Upgrade to Silo 4.12

[JustinPrivitera]

[cyrush]

.... and hdf5 1.14.4

[markcmiller86]

So, I am now testing on HDF5-2.0. Since its a totally new version of HDF5 than I have been using these few weeks, I wanna run Silo's whole test suite with it under Address Sanitizer and see what I see.

I've run VisIt's tests with HDF5-2.0 and Silo 4.12.0 and I am seeing one issue opening HDF5-1.8.4 files. I am hopeful it is something small. HDF5 2.0 can certainly open the files. But, Silo 4.12.0 can't at the moment.

[markcmiller86]

Pretty sure I know what problem I am running into here. Some of the data files in silo_hdf5_test_data.tar.xz utilize zlib compression on the datasets. These are from an old version of silo, 4.7.2, that wasn't too smart about what datasets it applied compression to. One of those datasets is a small, character dataset, _silolibinfo written to the root directory of a Silo file and intended to capture Silo library information at the time the file was created.

Well, DBOpen() will read that dataset, if it is present. However, that dataset is zlib compressed in the file and the version of HDF5 I have built does NOT have zlib compression enabled. So, the read fails and so the DBOpen() fails.

I kinda sort think this is a Silo library bug. I can make DBOpen() behave a bit better in this circumstance but the rest of the objects in the file will be unreadable without a zlib enabled HDF5 anyways.

[markcmiller86]

Ok, encountering a SEGV with the AMR tests in silo.py using Silo 4.12.0 and HDF5 2.0. All other tests in silo.py pass...

Well, wait a min...the SEGV is happening opening a PDB file. Hummm...still investigating.

[markcmiller86]

I had things incorrectly configured and that was leading to the SEGV in one of the silo.py tests. Fixed now. Proceeding with confirming remaining tests with VisIt pass.

[markcmiller86]

Ok, all silo.py tests pass with Silo 4.12.0 and HDF5 2.0...

silo_4120_hdf52_test.tar.gz

[markcmiller86]

ASAN testing on HDF5-2.0 also shows no issues.