* Move files for better setup.py packaging
* Update setup.py and configs for src layout
* Run tests on setup.py build
In effect, this tests that the setup.py configuration is correct.
coverage combine and coverage:paths are added to keep the displayed
coverage paths as src/ssh_audit/*.py instead of
.tox/$envname/**/site-packages/ssh_audit/*.py
* Remove unnecessary encoding declarations
Python 3 defaults to UTF-8 encoding.
https://docs.python.org/3/reference/lexical_analysis.html#encoding-declarations
* Remove shebang from colorama type stubs
Shouldn't need to be an executable.
Related: git has this file tracked as chmod -x.
* Ignore all flake8 warnings - one by one
Without ignoring, there are by far more than 1000 linting issues.
Fixing these warnings means possibly changing almost every line of
code, as single warnings can effect more than one line.
Doing this in one pull request is generally no good idea, and especially
not now, as the test suite is currently broken.
Instead of just deactivating flake8, or ignoring its exit code, the
warnings are ignored one by one.
This means, when one wants to work on the linting issues, one can just
remove one ignored warning, and fix the problems - which is not too much
work at once, and leads to an managable diff.
modified: tox.ini
* Unpin dependencies for mypy run
... as they could not be installed due to compilation errors.
modified: tox.ini
* Fix syntax error for mypy
When new code was added via
af663da838
the type hint was moved further down and so caused a syntax error, as
type hints have to follow the function declaration directly.
Now, the the type linter finally works and shows 187 errors.
modified: ssh-audit.py
* Update .gitignore for mypy
modified: .gitignore
* Let tox not fail on mypy errors
Currently, there are almost 200 typing related errors.
Instead of letting the tox run fail, the errors are still shown, but
the exit code gets ignored for now.
This way one can fix them one by one - if wanted.
modified: tox.ini
* Let tox not fail on pylint errors
Currently, there are more than 100 linting related errors.
Most of them will be fixed when flake8 gets fixed.
Instead of letting the tox run fail, the errors are still shown, but the
exit code gets ignored for now.
This way, one can fix them one by one.
modified: tox.ini
* Let vulture only fail on 100% confidence
Vulture is a tool to find dead code. Unlike Flake8, which also finds
unused imports and variables, Vulture does some guess work and finally
outputs a list of possible dead code with a confidence marker.
Already the first result ...
"ssh-audit.py:48: unused import 'Dict' (90% confidence)"
... is a false-positive.
As Flake8 also does a good job in detecting unused code, it makes not
much sense to let tox fail when vulture fails.
Instead of deactivating vulture, it was configured in a way to only
report results with 100% confidence.
modified: tox.ini
* Make timeout_set optional
When timeout_set was introduced in
1ec13c653e
the tests were not updated, which instantiated the Socket class.
While the commit message read "A timeout can now be specified", the
code enforced a `timeout_set`.
`timeout_set` now is `False` by default.
modified: ssh-audit.py
* Set default values for Socket's `ipvo` and `timeout`
Commit
f44663bfc4
introduced two new arguments to the Socket class, but did not update
the tests, which still relied on the socket class to only require two arguments.
While for `ipvo`the default of `None` is obvious, as in `__init__` it is
checked for it, for `timeout` it was not that obvious.
Luckily, in the README a default of 5 (seconds) is mentioned.
modified: ssh-audit.py
* Un-comment exception handling
While working on commit
fd3a1f7d41
possibly it was forgotten to undo the commenting of the exception
handling for the case, when the Socket class was instantiated with a
missing `host` argument.
This broke the `test_invalid_host` test.
modified: ssh-audit.py
* Skip `test_ssh2_server_simple` temporarily
After fixing all the other tests and make tox run again, there is one
failing test left, which unfortunately is not super easy to fix without
further research (at least not for me).
I marked `test_ssh2_server_simple` to be skipped in test runs
(temporarily), so at least, when working on new features, there is
working test suite, now.
modified: test/test_ssh2.py
* Do not pin pytest and coverage version
... but do use pytest < 6, as this version will have a breaking change
with junit/Jenkins integration
Also see https://github.com/jtesta/ssh-audit/issues/34
* Drop unsupported Python versions
... except Python 2.7, as this will need also changes to the source
code, and this pull request is already big enough.
Also, support for Python 3.8 was added.
The Travis configuration was simplified a lot, by leveraging the tox
configuration.
Also, the mac builds have been dropped, as they all took almost an hour
each, they failed and I have no experience on how to fix them.
The `appveyor` build only has been updated to reflect the updated Python
versions, as I have no access to the status page and no experience with
this build environment.
Also, removed call to `coveralls`, which seems to be a leftover from
the old repository.
modified: .appveyor.yml
modified: .travis.yml
modified: packages/setup.py
deleted: test/tools/ci-linux.sh
modified: tox.ini
