MadGraph5_aMC@NLO

MSSM launch with chargino fails on MacOSX

Bug #1022415 reported by Sho Iwamoto on 2012-07-09

This bug affects 1 person

Affects		Status	Importance	Assigned to	Milestone
	MadGraph5_aMC@NLO	Fix Released	Undecided	Johan Alwall

Bug Description

Dear MadTeam,

The following command dies with an error on MacOSX:
$ ./bin/mg5
> import model mssm
> generate p p > x1+ n1
> output foo
> launch -f foo --laststep=parton

The MSSM model and Cards are the default (attached) one.
This issue occurs for "p p > x1- n1" and "p p > n1 n4" but does not for "p p > n1 n1".

The error message says
>
> refine 10000
> Refine results to 10000
> P0_qq_x1pn1
> usage: mkdir [-pv] [-m mode] directory ...
> ././ajob1: line 22: ../madevent: No such file or directory
> usage: mkdir [-pv] [-m mode] directory ...
> ././ajob1: line 60: ../madevent: No such file or directory
> usage: mkdir [-pv] [-m mode] directory ...
> ././ajob1: line 98: ../madevent: No such file or directory
> ./ajob1 run in 0.026181 s
> Error detected in "generate_events run_01 -f --laststep=parton"
> write debug file /Users/misho/hoge/madgraph5/foo/run_01_tag_1_debug.log
> If you need help with this issue please contact us on https://answers.launchpad.net/madgraph5
> MadGraph5Error : ./ajob1 didn't stop properly. Stop all computation
> quit
>

This issue occurs on MacOSX (10.7/Lion). On Linux (Debian), no error reported and the generation finishes correctly.
I checked that MG5 ver.1.4.7 (and the latest trunk @216) reproduces this issue.

I attach the log file.

---

I investigated a little bit, and I noticed that the files "ajobN" begin with, e.g.
> #!/bin/bash
> k=run1_app.log
> script=ajob1
> j=
>
i.e. the variable $j is empty.
As these files seem to be generated by gen_ximprove, this issue occurs because gen_ximprove cannot produce the string for j.

Also I noticed that some generated directories have invalid names.
For "p p > x1+ n1", the directory "SubProcesses/P0_qq_x1pn1" has the following subdirectories:
???A%ADo_%83?A%ADo%BB%86?A%ADo/ %F0m0%C1Bp%C8%DD/ G1.1/ G1.2/ G2/ G3/
I suppose that this is because I use Japanese lauguage, and try using LANG=C option, but LANG=C does not stop generating these directories.

First I thought this invalid directory forbids gen_ximprove to generate $j.
However, for "p p > n1 n4", the directory "P0_qq_n1n4" has no directory with an invalid name.
G1.1/ G1.2/ G2/ G3/ G4/ G5/ G6/ G7/ G8/
but P0_qq_n1n4/ajob1 has empty $j and ajob1 fails as I described above.
In this case, I have ajob1 to ajob8. ajob1 has empty $j, and ajob2-8 have j=G6a, G6b, G6c, G3a, G3b, G5a, G5b, respectively.

---

Here I stopped investigation. I will let you know when i notice something.

Related branches

lp:~maddevelopers/mg5amcnlo/1.4.8

Merged into lp:~madteam/mg5amcnlo/trunk at revision 217

Johan Alwall (community): Needs Fixing on 2012-07-23

Olivier Mattelaer: Approve on 2012-07-20

Revision history for this message

Sho Iwamoto (misho) wrote on 2012-07-09:

run_01_tag_1_debug.log Edit (3.1 KiB, text/plain)

Revision history for this message

Johan Alwall (johan-alwall) wrote on 2012-07-09:

Hello Sho,

I can't reproduce your problem on my Mac OSX 10.6 with gcc 4.6.0, so we'll need you to do debugging (with our help).

Can you try going to one of the SubProcesses/P**** directories and run
../bin/internal/gen_ximprove
and copy the resulting output here?

Thanks,
Johan

Revision history for this message

Sho Iwamoto (misho) wrote on 2012-07-09:

Hello,

In the directory for the process "p p > x1+ n1" with the MSSM model,

> cd SubProcesses/P0_qq_x1pn1
> LANG=C ../../bin/internal/gen_ximprove
>
is terminated with no error (success) with the following output.

Hello,

In the directory for the process "p p > x1+ n1" with the MSSM model,

> cd SubProcesses/P0_qq_x1pn1
> LANG=C ../../bin/internal/gen_ximprove                                                                                                                                     
> 
is terminated with no error (success) with the following output.

Warning: parameter  nhel_refine  not found
          setting it to default value            0
Enter fractional accuracy (<1), or number events (>1), max processes per job
, and whether to split channels (T/F)
100 1 F
Generating      100. unweighted events.
Reading total xsection    0.158E-01
 Working on creating    120.000004768372       events.
 Effective Luminosity   7.58713858466814       fb^-1.
           1  9.999999999999999E+095  7.587138584668139E-096
           2   28.2000000000000       0.269047467541423     
           3   302.000000000000       2.512297544592099E-002
           4   462.000000000000       1.642237788889207E-002
 Resetting G1.1/multijob.dat                                                                                                                                     
 Resetting G1.2/multijob.dat                                                                                                                                     
 Resetting G2/multijob.dat                                                                                                                                       
 Resetting multijob.dat                                                                                                                                          
 Improving for error            1 ???-?w??-?wP?-?w?}?-?w??f???f???f?(!?f?  0.269047467541423        0.00000000000000        462.000000000000     
 Improving for error            2 ???-?w??-?wP?-?w?}?-?w??f???f???f?(!?f?  2.512297544592099E-002   0.00000000000000        462.000000000000     
 Number of diagrams to fix           2
 Opening file ajob1                         
 Channel   0.269047467541423                1        1000
 Cross section           1           0   0.00000000000000              445
 Opening file ajob2                         
 Channel   0.269047467541423                1        1000
 Cross section           2           0   0.00000000000000              445

Revision history for this message

Johan Alwall (johan-alwall) wrote on 2012-07-09:

Hello Sho,

Many thanks for your report, which helped me find what I think is the cause for the error. There is an array limit which was too small for this particular process. Can you try with the version in branch
bzr branch lp:~maddevelopers/madgraph5/1.4.8
and let me know if this fixes the problem on your computer?

Let me know if you want me to send it as a tar file instead.

Thanks again,
Johan

Changed in madgraph5:
status:	New → Confirmed
status:	Confirmed → In Progress
assignee:	nobody → Johan Alwall (johan-alwall)

Johan Alwall (johan-alwall) on 2012-07-09

Changed in madgraph5:
status:	In Progress → Fix Committed

Revision history for this message

Sho Iwamoto (misho) wrote on 2012-07-09:

Hello Johan,

very cool! your guess seems correct.
the change of the compare operator seems to have solved the issue.
I just have checked that the error do not occur for one process.
I will check for the other several processes tomorrow.

Sincerely yours,
Sho

Revision history for this message

Sho Iwamoto (misho) wrote on 2012-07-10:

Dear Johan,

All processes which failed before the modification are now executed with no error.
Thanks a lot.

Sincerely,
Sho

Johan Alwall (johan-alwall) on 2012-07-24