[Clas_offline] Fwd: unresolved issue with user_ana and Centos62 batch farm

[Paul Mattione]

I dunno if this fix ever got sent out to the listserv, and I lost the original email I sent, but here's one a few emails later.  

 - Paul

Begin forwarded message:

> From: Paul Mattione <pmatt at jlab.org>
> Date: September 9, 2013 5:05:52 PM EDT
> To: Dennis Weygand <weygand at jlab.org>
> Cc: Johann Goetz <theodore.goetz at gmail.com>
> Subject: Re: [Clas_offline] unresolved issue with user_ana and Centos62 batch farm
> 
> Well, the line it crashes on is the "read" line below (recsis/retcl_yield.F):
> 
>         if (nbyt .gt. 0) then
>          read(*,13, end=999)cline
> 
> So it's whatever fortran "read" with "*" as the first argument means (I thought it was command line, but maybe it is stdin, I dunno).  
> 
>  - Paul
> 
> On Sep 9, 2013, at 5:02 PM, Dennis Weygand wrote:
> 
>> Hello Paul,
>> 
>> Did you mean 'command line'? or stdin?
>> Dennis
>> 
>> 
>> On Sep 9, 2013, at 4:54 PM, Paul Mattione wrote:
>> 
>>> I have the same problem as well, and I think I may have found a kludge to avoid it (it doesn't "fix" it though).  
>>> 
>>> My understanding is that in this part of the code, recsis tries to read in tcl arguments from the command line.  Since we don't want to do this in a batch job anyway, in recsis/retcl_yield.F, just before the crash at line 100, I changed the lines (in my personal build):
>>> 
>>>       IF (LINTER) THEN
>>>        call asynck_c(nbyt)
>>>       else
>>>        nbyt = 1
>>>       endif
>>> 
>>> to:
>>> 
>>>       IF (LINTER) THEN
>>>        call asynck_c(nbyt)
>>>       else
>>>        nbyt = 0
>>>       endif
>>> 
>>> This skips the check-the-command-line section of the code entirely, and appears to be working for me (although I have some other issues to iron out on my monte-carlo so I'm not completely sure yet).  
>>> 
>>> I realize this isn't an ideal fix, but maybe it'll get things working for other people.  Let me know if it does.  
>>> 
>>> - Paul
>>> 
>>> On Sep 9, 2013, at 9:14 AM, Silvia Niccolai wrote:
>>> 
>>>> Dear Dennis and John,
>>>> we have managed to run the whole set of executables that are needed for 
>>>> our MC jobs (event generator, GSIM, GPP, user_ana) on the **interactive** 
>>>> Centos62 machine. Everything seems to be working fine and we are getting 
>>>> consistent results with what was obtained running the full simulation 
>>>> chain on Centos53.
>>>> 
>>>> However, there are problems when moving onto the batch farm. My 
>>>> student has met the same problem that was reported a week or so ago by 
>>>> Kijun Park (and Kijun has confirmed us that his problem has not been 
>>>> solved yet). The original email from Kijun is reported here. Did any of 
>>>> you have a chance to look into this?
>>>> 
>>>> It seems like user_ana crashes while the routine retcl_yield.F tries to 
>>>> read the tcl file. Why does this crash happen only when running on the 
>>>> batch farm and not on the interactive one? Is there any difference from 
>>>> ifarm1102 and the batch ones? I have no idea...
>>>> 
>>>> The line 100 of retcl_yield.F reads:
>>>> 
>>>>         read(*,13, end=999)cline
>>>> 13      format(a200)
>>>> 
>>>> My FORTRAN knowledge is a bit too stale lately to start looking into 
>>>> this in a useful way.
>>>> And it seems like Kijun and Nerses have already spent quite some time 
>>>> trying to look into this without finding a solution. If anyone has any 
>>>> idea, it would be great. It could be another step ahead towards 
>>>> actually managing to use the Centos62 farm (at today, used only at 
>>>> 40%, while the "old" farm is overloaded...).
>>>> 
>>>> Thank you all and best regards,
>>>> Silvia
>>>> 
>>>> 
>>>>> 
>>>>> On 08/30/2013 03:52 PM, KiJun Park wrote:
>>>>>> Hello All,
>>>>>> 
>>>>>> Do you have any idea to resolve the following issue ?
>>>>>> 
>>>>>> This is an error from "user_ana" running under a batch-farm.
>>>>>> Meantime, I didn't have any problem same code under
>>>> ifarm1102(interactively).
>>>>>> I have checked all envi. setup both cases but nothing wrong so far...
>>>>>> Thank you in advance for nay help.
>>>>>> 
>>>>>> At line 100 of file reconstruction/recsis/retcl_yield.F (unit = 5, file
>>>> = 'stdin')
>>>>>> Fortran runtime error: Sequential READ or WRITE not allowed after EOF
>>>> marker, possibly use REWIND or BACKSPACE
>>>>>> cp: cannot stat `clas.A00': No such file or directory
>>>>>> cp: cannot stat `clas.A00': No such file or directory
>>>>>> cp: cannot stat `eventtag': No such file or directory
>>>>>> cp: cannot stat `eventtag': No such file or directory
>>>>>> tput: No value for $TERM and no -T specified
>>>>>> stty: standard input: Invalid argument
>>>>>> 
>>>>>> Sincerely yours,
>>>>>> Kijun
>>>> _______________________________________________
>>>> Clas_offline mailing list
>>>> Clas_offline at jlab.org
>>>> https://mailman.jlab.org/mailman/listinfo/clas_offline
>>> 
>> 
>> --
>> Dennis Weygand
>> weygand at jlab.org
>> (757) 269-5926 (office)
>> (757) 870-4844 (cell)
>> 
>> 
>> 
>> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mailman.jlab.org/pipermail/clas_offline/attachments/20141120/dc72e185/attachment.html