Error in test_report.py

[jrood-nrel]

It seems like something may have changed in AMReX that is causing test_report.py to parse the some output wrong? I get errors in test_report.py:641 where in this code:

                if fields[0].startswith("variable"):
                    ht.header(fields)
                    continue

fields[0] is an out of bounds access. This was working until something changed March 4 and I don't update this repo nightly, but my testing updates AMReX nightly.

[drummerdoc]

No clue. Can you submit an issue in the amrex repo? I'll ask around when I get into the office.

[zingale]

this may be due to the change in the compare tools. Can you point to the html output from a case that did work for this test?

[jrood-nrel]

@zingale This repo https://github.com/AMReX-Combustion/PeleCRegressionTestResults/tree/master has the html output (I had to clone it locally to view the rendered HTML files). Everything but the Sod-1d test case seems to work. Looking at the HTML for the Sod-1d test case it has things like < variable not present in both files >

[jrood-nrel]

Here is some debugging output where I print line which is used to populate q and then fields:

                print(line)
                fields = [q.strip() for q in line.split("  ") if not q == ""]

                if fields[0].startswith("variable"):
                    ht.header(fields)
                    continue

debug.txt

[jrood-nrel]

It seems to me like fields needs to be guarded in all references to it in case it is null due to q being blank. Or the if not q == "" just needs to be removed?

[zingale]

The < variable not present in both files > arises when you have added variables to the plotfile in your code recently, that are not in the benchmark.

Can you send me the tarred up output file the comparison is failing on along with the benchmark so I can try to reproduce this? It is not obvious what the issue is.

[jrood-nrel]

I guess one thing I haven't done is update the benchmarks to try to solve this. Partly because changes have been made in PeleC that we need to verify before updating the benchmarks. Beyond trying to reproduce this, does #42 look like a fair solution? It seems like a lot of work to try to recreate this elsewhere, and the benchmarks are 120MB. Maybe this will go away once we update the benchmarks, so lets wait until I am able to do that.

[jrood-nrel]

https://amrex-combustion.github.io/PeleCRegressionTestResults/ shows the results using #42 .