GitLab

Originally I had hoped to make this a more encompassing change, but
for now this was a lot simpler.

Just so that services can be added to EventLoop jobs.

Mostly a copy-and-paste of AnaAlgorithm, with cleanup and updates to
inherit from AthReentrantAlgorithm.

That puts one degree of separation between EventLoop and the algorithm
classes, which besides cleaning up the code a bit should also allow to
add separate wrappers for re-entrant algorithms, public tools and
services.

This combines two updates into one: It creates a new manager, that
takes care of everything related to handling the submission directory
itself.  And it defines a series of modes for creating the submission
directory.

The modes for the submission directory are probably more interesting
for the users.  There are now for modes:

* "no_clobber": create the directory and if that fails raise an error.
  this is the default behavior we had so far.

* "overwrite": create the directory, removing anything in the way.
  this was the behavior so far when `optRemoveSubmitDir` was
  specified.

* "unique": append a unique suffix to the submit directory name.  this
  is mostly meant for unit-tests in which one may specify the same
  submit directory multiple times or in multiple tests.

* "unique-link": same as "unique" but create a link under the
  officially requested name to point to the newly created directory,
  replacing any previously existing link.  the idea is that this is
 ...

That allows to setup resources that are shared among all jobs.
Originally I wanted to put the storage there, but that turned out more
complicated and I'm leaving it for another day.

That makes it a little easier to write, and it allows the user to
replace the configuration file if they need something different.

This can't do anything too fancy, but if you have a shared file system
it ought to work as is.

Just as a first attempt to see how easy it is to run inside docker.
So far, so good.

This makes the setup both cleaner and more robust.

The goal is that ultimately histogram output files ought to be a
normal, albeit special output stream.

These 5 properties should be respected by all new code introduced for checking on
the memory behaviour of user jobs.

The ultimate goal here is to have it be one of the cornerstones of a
new simplified core configuration engine, but for now it is just a way
of making sure algorithms get streamed correctly.


Former-commit-id: ea45ebe6974e87bc15594ab64e2079b86c0efc09

Seems a fair rename, as Attila didn't like the name EventLoopComps.


Former-commit-id: ab6a6ccbce6199312212dce4cd4144a18dc881d0

- add optBatchSetupCommand to override the default asetup, guessing the release based on environment variables for Base, Top
- use a regular tar command that tarballs the /home/kratsg/rel21testing/build/x86_64-slc6-gcc62-opt and copies this into the condor submission folder for submitting
- clean up some code and make it a little more consistent


Former-commit-id: 823798f7a2576884f7e86533cc61c786584905e8

This should address a number of issues with the existing algorithm
base class.


Former-commit-id: b3ef0a61e9a5f21d1114350f562e381609b1279d

Former-commit-id: 87515bb9d0dc1086c5247e963b9f3bd8babccaef

Former-commit-id: f5472df765a81d46c43ed28f0fad0e49d24e8fb5

Former-commit-id: 18df04923918cab970d87b29cfe6af8994b22e8d

necessary for the analysis release(s).


Former-commit-id: baa5949f160718f55c99dbc034497a6265a7e300