Verifying/Testing generated root files

[cschwan]

It would be nice if were able regression test the root files generated by amcfast. The problem is that, when I run mg5_aMC+aMCfast+APPLgrid once, store the root file, and run the same commands again, I get a different root file. One difference is due to the automatic seeding of mg5_aMC, but this should be fixed by set run_card iseed 1 (it isn't, I still get a different file).

Intuitively, a run with the same seed should give the same events and therefore the same file. Does anyone know why this is not the case?

[scarrazza]

Ok, so lets parse the ROOT file and compare for each key/object its content.

[scarrazza]

In a few words we need a rootdiff program.

[cschwan]

I have a preliminary version of our rootdiff program. It's more or less tailored to *.root files which are in fact applgrids. One thing which complicates the comparison of two applgrids is the fact that they store a custom type TFileString which looks like a class from ROOT, but it's actually implemented in applgrid, in such a way that it's not visible from outside (there's no header file installed for TFileString which would allow to reconstruct this type). But I think that, for the time being, it's safe to ignore this entry in the .root file.

I have verified that the diff works if I run MG5_aMC twice with the same seed and that it shows differences if I modify the seed in one of the runs.

[scarrazza]

Sounds good, could you please open a pull request with the script so I can test?

[cschwan]

I added the python script in commit f742eab. This script is adapted from https://github.com/wiso/ROOTDiff.