Hi, I've built oifs43r3v1 both with and without xios. But for both cases when I copy the master.exe file to the t21test directory and try to run "./job.sh" I get the following error.  I had a look at the oifs_run script that job.sh calls, and to me it looks a bit like it ends up just running mpirun without any arguments. I would really appreciate if someone had an idea about what I'm doing wrong or could give me a tip about where I could start looking for the error. 

I don't know if this is relevant, but I did get some "[info] Use of uninitialized value in subroutine"-rows during the building, but it didn't seem to fail anywhere. Though, for the version that use dummy-xios the xios.mod seemed to be called before it was created, so I did the build once where it failed a lot, but managed to create xios.mod and xios.o. Then I created a new gnu-opt folder, moved the two files to oifs/include/ and oifs/o/ in that new folder, and then did the build again, and that went thru without errors.   


Thanks,

Ane


 

6 Comments

  1. Hi Ane,

    Another OpenIFS user has very recently reported this problem and Glenn Carver found the following solution (I quote directly):

    "There is a mistake in the oifs-config.editme.sh file. There is a line in there as:

    OIFS_RUNCMD='mpirun'

    this should be commented out, but it isn't. The value is meant as an example but is not correct. The defaults used in oifs_run ARE correct but because this variable is set, it overrides those defaults.

    Please:

    1. Comment out the variable in oifs-config.editme.sh (or any copy you have made):
      #OIFS_RUNCMD='mpirun'
    2. Type:
      unset OIFS_RUNCMD
      to clear the environment variable.

    It should then work."

    Could you please give this a try and let me know if it solves your problem?

    Regards, Marcus

  2. Hi Ane,

    Thanks Marcus for replying quickly. Yes, this is a typo that will be fixed in the next update (oifs43r3v2).

    Thanks for reporting.

    BR, Glenn

  3. Hi Ane,

    Regarding your other errors. "use of uninitialized value..". This is coming from perl which the build system in OpenIFS uses. It's an annoying but harmless error message.  Changing the perl version often fixes it, but you can safely ignore it.

    Regarding the xios build errors, yes I've spotted that one and it's been fixed in the next update of OpenIFS. There is a small fix in the dependency specification in the fcm configuration files, which I can send if you need it.

    Cheers,  Glenn


  4. Thanks a lot for the quick response, that solved the problem. I got thru the test now, only thing is that after the last time step is finished I get "cat: 'ICMGGepc8+*' :No such file or directory ", is it supposed to end like that?


    Best regards, Ane

  5. Hi Ane,

    Yes, that message is fine, it's just that the t21test doesn't output all of the files a normal forecast does. It's another small bug fixed in the forthcoming update.

    As long as the output from the res file says the model is behaving correctly, you can use it for your work.  If you need any initial data for forecasting, just create a new issue and we'll provide it.

    Best regards,  Glenn

  6. Okay, thanks again! We already have the initial data.


    Best regards, Ane