README for Lorry ================ Lorry is a tool to take upstream source code (in various formats, though preferably in version control) and converts it into a git repository. If you want to try this, use `--pull-only` and/or `--mirror-base-url-push` so that you do not accidentally overwrite important stuff for Baserock. (If you don't have direct commit access to Baserock on git.baserock.org then you're not dangerous.) See the manual page for instructions on using. The usual use is that a server hosts lorry and checks for updates, which it then lorries. Lorry should not generally be run from the developer's machine apart from testing. This is because Lorry has to keep the git trees to avoid it having to pull everything. You can find a lot of lorries to crib ideas from at . Dependencies ------------ Required: * **Python 3**: Minimum of 3.6. * **Git** * **cliapp**: Can be installed as the `python3-cliapp` package in Debian, or with: pip3 install https://gitlab.com/trovekube/cliapp/-/archive/cliapp-1.20180812.1/cliapp-cliapp-1.20180812.1.tar.gz or from the source at . * **PyYAML**: Can be installed from PyPI. or as the `python3-yaml` package in Debian. Optional: * **bzr-fastimport**: Needed if you want to import Bazaar (bzr) repositories. Can be installed as the `bzr-fastimport` package in Debian. It is also possible to use Breezy instead, but this should currently be considered experimental due to bugs in its implementation of `fast-export`. * **cmdtest**: Needed if you want to run the test suite. Can be installed as the `cmdtest` package in Debian, or from the source at . * **git-cvsimport**: Needed if you want to import CVS repositories. Can be installed as the `git-cvs` package in Debian. * **git-svn**: Needed if you want to import Subversion (svn) repositories. Can be installed as the `git-svn` package in Debian. * **hg-fast-export**: Needed if you want to import Mercurial (hg) repositories. Can be installed from the source at . * **OpenSSH**: The OpenSSH client is needed if you want to import or push to repositories using the SSH protocol. * **Perl**: Needed if you want to import tarballs. * **Git LFS**: Needed if you want to import raw files. Lorry file specification ------------------------ Lorry files are json dicts where the repository names are the keys and the values are dicts with the data required to mirror it. So a simple lorry that mirrors a git project looks like { "git": { "type": "git", "url": "git://github.com/gitster/git.git" } } Multiple repositories can be specified in the same .lorry file, in which case all of them will be processed by lorry. The following shows two repositories. { "git": { "type": "git", "url": "git://github.com/gitster/git.git" }, "curl": { "type": "git", "url": "git://github.com/bagder/curl.git" } } Lorry can import other version control systems into git. When the URL uses the `https:` scheme, Lorry will validate the SSL/TLS server certificate by default. If necessary, this can be disabled for a Bazaar, Git, and Mercurial server by adding the key: "check-certificates": false ### Mercurial Mercurial is very similar to git, just change the type field to "hg" { "sudo": { "type": "hg", "url": "http://www.sudo.ws/repos/sudo" } } ### Bazaar Repositories and branches in Bazaar mean different things to Git. The practical difference for Lorry is that it is not possible to have a url for a repository, urls map directly to branches. { "libpipeline": { "type": "bzr", "branches": { "trunk": "http://bzr.savannah.gnu.org/r/libpipeline/trunk" } } } For convenience if the project only needs one branch mirrored, the url is assumed to be the master branch. { "libpipeline": { "type": "bzr", "url": "http://bzr.savannah.gnu.org/r/libpipeline/trunk" } } ### Subversion To support all the branches and tags a layout needs to be specified as svn is very flexible with the possible layouts, however the most common is to have the working branch in a directory called trunk, and the branches and tags in respectively named subdirectories. Because this is so common "standard" can be used as the layout { "mpc": { "type": "svn", "url": "svn://scm.gforge.inria.fr/svn/mpc", "layout": "standard" } } This is equivalent to { "mpc": { "type": "svn", "url": "svn://scm.gforge.inria.fr/svn/mpc", "layout": { "trunk": "trunk", "branches": "branches/*", "tags": "tags/*" } } } Trunk is the path to the directory where the main branch is located. Branches and