WHIZARD gets stuck in a loop when generating new events after reading in old ones

[Christian Speckner]

Occasionally, while trying to generate new events after reading in old ones, WHIZARD gets stuck in a loop searching for a valid event. Unfortunately, I have not been able to build a testcase significantly simpler than the actual input file triggering the issue, so I am quoting the full input below. Note that, while the issue seems to arise erratically, it persists once it has surfaced - cancelling the run and retrying does not help. I have looked at the code but have been unable to identify the issue.

model = SM

alias pr = u:U:d:D:c:C:s:S:b:B:g
process drellyan = pr, pr => pr, pr

sqrts = 14 TeV
beams = p, p => lhapdf {$lhapdf_file = "cteq6l.LHpdf"}

mc = 0
ms = 0
mb = 0
mZ = 91.1875 GeV
mW = 80.45 GeV
wW = 0
wZ = 0

real iae = 137.0359895
GF = sqrt (2.) * pi / iae / mW**2 / (1 - mW**2 / mZ**2) / 2

?alpha_s_is_fixed = false
?alpha_s_from_lhapdf = true
scale = (eval Pt [extract index 1 [pr]]) / 2

cuts = all Pt > 100 GeV [pr] and 
   all abs (Eta) < 2.5 [pr]

integrate (drellyan) {iterations = 5:10000, 10:10000}

histogram ptdist (100 GeV, 4 TeV) {n_bins = 40}

analysis = record ptdist (eval Pt [extract index 1 [pr]])
n_events = 10
?unweighted = false
simulate (drellyan) {$event_normalization = "sigma/n"}

[kilian]

First observation: the loop is not endless, but after a few minutes you get

******************************************************************************
******************************************************************************
*** FATAL ERROR: Failed to generate a valid event after 100000 tries
******************************************************************************
******************************************************************************
WHIZARD run aborted.

which is exactly what happens. Have to check why, of course.

NB: This limit can be reset in limits.f90 as MAX_TRIES_FOR_SINGLE_EVENT.

[kilian]

This bug was only the tip of an iceberg which just surfaced: the various transfer routines that create the particle lists for cuts and analysis treat beams and beam remnants inconsistently. Fixing this requires some work, I'll have to do it before 2.0.4.

[kilian]

Required a thorough revision of the handling of the subevt object which is used by cuts etc. r2814 should cover all cases.