Warning: the reported coverage is much lower than before. But when reading
the previous reports, it seems the coverage was overestimated by large,
mostly because old report did not consider all files in woob dir.

Also for some reason pytest's JUnit XML uses `modules.<name>.<class>`
with the `modules.` prefix so we take it into account in testing grid.

Sometimes, it's desirable to run tests even if some modules can't be
loaded.

For user setups, a number of dirs are used, whose paths are
customizable.

For workdir, before:
- $WEBOOB_WORKDIR
- $XDG_CONFIG_HOME/weboob

and after:
- $WOOB_WORKDIR
- $WEBOOB_WORKDIR
- $XDG_CONFIG_HOME/woob
- $XDG_CONFIG_HOME/weboob

If the old name is found in $XDG_CONFIG_HOME, it will try to rename it.

For data dir, now:
- $WOOB_DATADIR
- $WOOB_WORKDIR
- $WEBOOB_DATADIR
- $WEBOOB_WORKDIR
- $XDG_DATA_HOME/woob
- $XDG_DATA_HOME/weboob

Also, $WOOB_BACKENDS then $WEBOOB_BACKENDS.

By forcing argv, "config update" was received as a command.

For now, run using weboob package from PYTHONPATH. Maybe we should
require setup.py to have been run before being able to start run_tests.
This way, we're sure everything is set up.

Same for make_man and weboob_lint.

Laurent Bachelier authored Oct 15, 2019 and hydrargyrum committed Nov 30, 2019

Under Python 3, __pycache__ is used, and is not cleaned up.
It is also not necessary to clean it up anymore, since those .pyc files
cannot be imported without a .py files unlike what happens under Python 2.

Sébastien Marie authored Apr 25, 2018 and hydrargyrum committed Apr 26, 2018

- use env(1) to found bash on the system
- remove bash dependency on few scripts

This is starting to get quite repetitive though.

This will generate a lot of warnings on modules since they are not
python3-compatible yet.
Use the same flag for run_tests, and use python3 by default for weboob_lint.

This commit adds the necessary files to run the CI using Gitlab-CI. For
now, it checks that Weboob builds, then runs the linting script
(checking that every module as an icon and some tests + PyFlakes) and
the unittests. Most of modules unittests cannot run because there is no
backend configured.

Some changes were needed in the pre-existing scripts:
* Edit `weboob_lint` to exit with non-zero code if it finds modules
without icons or tests, so that the build could fail in such a case.
* Edit `run_tests.sh` to set correct exit code on failure and rework
generation of XUNIT output. Also added some doc about useful environment
variables. Added a way to generate an xunit output file when running modules
unittests, passing a `XUNIT_OUT` env variable to `run_tests.sh` script.
* Modification of `setup.cfg` and `run_tests` scripts to handle code
coverage generation. The matching regex in Gitlab for the total code
coverage is `TOTAL: (\d+\%\s*)$)`.

I also added a script to generate a JSON module status matrix from
modules unittests, ready to be sent to a
[Weboob-CI](https://github.com/Phyks/weboob-ci) instance.

NOTE: Required Python modules are taken from the `setup.py` script.
`.ci/requirements.txt` contains the requirements to run the unittests
and the CI, whereas `.ci/requirements_modules.txt` contains the specific
Python modules required at runtime by Weboob modules. The latter could
eventually be replaced by a proper call to `debpydep` script.

Another command was run after executing nose, so the real exit code was
lost.

related to e93d724e

This way run_tests.sh follows what local_run.sh can do.

close #1403

Because of BSD.

It would cause issues if we had a different filename for WEBOOB_BACKENDS.

Laurent Bachelier authored Apr 09, 2013 and Romain Bignon committed Apr 10, 2013

This should make running tests faster, as we do not remove all .pyc
files anymore!

when run by buildslaves

Laurent Bachelier authored Nov 10, 2012 and Romain Bignon committed Nov 10, 2012

Some were not caught, and they were ran even for a single-module test.
Also, don't include them in the report.

Laurent Bachelier authored Nov 10, 2012 and Romain Bignon committed Nov 10, 2012

There is no processing of the reports for now,
but the submit part is done.

Get the explicit Python 2 version if there is one.

fixes #842