Fixes for restart bugs (issue #5)

[jobordner]

This PR addresses two bugs associated with Issue #5:

1. corrupted strings written to parameters.libconfig files after a restart (bug fix by Matthew Abruzzo)
2. EnzoBlock attribute values not getting written to HDF5 files after restart.

[mabruzzo]

This all looks great to me!

I have also taken some steps to confirm that this resolves issue #5 . In fact, the following steps confirm that both the configuration files and data dumps are completely identical both before and after restarting:

1. Make the following modifications to line 26 of input/checkpoint_ppm-8.in:
   • Replacing schedule { var = "cycle"; } with schedule { var = "cycle"; list=[10,20];}
   • Inserting the line dir = ["checkpoint_ppm-8-data-%d","cycle"]; somewhere between lines 25 and 28
2. Inserting the line dir = ["checkpoint_ppm-1-data-%d","cycle"]; somewhere between lines 28 and 32 in input/checkpoint_ppm-1.in
3. Install h5diff
4. Download this script and save it as test_checkpoint.sh.
5. From the base directory of the repository separately execute bash test_checkpoint.sh 1 and bash test_checkpoint.sh 8

I also confirmed that the outputs are identical before and after restarts if we use charmrun ++p 4 instead of charmrun ++local.

[bwoshea]

@jobordner could you please fix the merge conflicts so that we can merge this? Thank you!

[mabruzzo]

I looked into this and the source of the merge conflict is really simple. The PR #53 simply moved input/checkpoint_ppm-1.in and input/checkpoint_ppm-8.in to input/Checkpoint/checkpoint_ppm-1.in and input/Checkpoint/checkpoint_ppm-8.in. The include paths were also tweaked accordingly.

This is trivial to resolve. @jobordner would you like me to take care of it?

Edit: In fact, I've confirmed that git merge master will automatically resolve this conflict.

[mabruzzo]

@jobordner Out of curiosity, do you have any idea why the tests failed the first time they were run after you made the latest commit? I saw that they were timeout bugs - do you think that they were just an anomaly?

[jobordner]

@mabruzzo The adapt-L5-P1 test hung, which shouldn't be related to any of the changes in this PR. There were also error messages from circleci about violating concurrency limits, maybe it was somehow using more cores than requested at some point during the testing. Both single and double precision tests failed (I think I only checked where in one, however). When I reran it with ssh it still seemed to be slow. Not sure what happened, but it's something to keep an eye on.

[mabruzzo]

Thanks. I'll file an issue about it so that if that kind of issue pops up again, we'll have a record of it

@bwoshea I think that this is ready to be merged

[bwoshea]

2 approves, I'm going to merge1