A few fixes for Tracking stuffs: do not delete at endRun what is created in the constructor #1721
Conversation
…ted in the constructor
|
A new Pull Request was created by @perrotta for CMSSW_7_0_X. A few fixes for Tracking stuffs: do not delete at endRun what is created in the constructor It involves the following packages: RecoPixelVertexing/PixelTrackFitting @nclopezo, @cmsbuild, @thspeer, @slava77 can you please review it and eventually sign? Thanks. |
| @@ -37,14 +37,17 @@ | |||
| init(); | |||
| } | |||
|
|
|||
| PhotonConversionTrajectorySeedProducerFromQuadrupletsAlgo::~PhotonConversionTrajectorySeedProducerFromQuadrupletsAlgo() { | |||
| if(theRegionProducer!=NULL) | |||
| delete theRegionProducer; | |||
There was a problem hiding this comment.
delete NULL is a valid no-op by C++ standard.
|
+1 |
|
This pull request is fully signed and it will be integrated in one of the next IBs unless changes or unless it breaks tests. @ktf can you please take care of it? |
Reco fixes -- A few fixes for Tracking stuffs: do not delete at endRun what is created in the constructor
Advance RecoEgamma/ElectronIdentification data to V01-00-05.
In the modifications meant to port to multi-threaded fwk a few tracking related classes in PR #1628 (later become #1634) the creation of some objects was moved from begin run to the class constructor.
I verified that there were a few cases in which those objects were still destructed at the endRun, thus resulting in a series of segmentation violations reported, e.g., by the first three workflows in https://cmssdt.cern.ch/SDT/cgi-bin//showMatrixTestLogs.py/slc6_amd64_gcc481/www/sat/7.0-sat-02/CMSSW_7_0_X_2013-12-07-0200/pyRelValMatrixLogs/run/
This Pull Request fixes those cases. The wkfs 4.34, 4.6 and 144.6, as well as all those in the short matrix run now without crashing
Please Reco guys: check that what is in this PR still satisfy your wishes (and sign it if so)