cmssw: Relvals 537.0 and 538.0 failing in GCC10 IB

Two relvals are failing in gcc10 with the same stacktrace suggesting the same issue in ThePEG:


#4  <signal handler called>
#5  0x00002b94bf6c5305 in ThePEG::CFileLineReader::operator>> (this=0x2b94c1987de0, d=<error reading variable>) at CFileLineReader.cc:115
#6  0x00002b94c42e8b45 in ThePEG::FxFxFileReader::doReadEvent() () from /cvmfs/cms-ib.cern.ch/nweek-02694/slc7_amd64_gcc10/external/herwig7/7.2.2-39af8eacc71a05fc1859229135845ef8/lib/Herwig/HwFxFx.so
#7  0x00002b94c4317a8e in ThePEG::FxFxReader::readEvent() () from /cvmfs/cms-ib.cern.ch/nweek-02694/slc7_amd64_gcc10/external/herwig7/7.2.2-39af8eacc71a05fc1859229135845ef8/lib/Herwig/HwFxFx.so
#8  0x00002b94c43108a1 in ThePEG::FxFxReader::scan() () from /cvmfs/cms-ib.cern.ch/nweek-02694/slc7_amd64_gcc10/external/herwig7/7.2.2-39af8eacc71a05fc1859229135845ef8/lib/Herwig/HwFxFx.so
#9  0x00002b94c4315610 in ThePEG::FxFxReader::initialize(ThePEG::FxFxEventHandler&) () from /cvmfs/cms-ib.cern.ch/nweek-02694/slc7_amd64_gcc10/external/herwig7/7.2.2-39af8eacc71a05fc1859229135845ef8/lib/Herwig/HwFxFx.so
#10 0x00002b94c42ed6c9 in ThePEG::FxFxFileReader::initialize(ThePEG::FxFxEventHandler&) () from /cvmfs/cms-ib.cern.ch/nweek-02694/slc7_amd64_gcc10/external/herwig7/7.2.2-39af8eacc71a05fc1859229135845ef8/lib/Herwig/HwFxFx.so
#11 0x00002b94c42fb11a in ThePEG::FxFxEventHandler::initialize() () from /cvmfs/cms-ib.cern.ch/nweek-02694/slc7_amd64_gcc10/external/herwig7/7.2.2-39af8eacc71a05fc1859229135845ef8/lib/Herwig/HwFxFx.so
#12 0x00002b94bf72f55b in ThePEG::EventGenerator::doinit (this=0x2b9490563200) at EventGenerator.cc:262
#13 0x00002b94bf732be4 in ThePEG::InterfacedBase::init (this=0x2b9490563200) at ../include/ThePEG/Interface/InterfacedBase.h:246
#14 ThePEG::EventGenerator::setup (this=this@entry=0x2b9490563200, newRunName=..., newObjects=..., newParticles=..., newMatchers=...) at EventGenerator.cc:175
#15 0x00002b94bf777991 in ThePEG::Repository::makeRun (eg=..., name=...) at Repository.cc:316
#16 0x00002b94bf77ae79 in ThePEG::Repository::exec (command=..., os=...) at Repository.cc:786
#17 0x00002b94bf77be1f in ThePEG::Repository::execAndCheckReply (line=..., os=...) at Repository.cc:510
#18 0x00002b94bf77c0cb in ThePEG::Repository::read (is=..., os=..., prompt=...) at Repository.cc:566
#19 0x00002b94bf77c575 in ThePEG::Repository::read (filename=..., os=...) at Repository.cc:452

https://cmssdt.cern.ch/SDT/cgi-bin/logreader/slc7_amd64_gcc10/CMSSW_12_1_UBSAN_X_2021-08-20-2300/pyRelValMatrixLogs/run/537.0_DYToLL012Jets_5FS_TuneCH3_13TeV_amcatnloFxFx_herwig7+DYToLL012Jets_5FS_TuneCH3_13TeV_amcatnloFxFx_herwig7+HARVESTGEN/step1_DYToLL012Jets_5FS_TuneCH3_13TeV_amcatnloFxFx_herwig7+DYToLL012Jets_5FS_TuneCH3_13TeV_amcatnloFxFx_herwig7+HARVESTGEN.log#/

this follows the end of the discussion thread on https://github.com/cms-sw/cmssw/issues/34531 where the 537.0 was failing for another reason. I’m opening to separate it from the other issue which is now resolved

About this issue

Original URL
State: closed
Created 3 years ago
Comments: 21 (20 by maintainers)

Most upvoted comments

The Herwig author I spoke to says this difference wasn’t deliberate, just because thePEG code got updated, but nobody remembered to update the duplicated code FxFx code, so this patch shouldn’t cause any problems, and they’ll include it in a future release

Dominic-Stafford on Sep 1, 2021

I am testing this here https://github.com/cms-sw/cmsdist/pull/7261

smuzaffar on Sep 1, 2021

If I got it right that LesHouches.h substitutes FxFx.h resulting in diff types in it + the patch from above the whole thing is this:

diff --git a/MatrixElement/FxFx/FxFx.h b/MatrixElement/FxFx/FxFx.h
index a2133bb..23eba40 100644
--- a/MatrixElement/FxFx/FxFx.h
+++ b/MatrixElement/FxFx/FxFx.h
@@ -161,7 +161,7 @@ public:
     ISTUP.resize(NUP);
     MOTHUP.resize(NUP);
     ICOLUP.resize(NUP);
-    PUP.resize(NUP, std::vector<double>(5));
+    PUP.resize(NUP);
     VTIMUP.resize(NUP);
     SPINUP.resize(NUP);
   }
@@ -231,7 +231,7 @@ public:
    * Lab frame momentum (Px, Py, Pz, E and M in GeV) for the particle
    * entries in this event.
    */
-  vector< vector<double> > PUP;
+  vector< array<double,5> > PUP;
 
   /**
    * Invariant lifetime (c*tau, distance from production to decay im
diff --git a/MatrixElement/FxFx/FxFxReader.cc b/MatrixElement/FxFx/FxFxReader.cc
index 3dafe2c..d2340cd 100644
--- a/MatrixElement/FxFx/FxFxReader.cc
+++ b/MatrixElement/FxFx/FxFxReader.cc
@@ -1219,7 +1219,7 @@ bool FxFxReader::uncacheEvent() {
   pos = mread(pos, hepeup.MOTHUP[0], hepeup.NUP);
   hepeup.ICOLUP.resize(hepeup.NUP);
   pos = mread(pos, hepeup.ICOLUP[0], hepeup.NUP);
-  hepeup.PUP.resize(hepeup.NUP, vector<double>(5));
+  hepeup.PUP.resize(hepeup.NUP);
   for ( int i = 0; i < hepeup.NUP; ++i ) 
     pos = mread(pos, hepeup.PUP[i][0], 5);
   hepeup.VTIMUP.resize(hepeup.NUP);

and which I’m running to see if it works

mrodozov on Sep 1, 2021