[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Improving the Python build system.
From: |
Andreas Enge |
Subject: |
Re: Improving the Python build system. |
Date: |
Sun, 1 Sep 2013 15:27:26 +0200 |
User-agent: |
Mutt/1.5.21 (2010-09-15) |
On Mon, Aug 26, 2013 at 01:29:48AM +0200, Cyril Roelandt wrote:
> Andreas, could you please send us your Python modules that fail
> because of setuptools ? I think these packages require setuptools
> instead of distutils (which comes with Python), so we'll probably
> have to package setuptools. Both are quite similar, so the
> "python-build-system" can probably be used anyway. Otherwise, we'll
> have to write a build system per tool
> (distutils/distutils2/setuptools/bento).
Probably (once the python changes dust settles), it will be enough to simply
add setuptools as an input to the packages.
I would like to suggest a few modifications to the python build system
resulting from discussions with Brandon Invergo at the GHM. If I understood
correctly, then "setup.py check" only makes some basic checks on the package
and can be safely dropped. This is corroborated by
python setup.py --help-commands | grep check
on pytz, for instance, which displays
check perform some checks on the package
However, "python setup.py --help-commands | grep test":
test run unit tests after in-place build
I think this it what we would like to do in the check phase.
Then it is possible to also separate the build and install phases:
build build everything needed to install
I think we would like to add such a phase.
Stylistically, they should probably be obtained by a call to one function
returning a procedure calling setup.py with "build", "test" and "install",
respectively, instead of copy-pasting three times the same code.
There is another mild complication: The binary should be called "python3"
instead of "python" under Python 3.
What do you think?
Andreas
- Re: Improving the Python build system.,
Andreas Enge <=