bug fixes for Jet toolbox

[jstupak]

Forward port of Dinko's JetDeltaRValueMapProducer.cc, which correctly handles strange corners of phase space.

Changes to RecoJets/JetProducers/BuildFile.xml needed for compilation.

Corrected RecoJets/JetProducers/python/QGTagger_cfi.py to read from condDB rather than a local sqlite file.

Fixed various config files to use correct jet collections.

[cmsbuild]

A new Pull Request was created by @jstupak for CMSSW_7_1_X.

bug fixes for Jet toolbox

It involves the following packages:

CommonTools/RecoAlgos
RecoJets/JetProducers

@nclopezo, @cmsbuild, @thspeer, @StoyanStoynev, @slava77, @Degano can you please review it and eventually sign? Thanks.
@yslai, @TaiSakuma, @nhanvtran this is something you requested to watch as well.
You can sign-off by replying to this message having '+1' in the first line of your reply.
You can reject by replying to this message having '-1' in the first line of your reply.
@nclopezo, @ktf you are the release manager for this.
You can merge this pull request by typing 'merge' in the first line of your comment.

[cmsbuild]

+1
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-4021/1622/summary.html

[cmsbuild]

Comparison is ready
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-4021/1622/summary.html

[StoyanStoynev]

@jstupak what are the expected changes in performance, in particular with AK5->AK4 (comparison results?)? Why is this coming in 71X at this stage and not in 72X first (it's not exactly a bug fix, is it)?

[jstupak]

There is no expected change in performance, nor is this run as part of the standard reconstruction sequence. I would call this a bugfix, since the current code produces an incorrect result.

[StoyanStoynev]

+1
No significant changes observed in RECO.

When merging the PR however "errors" occur which I am told are harmless:

...
error: addinfo_cache failed for path 'CondTools/Hcal/interface/BoostIODBWriter.h'
error: addinfo_cache failed for path 'CondTools/Hcal/interface/BoostIODBReader.h'
Removing RecoPixelVertexing/PixelVertexFinding/python/pixelVertexCollectionTrimmer_cfi.py

...

Comment if I am mistaken.

[StoyanStoynev]

(the large font was unintentional)

[cmsbuild]

This pull request is fully signed and it will be integrated in one of the next CMSSW_7_1_X IBs unless changes (tests are also fine). @nclopezo, @ktf can you please take care of it?

[StoyanStoynev]

Actually ignore the additional comment - it is for another pull request (sorry).

[slava77]

"no significant changes" is not very clear: which variables are affected?
John commented that this code is not run in RECO. If so, and there are changes, do we have a bug?

[StoyanStoynev]

There are no changes in distributions except timing. There are always some variations in timing, here is a non-representative example (from this PR):

More timing tests (DQM based) didn't show anything suspicious, and in particular - about jets.
Then, the event size - there are few changes like this (actually I show the largest by far) in the compressed size:
...
Branch Name | Average Uncompressed Size (Bytes/Event) | Average Compressed Size (Bytes/Event)
...

< recoTracks_generalTracks__RECO. 282054 64705.3

recoTracks_generalTracks__RECO. 282054 64693.6

In fact all the minute changes in sizes seem unrelated to the PR. So to recap - nothing points to problems with the PR.

[slava77]

ok, got it.
what you describe qualifies as "no changes at all" ;)

[jstupak]

So does this mean we are ready for the merge?

[slava77]

it will be decided tomorrow at/after the release planning meeting

[jstupak]

okay, thanks!