GitLab

* java: (37 commits)
  Moved the guest-language tests to their own file
  [java] Given the Worker life-cycle we can assume that the wrapper has been compiled recently
  [java] Hidden option to check for type conversions
  [java] Note for later
  [java] More logical structure
  [java] The list was missing Long, so better to use the super-class
  [java] More checks of the debug level + minimum is WARN
  debug for travis
  [Java] Added "version" task in ant
  [Java/LongMult] Make a unique list of jobs because PostgreSQL doesn't have INSERT IGNORE
  Slower loop to allow the Java workers to spawn
  For this to work, Java obviously has to be installed
  typo
  Changed the default debug levels (higher level = more verbose)
  bugfix: was using the wrong argument
  bugfix: exec() doesn't like undefs
  Copyright update
  Moved the Java config file to Examples/ and included it to the test-suite
  .gitignore update to only match directories under wrappers/python3
  hides eHive jars in git
  ...

There is a trick ! They're in fact the same script :)
By looking at $0 the script can detect which type of config it should run

It should make the worker much faster

* master:
  The unit-test automatically finds all the LongMult flavours
  bugfix: the accu definition should have been changed at the same time as the others

* version/2.4:
  The unit-test automatically finds all the LongMult flavours
  bugfix: the accu definition should have been changed at the same time as the others

* master: (63 commits)
  bugfix: $pre_cond_groups could be an empty, which raises a warning
  bugfix: make sure synonymous Accu sinks of the same pipeline are correctly placed in their boxes. Thanks to Holly for reporting
  fixed kmer.t pipeline urls to loop over multiple dbs
  Added new kmer example PipeConfigs, Runnables, and tests
  More comments about the use of INPUT_PLUS in the LongMult pipelines
  allow refreshing of Storable/Cacheable objects and refresh Analysis in Worker to check for HIVE_OVERLOAD
  bugfix: rounding the float before storing as an integer to avoid warnings; thanks to Michael Paulini for reporting
  shut up the chatty "check-point" log message generator of the tail trimmer
  bugfix: make sure we pause the timer to avoid accumulating running totals
  added optional Accumulator::accu_input_variable parameter, changed examples+tests accordingly
  removed the complicated analysis_id2rc_name mapping that needed extra db activity -- doing it via collections now
  strict check for all provided arguments to help eliminate human error
  Documentation update to reflect the latest changes
  bugfix: more accurrate way of detecting that there is nothing to dump
  Don't die if a table doesn't exist
  bugfix: a file should be created even if there is nothing to dump
  bugfix: die if the input_id is not a valid hash
  [test-suite] Obviously the unit-test has to match the code ...
  bugifx: don't mark the job as DONE if it hasn't successfully dataflown to #-1
  Fixed triggers to work with role table and added PostgreSQL trigger support. Works in simple cases, needs heavier testing.
  ...

* version/2.4:
  bugfix: $pre_cond_groups could be an empty, which raises a warning

bugfix: make sure synonymous Accu sinks of the same pipeline are correctly placed in their boxes. Thanks to Holly for reporting

allow refreshing of Storable/Cacheable objects and refresh Analysis in Worker to check for HIVE_OVERLOAD

bugfix: rounding the float before storing as an integer to avoid warnings; thanks to Michael Paulini for reporting