-
Notifications
You must be signed in to change notification settings - Fork 670
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Travis run the test in minimal and full config #570
Conversation
- [email protected] | ||
- MDA_DOCDIR=package/doc/html | ||
matrix: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I prefer if this was at the top of the file where we define the python versions we use.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
On 08/12/15 00:00, kain88-de wrote:
In .travis.yml
#570 (comment):\ No newline at end of file
- global:
- secure="HIj3p+p2PV8DBVg/KGUx6n83KwB0ASE5FwOn0SMB9zxnzAqe8sapwdBQdMdq0s XB7xT1spJqRxuxOMVEVn35BNLu7bxMLfa4287C8YXcomnvmv9xruxAsjsIewnNQ80vtPVbQddBPxa4 jKbqgPby5QhhAP8KANAqYe44pIV70fY="
- GH_DOC_BRANCH=develop
- GH_REPOSITORY=github.com/MDAnalysis/mdanalysis.git
- GIT_CI_USER=TravisCI
- GIT_CI_EMAIL=[email protected]
- MDA_DOCDIR=package/doc/html
- matrix:
I prefer if this was at the top of the file where we define the python
versions we use.—
Reply to this email directly or view it on GitHub
https://github.com/MDAnalysis/mdanalysis/pull/570/files#r46891970.I prefer it at the top too. But it was at the end so I left it there. I
can move it up.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
By all means, move it up. I might have added it and I just followed other examples without much thinking.
This is awesome, thanks. It looks like there's just a few things that need decorating with knownfail |
- source activate pyenv | ||
- conda install --yes python=$TRAVIS_PYTHON_VERSION cython biopython matplotlib networkx netcdf4 | ||
- if [[ $SETUP == 'full' ]]; then conda install --yes python=$TRAVIS_PYTHON_VERSION cython biopython matplotlib networkx netcdf4; fi | ||
- if [[ $SETUP == 'minimal' ]]; then conda install --yes python=$TRAVIS_PYTHON_VERSION cython biopython networkx; fi |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
minimal should work without cython.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Hm depending on when #560 get's merged maybe we can also say that a 'full' setup means we cythonize
the files again and in the minimal setup we use the shipped *.c
files.
This commit asks Travis to run the tests twice: once with all the requirements installed (SETUP=='full'), and once with none of the optional dependencies installed (SETUP=='minimal'). Note that the minimal setup, runs without the MKL as the conda mkl package pulls scipy with it.
Tests often fail in ways that seem related to timeout issue. Indeed, Travis can be slow sometime. This commit increases the timeout for tests from 120 seconds to 300 seconds.
Build the full setup with and without cythonize; build the minimal setup only without cythonize.
This makes the file easier to understand as a reader see where the variable come from before they are used.
I rebased against develop to account for #560. I also adapted the matrix to follow @richardjgowers suggestions in #560; 3 versions are built:
I also removed cython from the packages installed in the minimal setup, and I moved the environment variable at the top of the file to improve readability. |
I don't think we need to run 3 versions. All the cython-compiled C-files will also be tested in the minimal install since none of them is optional (and they shouldn't be since then people who have installed cython but not the optional dependencies might have problems). And also consider that we will have double the number of build-jobs as soon as we support python 3 as well. With this in mind and travis asking nicely that we keep the build-matrix minimal I think we can skip full install without cythonize.
|
I don't mind removing the full install without cythonize. It is just a matter of removing one line. What are the cases where the build with cythonize can pass while the build whithout fails? Or vice-versa? |
It can happen that someone doesn't update the cython generated C-files. If a bug was fixed in the pyx files that can lead to the tests failing for that bug. |
I don't mind either way, so I'll let @kain88-de decide |
Remove the build with full install and no cythonize as suggested by @kain88-de in MDAnalysis#570.
If @kain88-de decides, then I remove the useless build 😛 |
Travis run the test in minimal and full config
This commit asks Travis to run the tests twice: once with all the requirements installed (SETUP=='full'), and once with none of the optional dependencies installed (SETUP=='minimal'). The minimal setup, has to run without the MKL as the conda mkl package pulls scipy with it.
The build fails which is expected. Indeed, until #568 get merged, some tests fail if scipy or netcdf are not installed.
Fix #483