so build cgit against 5.1 to 5.4 if you use the auth or email filters.
`link-commits.lua` needs nothing beyond Lua itself.


Web server configuration
------------------------

cgit is a CGI program. Complete, commented configurations for nginx, Apache and
lighttpd live in `custom/servers/`, each explaining how that server routes
requests to the binary and serves the static assets off disk.


URLs and query parameters
-------------------------

cgit serves every page from one CGI program, so a url names the repository, the
page and a path inside that page. There are two forms and they carry the same
information. With virtual-root set, a request is a path, written as
/<repo>/<page>/<path> with anything else as a query string. Without it, the
whole request lives in the query string as ?url=<repo>/<page>/<path>, which is
the form the test suite uses. In the path form the first extra argument opens
the query string with a question mark, and in the url= form it continues the
existing one with an ampersand, which is why links in the two forms are spelled
differently.

The page name is the second element and is one of about, atom, blame, blob,
commit, diff, log, patch, plain, rawdiff, refs, snapshot, stats, summary, tag or
tree. Leaving it out gives the repository summary, and leaving the repository
out as well gives the index of repositories. Everything after the page name is
the path argument, so /demo/tree/src/main.c asks for the tree page at
src/main.c. A repository whose name contains a character that is special in a
url has it percent-encoded, and a plus sign has to be written %2b because a bare
plus decodes to a space.

Every parameter is optional. The pages column names the pages that read each
one, where a value of all marks the few not tied to particular pages.

+----------+----------------------+---------------------------------------------+
| param    | pages                | description                                 |
+----------+----------------------+---------------------------------------------+
| url      | all                  | repository, page and path in one value      |
| h        | all                  | the branch or ref to read                   |
| id       | all                  | pin the page to one commit or object        |
| id2      | diff patch rawdiff   | the second object of a diff, paired with id |
| ofs      | index log refs stats | offset into a paged listing                 |
| path     | all                  | restrict the page to one path               |
| q        | log                  | the search term                             |
| qt       | log                  | one of grep, author, committer or range     |
| s        | index                | sort key for the listing                    |
| showmsg  | log                  | show full commit messages                   |
| period   | stats                | the statistics window, one of w, m, q or y  |
| dt       | diff commit          | diff type, unified, side by side or raw     |
| ss       | diff commit          | shorthand for the side by side diff         |
| all      | atom                 | include every ref rather than one branch    |
| context  | diff commit          | lines of context around a diff hunk         |
| ignorews | diff commit          | ignore whitespace when diffing              |
| follow   | log                  | follow a single path across renames         |
+----------+----------------------+---------------------------------------------+

A few endpoints are not ordinary pages. The snapshot page takes a filename
rather than a ref, so /demo/snapshot/demo-1.0.tar.gz names both the ref and the
archive format through the suffix, and the formats on offer are set by the
snapshots option. The plain page serves a blob as its own bytes under headers
that stop a browser treating repository content as markup. The atom page is a
feed rather than a page and accepts h, path and all. The clone endpoints under
info and objects implement the dumb HTTP protocol and are only present when http
clone is enabled.

A url fragment can address lines inside a file. The tree and blame pages give
every line an anchor named n followed by its number, so #n17 jumps to line 17
and highlights it, and a range written #n17-n24 highlights the whole run and
scrolls it into view.

Some worked examples, in the url= form.

    ?url=                              the repository index
    ?url=demo                          summary for the demo repository
    ?url=demo/log&h=next               log of the next branch
    ?url=demo/log&qt=author&q=alice    commits authored by alice
    ?url=demo/tree/src&h=v1.0          the src directory at tag v1.0
    ?url=demo/commit&id=HEAD~3         one commit, pinned
    ?url=demo/diff&id=main&id2=next    diff between two branches
    ?url=demo/plain/README.md          the raw bytes of one file
    ?url=demo/atom&h=main              the commit feed for a branch


Runtime configuration
---------------------

The file `/etc/cgitrc` is read by cgit before handling a request. In addition to
runtime parameters, this file may also contain a list of repositories displayed
by cgit (see `MANUAL.txt` for further details). A fully commented starting point
with every option at its default is in `custom/cgitrc`.


The cache
---------

When cgit is invoked it looks for a cache file matching the request and returns
it to the client. If no such cache file exists (or if it has expired), the
content for the request is written into the proper cache file before the file is
returned.

If the cache file has expired but cgit is unable to obtain a lock for it, the
stale cache file is returned to the client. This is done to favour page
throughput over page freshness.

The generated content contains the complete response to the client, including
the HTTP headers.