Using test/baseline/rebase.py to rebase test results

[markcmiller86]

I had a slew of test results from a newly added test that I generated on a macOS system. But the nightly test suite runs on Linux, with different compilers and I suspect even different rendering code paths. So, many of the files do not match pixel-for-pixel with what is produced in nigthly testing on Linux. What to do?

The answer is we have a rebase tool in test/baseline/rebase.py that makes this very easy. It will copy files from the dashboard to their appropriate place in the baseline directory tree. Here are the steps...

$ pwd
visit
$ git checkout develop
$ git pull
$ git checkout -b task-mcm86-11nov24-rebase-annot-macros
$ cd test/baseline
$ ls
databases	hybrid		operators	plugins		quickrecipes	rendering	simulation
faulttolerance	meshtype	plots		queries		rebase.py	session		unit
$ ./rebase.py --help
.
.
.
Examples...

To rebaseline *all* files from oldsilo test from date tag 2018-04-07-09:12
 
    ./rebase.py -c databases -p oldsilo -m serial -d '2018-04-07-09:12'

To rebaseline silo_00.png & silo_01.png files from oldsilo test from same tag 

    ./rebase.py -c databases -p oldsilo -m serial -d '2018-04-07-09:12' 'oldsilo_0[0-1].png'

To be interactively prompted upon each file to rebaseline from oldsilo test

    ./rebase.py -c databases -p oldsilo -m serial -d '2018-04-07-09:12' --prompt

I wanted to rebase from results shown on the dashboard with date, 2024-11-10-22:00, in the URL. I wanted to rebase all restuls from the annot_macrosy.py test which is in the rendering category. The command was...

$ ./rebase.py -c rendering -p annot_macros -m serial -d '2024-11-10-22:00' 'annot_macros*.png'

As it copies files, it will warn about file size change because the local file is probably an LFS file (always around about 130 bytes) and the rebased file will be much larger because it is the actual .png file. That is ok.

Once the rebase is complete, commit and push changes and create a PR.

[JustinPrivitera]

Very cool