GitLab

jdk13 is early access (not even release candidate). We should stick to a released version, e.g. jdk12

Version 2.6.1 suffers from 5 high-severity and 1 moderate-severity
security vulnerabilities. Use the first version known to fix all of
these vulnerabilities.

Incidentally, all but one of these vulnerabilities date back to
2018 and earlier so one can't help but wonder why, seeing as this file
was only introduced mid-May 2019, we used such an old and buggy version
in the first place.

Feature/open jdk

pom.xml now only lists the direct dependencies (since maven takes care of
the indirect ones). This solves two issues that pom.xml had:
- different versions of slf4j required
- duplicated entries

For instance, it fixes this warning,
```
[WARNING]
[WARNING] Some problems were encountered while building the effective model for org.ensembl.hive:eHive:jar:1.0
[WARNING] 'dependencies.dependency.(groupId:artifactId:type:classifier)' must be unique: org.slf4j:slf4j-api:jar -> version 1.7.5 vs 1.7.12 @ line 65, column 17
[WARNING]
[WARNING] It is highly recommended to fix these problems because they threaten the stability of your build.
[WARNING]
[WARNING] For this reason, future Maven versions might no longer support building such malformed projects.
[WARNING]
```

I have also changed the versions to be the same ones as in the original
build.xml. The PR is about changing the build system, not the list /
versions of the dependencies.

* Bugfix:
  If user configured RunWorkerPath to be empty string,
  the runWorker script got prefixed with "/runWorker.pl"

Since we are comparing different databases (i.e.
travis_ehive_test_old_patched vs travis_ehive_test_new), we can't use
TABLE_CAT in the comparison

bugfix: without $self this is a function call, not a method call, and $self in synchronize_AnalysisStats is $stats instead of the Queen.

Because synchronize_AnalysisStats doesn't use $self and works when $stats
is missing, the call was basically doing nothing.

* version/2.4:
  Travis: comment on which encrypted Slack token is which
  [ENSCORESW-3077] Travis: Unify configuration of Slack notifications
  Travis: comment on which encrypted Slack token is which
  [ENSCORESW-3077] Travis: Unify configuration of Slack notifications

* version/2.2:
  Travis: comment on which encrypted Slack token is which
  [ENSCORESW-3077] Travis: Unify configuration of Slack notifications

...but keep notifying #ehive-commits.

...but keep notifying #ehive-commits.

sphinx.ext.pngmath is deprecated, sphinx.ext.imgmath is a drop-in
replacement for it (its default output format is PNG and under the
bonnet both use dvipng), and we could now easily switch to SVG
output by passing '-D imgmath_image_format=svg' to sphinx-build.

* version/2.4:
  Update README.md
  bugfix: add the missing print statements for tweaks that add control-flow rules

* version/2.3:
  Update README.md

* version/2.2:
  Update README.md

* version/2.4:
  bugfix: don't try to delete a parameter that doesn't exist
  bugfix: show the correct tweak query (i.e. with the meadow type)

The links previously used here were, in spite of having .txt at the end,
to respective GitHub HTML pages, which could come as a surprise to
someone who saved the links directly from the ReadTheDocs page and then
tried to feed the resulting files - with names still ending in .txt - to
pip. Using direct links both is less confusing and saves one click each,
regardless of whether one downloads the files or just passes links
directly to 'pip -r'.

Our documentation builder calls external scripts to produce Graphviz
dot code representing pipelines.

On the one hand, under both Python2 and Python3 the subprocess output
is a bytestring. On the other, some internals Sphinx Graphviz module
expects dot code to have string representation, which is NOT the same
as bytestrings under Python3. As a result, when running under Python3
the Graphviz module fails to render dot code due to being unable to
concatenate a string with a bytestring.

Now subprocess output is decoded to string, using the default encoding
of the given Python version. This makes the Graphviz module happy.

One way of doing this would be to refactor get_command() so that it
never outputs bytestrings, that said it feels like putting too much
effort into maintaining compatibility with a Python version that will
reach end of life within 2 years. Just choose the right string type
depending on the major version of Python, that way all that will have to
be done once we have decided to drop python2 support is to revert to
unconditionally checking _command_ against a single type.

Seems to work fine under both Python2 and Python3 now. Switching to
non-binary mode was needed due to Python3 strings being Unicode by
default.