To CMSSW_10_2_4_patch1

[dabercro]

We need to update CMSSW versions for the new data. Of course, CMSSW_10_1_4 -> CMSSW_10_2_4 isn't backwards compatible.

A naive attempt to get things to compile gives:

• data with bad PUPPI weights (http://t3serv001.mit.edu/~jenkins/relval/dabercro/PandaProd/2018D/4/SingleMuon_2018A_0/pfCandidates/puppiW_log.pdf)
• hanging during PFCandsFiller or HLTFiller. Unfortunately, running over the same file does not produce consistent locations for the hanging.

[dabercro]

Just happened to ask Phil. Apparently the newer data literally has a different format, so we'll have to run separate releases for the different datasets.

[yiiyama]

Actually - central NANOAOD production uses 102X for all Run2 samples (https://twiki.cern.ch/twiki/bin/viewauth/CMS/PdmVAnalysisSummaryTable). It should be possible to process all three years with panda 014. What's the format difference you were referring to?

[dabercro]

There was a change in the way Puppi Weights were packed. It went from something like using a smaller float format to using a short to express a fraction between -1 and 1 (weight = value/short::max) or vice versa. I would have to check. Unfortunately I cleaned up my plot there...

[yiiyama]

I see, so we need to add a switch to the filler then.

[dabercro]

I can't seem to duplicate the problem. I don't remember exactly how we fixed it before, but maybe everything is fine now.