Re: BLCR tests problem with MC

From: Paul H. Hargrove (PHHargrove_at_lbl_dot_gov)
Date: Thu Mar 12 2009 - 06:08:30 PDT

  • Next message: Igor Volovichev: "Re: BLCR tests problem with MC"
    Igor,
    
       Thanks for reporting your "make check" failures before proceeding to use 
    BLCR.  To answer your last question first: Yes, it should be safe to use BLCR 
    with OpenMPI and Torque on your system.
    
       Before today I had not used MC.  However, when I tried it I was able to 
    reproduce your problem on the first try, using a system that has not shown any 
    BLCR test failures in the past.  So, to answer your first question: this 
    *does* appear to be "normal" behavior, or at least this is not something 
    specific to your environment.
    
    Running inside MC I saw the following failures and no others:
    
    > Detected 3 failures in ./cs_enter_leave.st:
    > #ST_ALARM:60
    > !!! Alarm clock expired
    > !!! Missing final DONE
    > !!! Test killed unexpectedly by signal 9
    > FAIL: cs_enter_leave.st
    > Detected 3 failures in ./cs_enter_leave2.st:
    > #ST_ALARM:60
    > !!! Alarm clock expired
    > !!! Missing final DONE
    > !!! Test killed unexpectedly by signal 9
    > FAIL: cs_enter_leave2.st
    > Detected 3 failures in ./cr_tryenter_cs.st:
    > #ST_ALARM:60
    > !!! Alarm clock expired
    > !!! Missing final DONE
    > !!! Test killed unexpectedly by signal 9
    > FAIL: cr_tryenter_cs.st
    
    These three tests are written very similarly and while I don't know the source 
    of the problem yet, it is almost certainly the same cause for all three.  I 
    hope to understand (and then resolve) the root cause, but for now you are safe 
    ignoring this problem (as long as the tests all pass when run under bash).
    
    Could you please confirm that you see only these three tests fail?
    
    -Paul
    
    Igor Volovichev wrote:
    > I faced strange problem: when I do "make check" under Midnight Commander
    > some test fail (e.g. sc_enter_leave*, etc.). Whereas from bash promt all test
    > pass. Moreover, in MC redirecting stderr to a file (make check
    > 2>/tmp/some_file) make test pass too. Is it normal behavior or there is
    > something wrong with blcr in my environment? Is it safe to use blcr with
    > OpenMPI and Torque? Thanks to all for an advice.
    
    -- 
    Paul H. Hargrove                          PHHargrove_at_lbl_dot_gov
    Future Technologies Group
    HPC Research Department                   Tel: +1-510-495-2352
    Lawrence Berkeley National Laboratory     Fax: +1-510-486-6900
    

  • Next message: Igor Volovichev: "Re: BLCR tests problem with MC"