[Clas_offline] compiling/running gsim on centos62

Thu Apr 4 04:14:41 EDT 2013

Hello Johann, 
thanks for your reply. Here are the results of my test. 
I logged into ifarm1102. But the environment script for tcsh gave me the 
following error:

source /group/clas/builds/test3/environment.csh
dirname: invalid option -- 't'
Try `dirname --help' for more information.

Just for trying, I ran gsim_bat after this error, but I got segmentation 
fault.

So I moved to bash shell, and here the environment script gives no error.
But when I tried to run gsim_bat, I got segmentation fault again.

Here is my command line:

/group/clas/builds/test3/bin/gsim_bat -ffread gsim_bat_eg1dvcs_b.ffread 
-mcin pi0_2000.bos -bosout pi0_2000_newgsim.gsim

And here are the details of the the error message:

GUSER_PHYS I: *** Default version of guser_phys called ***
GUSER_BOOK I: *** Default version of guser_book called ***
GUSER_BEG  I: *** Default version of guser_beg called ***

  DZVERI --- UGINIT: check bank structure 
OPTIONS : CLFSU          OK
file[ 0]->pi0_2000.bos<-
bosClose : close local file
bosClose : Call cclose(stream=1,iobptr=0,medium=2)
-100

Program received signal SIGSEGV: Segmentation fault - invalid memory 
reference.

Backtrace for this error:
#0  0x7F87338E3E77
#1  0x7F87338E4444
#2  0x3345E328FF
#3  0x3345E8CAB7
#4  0x6529B9 in ctimef_
#5  0x461CEE in gsimpar_2_bos_
#6  0x40C201 in uginit_
#7  0x40B2ED in MAIN__ at gsim_bat.F:0
Segmentation fault

I saw this kind of error before, and I actually found an email that I had 
sent three years ago to clas_offline, reporting the same error after 
compiling gsim on ifarml64. At that time I got no reply ;-), but I finally 
managed to compile it and have a working executable, playing with the 
various paths to libraries and scons options.

So, maybe we could try this way: if you give me an environment script that 
works in tcsh (probably the error I get is a trivial one, but I am not 
familiar with this stuff) and the commands you used to compile the gsim 
packages and gsim_bat, I'll try to compile them myself.

Maybe at the same time you could try run my same test on your machine 
with your executable, and see if you get my same error.
My test input file in /volatile/clas/claseg1dvcs/gsim/pi0_2000.bos , and 
the ffread is in the same place, 
/volatile/clas/claseg1dvcs/gsim/gsim_bat_eg1dvcs_b.ffread .
If you don't get errors, there are clearly issues only on my side, maybe 
some mixup in my environment.

Thank you again for your help.
best regards,
Silvia

On Wed, 3 Apr 2013, Johann Goetz wrote:

> Hi Silvia,
> Sorry it took so long to get back to you. It seems the problems you are
> seeing come from the cernlibs. I have tested some very basic simulation
> using gsim_bat but I would like you to verify that this works for you. I
> put it into /group/clas/builds/test3 and you should source this environment
> script before running anything:
>
> ssh ifarm1102        #  <-- this is the only rhel6 interactive node
>
> source /group/clas/builds/test3/environment.csh
>
> After this is done, you should see that gsim_bat is in your path and it
> should be located here:
> which gsim_bat
> /group/clas/builds/test3/bin/gsim_bat
>
> Once you test it, I will move the environment script to what I hope to be
> the "canonical" location of the CLAS run-time environment:
> /group/clas/builds/environment.csh       # <-- for tcsh and csh
> /group/clas/builds/environment.sh         # <-- for bash and sh
>
> But know that this will ONLY work on rhel6 systems and there is only one
> interactive node: (ifarm1102). This should fail on ifarm1101 which is the
> older rhel5 system.
>
>
> -- Johann <https://sites.google.com/site/theodoregoetz/>
>
>
> On Wed, Apr 3, 2013 at 8:36 AM, Silvia Niccolai <silvia at jlab.org> wrote:
>
>> Hello Dennis and Johann,
>> last week I sent the email below to the clas_offline mailing list, but so
>> far I had no reply. So I am retrying, sending it only to the two of you :-)
>>
>> I am quite concerned - but maybe it is just because I have no informations
>> on the actual situation - about the transition of the farm to centos62. If
>> all nodes are going to be replaced by the early fall (this is what I
>> understood, I'll be glad if you correct me if I am wrong) we do need to
>> have running executables for all the "standard" CLAS analysis and
>> simulation codes, mainly GSIM, user_ana, and gpp.
>>
>> If this is not achievable, for reasons that are beyond my very little
>> understanding of computing, compiling and the like, we should probably ask
>> Computer Center not to replace all farm nodes, to leave things how they
>> stand now or at least to leave enough nodes with the centos52 machines for
>> us to be able to run jobs of CLAS6 analyses.
>>
>> I think this is an important issue that doesn't concern only my group
>> (still heavily involved in CLAS6 analyses, with PhD students and post-docs,
>> at different stages of advancement), but I presume most of the CLAS
>> collaborators.
>>
>> Could you guys please give me an idea of the actual situation, and the
>> next steps? Or, if there is actually no problems and the codes can be
>> compiled fine, could you please tell me how to do that?
>>
>> Thank you very much and best regards,
>>
>> Silvia
>>
>>
>> ---------- Forwarded message ----------
>> Date: Thu, 28 Mar 2013 06:15:54 -0400 (EDT)
>> From: Silvia Niccolai <silvia at jlab.org>
>> To: clas_offline at jlab.org
>> Subject: compiling/running gsim on centos62
>>
>> Hello, I have been trying to compile and run gsim on centos62, without
>> success.
>>
>> I first tried the available executable in the "official" bin directory,
>> but it crashes with a segmentation fault.
>>
>> I then checked out from svn the latest gsim and tried to compile it. The
>> gsim packages compiles fine (both at 32 and at 64 bits), but errors arise
>> when compiling gsim_int and gsim_bat. I tried at 32 bits first (problems of
>> missing libraries, then once I found the right paths to the libraries there
>> were errors related with gfortran...) and at 64 bits (all cernlib-related
>> errors, which didn't go away no matter which available path to cernlibs at
>> 64 bits I tried).
>>
>> As there is soon to be the migration of the farm to centos62, and several
>> analyses of 6 GeV data are still ongoing, it would be really good if we
>> could have some tips on how to produce working executables of all the
>> "classic" 6-GeV codes (gsim, user_ana, gpp,...) on centos62.
>>
>> Every suggestion or offer for help is more than welcome.
>>
>> Thank you very much and best regards, Silvia
>>
>>
>