The regex part of the C++11 standard was not implemented before june 2015.
To avoid forcing users to have a very recent compiler, switching to boost instead.

The problem is probably related to some IO buffering issue.
Batsim is now run in quiet mode in the tests, which hides the problem :(

Merge branch 'data_storage' of git+ssh://scm.gforge.inria.fr/gitroot/batsim/batsim into data_storage

Merge branch 'data_storage' of git+ssh://scm.gforge.inria.fr//gitroot/batsim/batsim into data_storage

The bash scripts generated by experiment management scripts seems fine,
but not when they are executed from the experiment management script themselves...

Running the old version of pybatsim_filler_tiny.yaml worked.

Running the old version of pybatsim_filler_medium.yaml didn't worked.
The python script was not able to recover batsim's command stdout/stderr, but
Batsim was executed. Batsim's execution was strange, it stopped depending on
the number of characters printed by XBT_INFO before the end of function
Job * Job::from_json(const rapidjson::Value & json_desc, Workload * workload)...
By "stopped", I mean being stuck in some blackhole, as the process did not
crash but maybe stuck in a blackhole inside XBT_INFO.

Simply modifying the batsim_command in these scripts stresses the fact that
the way the python scripts run the bash ones is bugged. Valgrind is executed
correctly if the bash scripts are executed manually, but cannot be executed
because of parsing options errors when python executes bash.

So, to conclude the investigations done this evening/night:
  - Batsim may contain some bufferoverflow-like issues, which can only
    be observed under certain circumstances, like being executed by
    python->bash->batsim but not bash->batsim nor batsim. Couldn't
    find any tonight but this kind of bug is tricky to detect.
  - SimGrid may generate Black holes in XBT_INFO. Unlikely.
  - The way the bash scripts are executed by Execo is buggy.
    Further investigations must be done to determine why some bash
    files work and others don't

+ task class modification I forgot to submit

Merge branch 'data_storage' of git+ssh://scm.gforge.inria.fr/gitroot/batsim/batsim into data_storage

Merge branch 'data_storage' of git+ssh://scm.gforge.inria.fr//gitroot/batsim/batsim into data_storage

Merge branch 'data_storage' of git+ssh://scm.gforge.inria.fr/gitroot/batsim/batsim into data_storage

Merge branch 'data_storage' of git+ssh://scm.gforge.inria.fr/gitroot/batsim/batsim into data_storage

Merge branch 'data_storage' of git+ssh://scm.gforge.inria.fr/gitroot/batsim/batsim into data_storage

Moved internal IDs in Redis to be of format WORKLOAD!NUMBER

Merge branch 'data_storage' of git+ssh://scm.gforge.inria.fr//gitroot/batsim/batsim into data_storage

Merge branch 'data_storage' of git+ssh://scm.gforge.inria.fr/gitroot/batsim/batsim into data_storage

Conflicts:
	src/job_submitter.cpp

Pybatsim should now be able to handle the new job_identifier syntax WLOAD!ID.

Batsim now stores the number of simulated machines into Redis.
Since the scheduler does not need to read the workload file anymore,
this argument has been removed from pybatsim's launcher.py.