I removed the dependency and announced the changes.

Thanks, Jendrik. I made a small additional commit to be more consistent about
our naming conventions, so please pull.

I've tried out the calibarion-test and make-ipc-submission scripts, and
unfortunately the latter doesn't work for me:

$ ./make-ipc-submission seq-sat-lama-2011 64
Making sure destination is clear...
Exporting code...
Preparing distribution directory...
Removing LP code...
Changing -m32 compiler option to -m64
Packaging...
./make-ipc-submission: line 77: realpath: command not found

This is on an Ubuntu 14.04 machine. From looking around a bit, "realpath"
doesn't seem to be a very common tool. (I eventually found it with apt-file, but
not in a common package.) I'd prefer to avoid dependencies we don't need -- can
we simply get rid of the use of it? I don't think it really gives us much, does it?

Once this is done, I think this issue is also worth an email to downward-dev and
perhaps to the Fast Downward mailing list, pointing out what is gone, what lives
where now, and what the suggested replacements for the things that are gone is.

Merged and pushed.

Yes.

I incorporated your last comment as well, Malte. Can this be merged?

Review done.

I have finished the tasks from msg3090 and started a code review with the diff 
that I made after your changes, Malte, at 
https://codereview.appspot.com/89740043. The code is also uploaded in my 
bitbucket repo: https://bitbucket.org/jendrikseipp/downward/branch/issue17

Will do.

We decided:

1. We will remove plan-ipc.

2. We will keep run-calibration for now, moving it to misc and cleaning it up.

3. We will remove the things from src/dist that we don't need, clean the
remaining things up and move them to misc.

4. We will move the portfolio scripts to misc for now to be able to close the
issue, but Jendrik and Gabi will discuss what to do with them afterwards.

Jendrik, can you deal with these points and invite me to a code review? The
branch for this issue is in the master repository.

We should also think about what to do with the src/dist scripts.

In the branch for this issue, I removed the scripts that are no longer needed
with the new buildbot scripts. I also merged in the latest changes from default.
Regarding the run-calibration script, I would prefer to have its future
discussed with the others. There are opportunities for merging some of the
functionality with the new scripts from issue414, there is a potential
relationship to the buildbot, and there is a relationship to the
"speed.pypy.org" plans we have been discussing.

Since the buildbot stuff has already been moved to misc/buildbot in issue133, 
only run-calibration and the portfolio builders remain. run-calibration can 
probably be moved to misc. Gabi, where should the portfolio builders go? I would 
suggest to use a dedicated repo for them, but putting them under misc/fdss also 
seems reasonable.

Added Gabi to the nosy list since she can probably judge best how to handle the 
portfolio builder scripts.

I had a look at scripts and new-scripts. Here are my thoughts:

1. I think that everything in "scripts" can go. There is some useful and
historically interesting stuff in there (e.g. the SAS+ parser, the ASP
competition things, some experiments for old papers), but I don't think we can
afford the maintenance burden. I suggest removing these completely.

2. In "new-scripts", there are a few apparently unused bits and pieces and many
"historical" experiments that won't work any more once the "new-scripts"
experiment builders are gone. I suggest removing these completely.

3. There are also a few things in "new-scripts" that we might consider keeping:

- run-calibration-test, which we've used (or wanted to use) to test if our ports
to different operating systems produced reasonable performance

- the buildbot stuff

- the portfolio builders

In the branch for this issue, I've removed the stuff for 1.+2., but left 3.
alone for now. If anyone thinks something in 1.+2. should be migrated, please
speak up.

Regarding 3., I think these need to be migrated to use lab, since we don't want
to maintain two separate ways of running experiments indefinitely. I'd like to
be involved in this, but I don't think I'll be able to do the bulk of the work.

This issue should be generalized to get rid of the "scripts" and "new-scripts"
directory. Copying Jendrik's suggestion from issue283 which was merged into this
one:

    Should we add a README that points to the lab package in the "scripts" dir?

Adding a README about lab would be a good idea. I think we should get rid of
"scripts", too, though.

I think we'll resolve this more quickly if we discuss the issue in person.

OK, as of now (r3316), only the following scripts directories remain:

 * trunk/scripts
 * branches/landmarks/scripts
 * branches/preferred/scripts
 * branches/lama/trunk/scripts
 * branches/lama/branches/lama-integration/scripts
 * branches/lama/branches/experimental/scripts

The "landmarks" and "preferred" branches contain some paper-specific things that
we should keep, but that requires sorting out the wheat from the chaff, so that
will be a bit of work. Silvia, what shall we do with the scripts in the LAMA
directory. Can some of these be removed?

The different "scripts" directories should be merged into the trunk.

The versions currently used in the "everything" branch should more or less
supersede everything, so for the most part we should go with those.

However, we should also keep the paper-specific scripts that we have in some
branches for documentation and archival purposes. The current plan is to put
these into subdirectories of the main script directory.