Code review comment for lp:~maddevelopers/mg5amcnlo/3.0.1

Thanks,

Update on my side:
1) the IOTest suite are bugged (not critical for a release) I have ask Valentin to look at it
2) the (fast) parrelel test leads to their own discovery of bug:

2.1) e+ e- > t t~ [real=QCD]
crashes with
FKSProcessError: This process does not have any correction up to NLO in QED,QCD
[ test: test_short_check_generate_events_nlo_py6pt_fsr]
same for e+ e- > p p p [real=QCD] in test: test_short_check_eejjj_lo_lhapdf]

2.2) A division by zero occurs in the jet veto of rik [test_short_jet_veto_xsec]

p p > e+ ve [QCD]
generate_events NLO
                 set ickkw -1
                 set ptj 10

2.3) The value of some squared order amplitude does not match anymore [test_short_sqso ]
Will look at this one (maybe with Valentin).

Cheers,

Olivier

On 9 Nov 2018, at 22:00, marco zaro <<email address hidden><mailto:<email address hidden>>> wrote:

Hi Again,
actually, doing p p > e+ ve [LOonly] works ok, at LO, NLO, aMC@LO, aMC@NLO
So it may be something relate to the phase-space of 2->1?
the reason why p p > w+ [LOonly] returns zero is because the function passcuts always returns False, exiting here
c Make sure have reasonable 4-momenta
     if (p(0,1) .le. 0d0) then
        passcuts=.false.
        return
     endif

with P(0,1) = -99

So clearly something goes wrong in the phase space generation. Rikkert, anything that could be related to the granny stuff?

also note that
p p > w+ [QED] has one test failing (soft test for photon emission off the W), where a lot of NaN’s appear
p p > e+ ve [QCD] seems ok though

Cheers,

Marco

On 9 Nov 2018, at 21:42, marco zaro <<email address hidden><mailto:<email address hidden>>> wrote:

Hi Olivier,
it is strange, it seems that the LOonly mode is broken (it always returns zero)
I will have a look on monday, and let you know.

cheers,

Marco

On 9 Nov 2018, at 21:10, Olivier Mattelaer <<email address hidden><mailto:<email address hidden>>> wrote:

Hi Marco,

The following test is also not passing (going trough ok with 2.6.4):
test_madspin_LOonly

Even though this is a madspin designed tests, it actually crash within the LOonly module.
If you just do
geneate p p > w+ [LOonly]
It will lead to a crash.

I have fixed all the other tests (but a couple of IOtests and i still have to run the short paralel test).

Since this is a beta version, we can also decide to not fix this in this version.
--
https://code.launchpad.net/~maddevelopers/mg5amcnlo/3.0.1/+merge/356133
You are requested to review the proposed merge of lp:~maddevelopers/mg5amcnlo/3.0.1 into lp:~maddevelopers/mg5amcnlo/FKS_EW_granny.

--
https://code.launchpad.net/~maddevelopers/mg5amcnlo/3.0.1/+merge/356133
You are requested to review the proposed merge of lp:~maddevelopers/mg5amcnlo/3.0.1 into lp:~maddevelopers/mg5amcnlo/FKS_EW_granny.

--
https://code.launchpad.net/~maddevelopers/mg5amcnlo/3.0.1/+merge/356133
You are requested to review the proposed merge of lp:~maddevelopers/mg5amcnlo/3.0.1 into lp:~maddevelopers/mg5amcnlo/FKS_EW_granny.

« Back to merge proposal