Yes, let's close this.

Jendrik, in my view we can close this due to lack of activity. Since you were also involved in the discussion, I'll leave the closing to you if you agree.

Unassigning this from myself.

This has been clarified offline: I misunderstood the requirements here and indeed 
the portfolio code does not have the issue anymore, but there are other places in 
the code that need to take externally set limits into account (e.g. invariant 
synthesis and soon limits for building abstractions).

Can you summarize the current status before we close this? I'd be surprised to
hear that the translator issue has been addressed.

What do people need to do if they want to run the planner with a certain time
and/or memory limit? There are probably three cases to consider:

- regular configurations
- anytime configurations (e.g. LAMA-2011 config)
- portfolios

Even if all this is already working, I don't think it's documented properly.

This issue can probably be closed, right?

I've made a band-aid fix to this in the ipc-2011-fixes branch (changeset
4aaeb96a7a65) which simply sets a memory limit of 6 GB minus 64 MB to hopefully
avoid being killed by the IPC organizers' memory limit enforcers. The more
general problem still remains unaddressed.

For the IPC, our complete planner run will be killed if one process exceeds the
externally set memory limit. This is bad for our portfolio configurations: if
one of them exceeds the memory limit, we would like to try out the other
configurations, but can't if our planner is killed.

Related to this, if someone sets an external time limit other than 1800 seconds
(e.g. for probing runs), our portfolio code and the invariant synthesis timeout
in the translator aren't prepared for that. We should make all of them more
flexible in this regard.