Tags are glob expressions to allow finer control over which paths are used. Trunk is mandatory, but Branches and Tags are optional. Texlive keeps a lot of resources in their svn repository, we are only concerned with the source code, so this layout should select the correct subdirectory for each branch. { "texlive": { "type": "svn", "url": "svn://tug.org/texlive/", "layout": { "trunk": "trunk/Build/source", "branches": "branches/*/Build/source", "tags": "tags/*/Build/source" } } } Brace expansions can be used to specify subsets of paths. Netpbm for example, keeps all its branches in the root directory { "netpbm": { "type": "svn", "url": "https://netpbm.svn.sourceforge.net/svnroot/netpbm", "layout": { "trunk": "trunk", "branches": "{advanced,stable,super_stable}", "tags": "release_number/*" } } } Note that git-svn can provide better history tracking if the url is as close to the root of the repository as possible, so it may be more effective if the lorry was specified similar to this, assuming svnroot is the real root of the repo { "netpbm": { "type": "svn", "url": "https://netpbm.svn.sourceforge.net/svnroot/", "layout": { "trunk": "netpbm/trunk", "branches": "netpbm/{advanced,stable,super_stable}", "tags": "netpbm/release_number/*" } } } ### CVS The url for CVS repositories is the CVSROOT string. The module is required as cvs repositories usually have multiple modules, the module is usually the same as the project name. { "openssl": { "type": "cvs", "url": "anonymous@cvs.openssl.org:/openssl-cvs", "module": "openssl" } } ### Tarball Lorry can import a tarball fetched from a URL. The contents will be committed on a branch named after the basename of the tar file (e.g. bc-1.06.tar.gz will be imported into a branch named 'bc-1.06'. The import is done by the `lorry.tar-importer` subprocess. It can detect and handle common compression formats including gzip, bz2, xz and lzma. It will also detect if there is a 'top directory' that contains the tarball contents and strip this out of the imported filenames. Tarball imports once required 'compression' and 'strip' to be specified. These are obsolete now and are ignored by Lorry. { "bc": { "type": "tarball", "url": "http://ftp.gnu.org/gnu/bc/bc-1.06.tar.gz" } } ### Zip Lorry can import a zip file fetched from a URL. The contents will be extracted in place and committed directly to master. The new commit will be tagged with the basename of the imported zip file (e.g. docbook-xml-4.5.zip will be tagged as 'docbook-xml-4.5') { "docbook-xml": { "type": "zip", "url": "http://www.docbook.org/xml/4.5/docbook-xml-4.5.zip" } } ### Gzip Lorry can import a gzip file fetched from a URL. The file will be extracted in place and committed directly to master after removing the .gz extension from the filename. The new commit will be tagged with the basename of the imported zip file (e.g. bkai00mp.ttf.gz will be tagged as 'bkai00mp.ttf') { "ttf-bkai00mp": { "type": "gzip", "url": "ftp://ftp.gnu.org/non-gnu/chinese-fonts-truetype/bkai00mp.ttf.gz" } } ### Raw File Lorry can store raw files in a git LFS repository, which may allow your git server to serve those files via its repository browser. For convenience, raw file lorries can specify multiple sources to store in the same repository. By default each raw file will be stored at the root of the repository, unless otherwise specified by the optional `destination` field, which specifies the subdirectory under which the raw file will be stored. { "raw-file-repo": { "type": "raw-file", "urls": [ { "url": "http://geoportal1-ons.opendata.arcgis.com/datasets/0b09996863af4b5db78058225bac5d1b_0.kml", "destination": "authorities/england-and-wales" }, { "url": "https://fsadata.github.io/radiological-monitoring-data/data/provisional-2016-bi-annual-aquatic-and-terrestrial-monitoring-results-26062017-raw-results.csv", "destination": "radiological-monitoring-data" } ] } } Tips ---- 1. Use upstream's git repository whenever possible Importing from foreign version control systems is always slower than a git mirror (with the exception of tarballs because they have less history). 2. GNU Projects often have a git repository Most GNU projects are old compared to git, so were mainly developed in CVS. Many official websites only mention the CVS or SVN repositories. They will tend to have a git repository as well though, especially if they are hosted on savannah. Legal stuff ----------- See the LICENSE file.