devinabox -- Bootstrapping a core Python sprint
The devinabox project helps an experienced CPython developer produce a directory which contains everything necessary to enable sprint participants to get set up quickly (regardless of OS).
This README outlines two things:
- what to download to create a devinabox
- what is provided to help new contributors
This document provides instructions for sprint leaders; it does not provide instructions for new contributors. If you are a new contributor, ask your sprint leader(s) about how to get started.
Things to download to create a devinabox
The following sections outline various files to download and repositories to clone into your devinabox including:
- version control tools
Be careful NOT to change the destination directories that repositories are cloned into. These default directory names are assumed by the other files in devinabox.
When you are done you should have in the destination directory everything
someone needs to contribute. Simply copy the whole directory to some sort of
media (USB 3 drive and a CD tend to work well) and then pass it around for
people to copy somewhere on to their system. They can run
hg pull -u to
get updates, sparing the probably taxed internet connection at the sprint from
doing complete repository cloning.
If recreating from an old checkout,
hg purge --all in the individual
clones is a handy way to ensure old build artifacts have been removed.
You will need to enable the purge extension in
Also make sure to not simply copy your own repositories to the box! Otherwise the clones will most likely have paths which use SSH and the hg account on hg.python.org which only core developers can use. It's easier to make the clones from scratch.
You will want to download the latest release of Mercurial (http://pypi.python.org/pypi/Mercurial) and TortoiseHg for Windows users (http://tortoisehg.bitbucket.org/download/). OS X users can be told that Mercurial is available through Homebrew if they prefer (if they use MacPorts or any other package manager ask them to use the download of Mercurial you have provided to save time).
Providing Mercurial guarantees there is no issue with new contributors trying to update repositories or generating patches.
If you receive questions about compilers, here are some suggestions.
OS X users should be told to download XCode from the Apple App Store ahead of
time. It's on the order of a couple GiB in size, so you don't want to have
people downloading it at the sprint. After installation they should also make
sure to install the command-line tools (e.g. in Mavericks,
If new contributors think they may be doing C development, suggest the use of LLVM + clang as this provides better error reporting than gcc.
For Windows users, ask them to download and install Visual Studio Community edition ahead of time.
Clone the CPython repository and build it (you will be cleaning up your build later, though as a final step).
Also make sure to build the documentation. This alleviates the need for sprint participants to build it from scratch. To build the documentation, create a venv with sphinx installed and point the Doc Makefile at the Python linked to in the venv.
All of this can be done by doing:
# Assuming at the root of the devinabox directory python build_cpython.py ./cpython/python -m venv venv ./venv/bin/pip install sphinx cd cpython/Doc make html PYTHON=../../venv/bin/python
Clone the PEP repository and build it (use the venv you created to build the CPython docs if necessary). This allows sprinters a local copy to reference for a PEP and it allows using the easier-to-read HTML version.
No specific guidelines for building the PEPs are provided since there is only a slim chance sprint participants will be editing a PEP.
Clone the devguide repository and build it (again, use the venv created to build the CPython docs if necessary). This gives sprinters a local copy to use rather than having to use the (often slow) internet connection at the sprint.
- Download coverage (need a special file that is not part of the normal distribution of coverage, so can't just use pip)
- Build CPython:
- Create an venv:
./cpython/python -m venv venv
- Extract coverage:
tar -x -f coverage-*.tar.gz
- Install coverage in the venv:
./venv/bin/python coverage-*/setup.py install
- Set PYTHONPATH to
fullcoverage(need to change your directory to the venv):
unset CPPFLAGSin order to avoid using system Python header files
- Run coverage from the venv:
./bin/python -m coverage run --pylib -m test
- Unset PYTHONPATH:
- Generate coverage report:
./bin/python -m coverage html --directory=../coverage_report -i --include="../cpython/Lib/*" --title="CPython test coverage report"
Do be aware that this step takes a few hours. If you find report generation is the bottleneck you can try using PyPy3 or your installed Python 3 interpreter to generate the report.
Helpful files for sprint participants
Helpful files are included in order to make things a little bit easier for you, the sprint leader, as well as sprint participants and new contributors.
An HTML file with links to:
- documentation which you built previously
- the helper scripts
On UNIX-based OSs this file builds the CPython repository. On all platforms it verifies that the expected CPython binary exists.
While the devguide includes instructions on how to build under UNIX, this script simplifies the process for sprint participants by having a single command to configure and build CPython. It also uses reasonable defaults (e.g. all cores on the CPU).