MLM pythia6 test stuck

[Juergen Reuter]

It seems that PYINIT calls make MLM_pythia6 test getting into a deadlock.

[Bijan Chokoufe Nejad]

It works with LHAPDF5 and with no LHAPDF (test is skipped). With LHAPDF6 I get right now

PYINIT: initialization of PYTHIA routines *

LHAPDF6 USING PYTHIA-TYPE LHAGLUE INTERFACE

terminate called after throwing an instance of 'LHAPDF::ReadError?'

what(): Info file not found for PDF set 'cteq6l1'

After installing cteq6l1, I see different numbers for integration :/. I will reduce this test again.

[Bijan Chokoufe Nejad]

Should be fixed with r6888 but JR will confirm.

[Juergen Reuter]

After splitting the WHIZARD ISR MLM matching from the PYTHIA ISR MLM matching tests JRR still sees the problem of PYTHIA hanging. Also it is strange that PYTHIA tries to use multiple interactions. Several issues summarized in this PYTHIA output:

1****************** PYINIT: initialization of PYTHIA routines *****************
==== LHAPDF6 USING PYTHIA-TYPE LHAGLUE INTERFACE ====
LHAPDF 6.1.5 loading /usr/local/share/LHAPDF/cteq6l1/cteq6l1_0000.dat
cteq6l1 PDF set, member #0, version 4; LHAPDF ID = 10042

 ==============================================================================
 I                                                                            I
 I       PYTHIA will be initialized for p+ on pbar- user configuration        I
 I            with    980.000 GeV on    980.000 GeV beam energies             I
 I                                                                            I
 I           corresponding to   1960.000 GeV center-of-mass energy            I
 I                                                                            I
 ==============================================================================
 Warning: requested subprocess  96 has vanishing cross-section.
 Process switched off!

 ******** PYMAXI: summary of differential cross-section maximum search ********

           ==========================================================
           I                                      I                 I
           I  ISUB  Subprocess name               I  Maximum value  I
           I                                      I                 I
           ==========================================================
           I                                      I                 I
           I    4   User process 1                I    1.0000D-12   I
           I   96   Semihard QCD 2 -> 2           I    0.0000D+00   I
           I                                      I                 I
           ==========================================================

 ****** PYMULT: initialization of multiple interactions for MSTP(82) = 4 ******
        pT0 = 1.84 GeV gives sigma(parton-parton) =      NaN mb: accepted

 ****** PYMIGN: initialization of multiple interactions for MSTP(82) = 4 ******
        pT0 = 1.84 GeV gives sigma(parton-parton) =      NaN mb: accepted

 ********************** PYINIT: initialization completed **********************

[Juergen Reuter]

Resetting the mode for the underlying event MSTP(82) = 1 solves the problem. However, the details why this is necessary are not known. Nevertheless, closing this ticket.

[Juergen Reuter]

Now it fails for LHAPDF5. There need to be two versions, one for LHAPDF5 and one for LHAPDF6.

[Bijan Chokoufe Nejad]

I just realized why multiple interactions are activated. Rick Field's Tune sets

MSTP(81) = 1, MSTP(82) = 4

We shouldnt activate this tune for unit testing. I still hope we can avoid two versions of the test for LHAPDF5/6 but if even the sum of momenta is different, I guess this is not possible.

[Bijan Chokoufe Nejad]

OK with the stripped version (r6896) it mlm_pythia6_isr works with LHAPDF5/6 for me. JR might want to confirm. But in mlm_matching_isr I just spotted an error in the transferral (one of outgoings somehow is assigned as incoming). Will be fixed by me.

[Bijan Chokoufe Nejad]

Done in r6898. Hopefully all bugs have been fixed. It remains to be tested if mlm_matching_isr works under 32 bit as it is numerically quite sensitive. If it does not work, we can

a) go back to only grepping for momenta or even

b) not look at the event record at all (not preferred but maybe necessary)

[Juergen Reuter]

At least that one seems to be fixed now.