[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Upstream CI: send patches to use Debian packages when existing?



Hello,

I am not really sure about how this would be perceived by upstream since
they do not only want to ensure that their software works on
Debian/Ubuntu .. but that their users can repeat the installation on
their platform, too. There may hence be a point to install software from
pip or via conda, both being mostly cross-platform, instead of using the
same package from the Linux distribution.

But then again ... let's discuss a few:

https://github.com/nanoporetech/pychopper/blob/master/.gitlab-ci.yml

before_script:
    - apt-get update
    - apt-get remove python
    - apt-get install -y python3 python3-pip make python3-numpy
python3-matplotlib hmmer
    - update-alternatives --install /usr/bin/python python
/usr/bin/python3 10
    - alias python=python3; pip3 install --upgrade pip
    - hash -r pip3
    - alias python=python3; pip3 install sphinx sphinx-argparse
sphinx_rtd_theme pytest pandas
    - alias python=python3; pip3 install -e ./

So, if the above installs numpy from "us", it can also install sphinx,
pytest and pandas from us. Right?

https://github.com/nanoporetech/medaka/blob/master/.gitlab-ci.yml
is completely debianised already.

https://github.com/nanoporetech/flappie/blob/master/.travis.yml
debianised but
https://github.com/nanoporetech/flappie/blob/master/.gitlab-ci.yml
cheats with their own scrappie-docker image

https://github.com/nanoporetech/qcat/blob/master/Dockerfile
starts out with Ubuntu 16:04 and updates a wheels and setuptools via pip.

https://github.com/arq5x/poretools/blob/master/Dockerfile
was last edited 2014. Ouch. Ubuntu trusty this is, right? From pip it
fetches what today is python3-numexpr and the R packages we have, too.

https://github.com/wdecoster/nanomath/blob/master/.travis.yml
nothing bound to any distribution - we have flake8 which is installed
via pip

https://github.com/python-streamz/streamz/blob/master/.travis.yml
installs conda (and the requirements.txt that Debian also has with it)
but gets python3-distributed from git
Don't see anything particularly Debianish here.

I sense that there are two major categories of testing/distribution
environments:
 A) those that give all they can to be platform independent - like only
using Python associated tools
 B) those that start out with a Debian environment (typically Ubuntu)
and then mix downstream from there

We should not bother A) but the B)-types - yes, they should get pull
requests from us, especially so when
their package is distributed with Debian already.

Opinions?

Steffen


Reply to: