Add test and improve error message when OutputModule SelectEvents refers to a Process that does not exist in history

[makortel]

The StdException failure reported in https://cms-talk.web.cern.ch/t/issues-with-run-379404/39046/1 was caused by an OutputModule SelectEvents to refer to a Process for which a TriggerResults data product did not exist. The exception message was just

----- Begin Fatal Exception 14-Apr-2024 18:15:47 CEST-----------------------
An exception of category 'StdException' occurred while
   [0] Calling EventProcessor::runToCompletion (which does almost everything after beginJob and before endJob)
Exception Message:
A std::exception was thrown.
vector::_M_range_check: __n (which is 4294967295) >= this->size() (which is 14)
----- End Fatal Exception -------------------------------------------------

We should make the exception message more clear, and also add tests for this case.

[makortel]

assign core

[cmsbuild]

New categories assigned: core

@Dr15Jones,@makortel,@smuzaffar you have been requested to review this Pull request/Issue and eventually sign? Thanks

[cmsbuild]

cms-bot internal usage

[cmsbuild]

A new Issue was created by @makortel.

@rappoccio, @antoniovilela, @Dr15Jones, @makortel, @smuzaffar, @sextonkennedy can you please review it and eventually sign/assign? Thanks.

cms-bot commands are listed here

[makortel]

Here is a minimal reproducer of the present exception message

import FWCore.ParameterSet.Config as cms

process = cms.Process("TEST")

process.source = cms.Source("EmptySource")
process.maxEvents.input = 3

process.out = cms.OutputModule("SewerModule",
    SelectEvents = cms.untracked.PSet(
        SelectEvents = cms.vstring("nonexistent_path:NONEXISTENT_PROCESS")
    ),
    name = cms.string("out"),
    shouldPass = cms.int32(0)
)

process.ep = cms.EndPath(process.out)

[makortel]

One test we did was to change these loops

cmssw/FWCore/Framework/src/OutputModuleCore.cc

Lines 277 to 282 in d196077

	for (unsigned int i = 0; i < n; ++i) {
	returnValue.emplace_back(uncheckedIndexFrom(s.token(i)));
	}
	for (auto token : tokensForEndPaths_) {
	returnValue.emplace_back(uncheckedIndexFrom(token));
	}

to filter out the indices for which productResolverIndex() != ProductResolverIndexValues::ProductResolverIndexInvalid. In that case the OutputModule accepted the events where the Path and Process did not exist.

We should really think and decide what the behavior should be in this case.

[makortel]

If the job reads a file, and the OutputModule SelectEvents asks for a non-existent Path from an existing Process, the OutputModule selects the events.

The behavior of non-existent Path and non-existent Process should be consistent. But we're not sure if silently accepting the Events is the correct or best approach.

[makortel]

Stack trace of the exception

#0  0x00007ffff5ba52f1 in __cxxabiv1::__cxa_throw (obj=0x7ffff1d31540, tinfo=0x7ffff5d0b1a8 <typeinfo for std::out_of_range>, dest=0x7ffff5bba260 <std::out_of_range::~out_of_range()>)
    at ../../../../libstdc++-v3/libsupc++/eh_throw.cc:81
#1  0x00007ffff5b9c355 in std::__throw_out_of_range_fmt(char const*, ...) [clone .cold] () from /cvmfs/cms.cern.ch/el8_amd64_gcc12/external/gcc/12.3.1-40d504be6370b5a30e3947a6e575ca28/lib64/libstdc++.so.6
#2  0x00007ffff7c59326 in std::vector<edm::propagate_const<std::shared_ptr<edm::ProductResolverBase> >, std::allocator<edm::propagate_const<std::shared_ptr<edm::ProductResolverBase> > > >::_M_range_check (
    this=<optimized out>, __n=<optimized out>) at /cvmfs/cms.cern.ch/el8_amd64_gcc12/external/gcc/12.3.1-40d504be6370b5a30e3947a6e575ca28/include/c++/12.3.1/bits/stl_vector.h:1153
#3  std::vector<edm::propagate_const<std::shared_ptr<edm::ProductResolverBase> >, std::allocator<edm::propagate_const<std::shared_ptr<edm::ProductResolverBase> > > >::at (__n=<optimized out>, this=<optimized out>)
    at /cvmfs/cms.cern.ch/el8_amd64_gcc12/external/gcc/12.3.1-40d504be6370b5a30e3947a6e575ca28/include/c++/12.3.1/bits/stl_vector.h:1194
#4  edm::Principal::prefetchAsync (this=<optimized out>, task=..., index=<optimized out>, skipCurrentProcess=<optimized out>, token=..., mcc=<optimized out>) at src/FWCore/Framework/src/Principal.cc:645
#5  0x00007ffff7cb683a in edm::Worker::prePrefetchSelectionAsync (this=0x7fffdef3fdc0, group=..., successTask=<optimized out>, token=..., id=..., iPrincipal=0x7fffdeee8790)
    at src/FWCore/Framework/src/Worker.cc:198
#6  0x00007ffff7c55432 in edm::Worker::doWorkAsync<edm::OccurrenceTraits<edm::EventPrincipal, (edm::BranchActionType)1> > (this=0x7fffdef3fdc0, task=..., transitionInfo=..., token=..., streamID=...,
    parentContext=..., context=0x7fffe406d090) at /build/mkortela/debug/tier0_Run379404/CMSSW_14_0_4/src/FWCore/Framework/interface/maker/Worker.h:1045
#7  0x00007ffff7c5582d in edm::WorkerInPath::runWorkerAsync<edm::OccurrenceTraits<edm::EventPrincipal, (edm::BranchActionType)1> > (context=0x7fffe406d090, streamID=..., token=..., info=..., iTask=...,
    this=<optimized out>) at /build/mkortela/debug/tier0_Run379404/CMSSW_14_0_4/src/FWCore/Framework/interface/WorkerInPath.h:119
#8  edm::Path::runNextWorkerAsync (this=0x7fffe4024b00, iNextModuleIndex=<optimized out>, iInfo=..., iToken=..., iID=..., iContext=0x7fffe406d090, iGroup=...) at src/FWCore/Framework/src/Path.cc:349
#9  0x00007ffff7c55b0f in edm::Path::processOneOccurrenceAsync (this=0x7fffe4024b00, iTask=..., iInfo=..., iToken=..., iStreamID=..., iStreamContext=0x7fffe406d090)
    at /build/mkortela/debug/tier0_Run379404/CMSSW_14_0_4/src/FWCore/Concurrency/interface/WaitingTaskHolder.h:78
#10 0x00007ffff7ca570c in edm::StreamSchedule::processOneEventAsync (this=0x7fffe406ce90, iTask=..., info=..., serviceToken=..., pathStatusInserters=...) at src/FWCore/Framework/src/StreamSchedule.cc:1088
#11 0x00007ffff7c85c41 in edm::Schedule::processOneEventAsync (this=<optimized out>, iTask=..., iStreamID=<optimized out>, info=..., token=...) at src/FWCore/Framework/src/Schedule.cc:1202

[makortel]

This line

cmssw/FWCore/Framework/src/Worker.cc

Line 196 in d196077

if (productResolverIndex != ProductResolverIndexAmbiguous) {

should also require the index to not be ProductResolverIndexValues::ProductResolverIndexInvalid

[makortel]

The behavior of non-existent Path and non-existent Process should be consistent. But we're not sure if silently accepting the Events is the correct or best approach.

We think the job should fail if the SelectEvents refers to a non-existent TriggerResults data product.

[makortel]

The exception for a non-existent Path is thrown via

#0  0x00007ffff5ba52f1 in __cxxabiv1::__cxa_throw (obj=0x7fffcb3c4d00, tinfo=0x7ffff79a6610 <typeinfo for edm::Exception>, dest=0x7ffff7970020 <edm::Exception::~Exception()>) at ../../../../libstdc++-v3/libsupc++/eh_throw.cc:81
#1  0x00007ffff7b9d31e in edm::EventSelector::initPathNames (this=0x7fffdeeaf440, pathNames=...) at src/FWCore/Framework/src/EventSelector.cc:167
#2  0x00007ffff7c1a7d8 in edm::EventSelector::acceptEvent (this=0x7fffdeeaf440, tr=...) at src/FWCore/Framework/src/EventSelector.cc:246
#3  0x00007ffff7cb60ba in edm::detail::NamedEventSelector::match (product=..., this=0x7fffdeeaf3c0) at /build/mkortela/debug/tier0_Run379404/CMSSW_14_0_4/src/FWCore/Framework/interface/TriggerResultsBasedEventSelector.h:35
#4  edm::detail::TriggerResultsBasedEventSelector::wantEvent (this=<optimized out>, ev=...) at src/FWCore/Framework/src/TriggerResultsBasedEventSelector.cc:158
#5  0x00007ffff7c4ce9e in edm::core::OutputModuleCore::prePrefetchSelection (this=0x7fffcb08e438, id=..., ep=..., mcc=<optimized out>) at src/FWCore/Framework/src/OutputModuleCore.cc:306
#6  0x00007ffff7cbb72d in operator() (__closure=0x7fffcb455960) at src/FWCore/Framework/src/Worker.cc:168
#7  edm::FunctorWaitingTask<edm::Worker::prePrefetchSelectionAsync(tbb::detail::d1::task_group&, edm::WaitingTask*, const edm::ServiceToken&, edm::StreamID, const edm::EventPrincipal*)::<lambda(const std::__exception_ptr::exception_ptr*)> >::execute(void) (this=0x7fffcb455940)

[makortel]

+core

[cmsbuild]

This issue is fully signed and ready to be closed.