Re: Memory leak somewhere.

From: Christian Holm Christensen (cholm@hehi03.nbi.dk)
Date: Wed Oct 31 2001 - 14:11:53 EST

  • Next message: Hironori Ito: "Vertex by ZDC, BB and TPM1"

    On Wed, 31 Oct 2001 13:44:30 -0500
    Jens Ivar Jordre <jensivar.jordre@fi.uib.no> wrote
    concerning "Memory leak somewhere.":
    > Howdy brats.
    > 
    > I'd just like to inform you about a memory leak that Hiro and I have
    > spotted.  Thereby you may look out for it, and clearly N pairs of
    > eyes more easily spots the problem than 2, provided N > 2 of
    > course. :)
    
    Using bratmain, with the global RDO modules + some user code, I've not
    seen _any_ memory leaks - it's rock steady at 22MB (or something like
    that anyway). 
    
    > We find that the leak occurs when using BrEventIO in tree mode for
    > output or when useing BrOutputTreeModule. This could point to
    > somewhere in their common mother class BrIOModule, but it is not
    > certain that the problem lies here. Trying to quantify the leak I
    > lose about 16 MB in 7000 events. But don't put to much into this
    > number, as one would imagine that the number of bytes per event
    > depends on the content of the output event node.
    
    Four very important pieces of information: 
    
    * What ROOT version are you using?  Anything before 3.02/00 is bad! 
    * What BRAT version are you using? 
    * Are you using User defined classes? 
    * How does you job look like?  I.e., what's in your configuration
      script? 
      
    > Hope that someone stumbles across the source of the leak one day.
    
    Something as serious as memory leaks is not something to be put off to
    another day - it needs fixing here and now! 
    
    Why don't _you_ do a more serious attempt to find the problem?  Like
    running with a high debug level to get the print out from
    TObjectTable, run your job in GDB, compile BRAT without optimisation
    and debuging enabled, compile ROOT wit debugging symbols (if needed),
    in that order.  
    
    It's good that you alarm us, but even better if you make an effort to
    try to find the error - you have jsut as much time to do this as the
    rest of us; that is, no time.   
    
    Yours, 
    
    Christian Holm Christensen -------------------------------------------
    Address: Sankt Hansgade 23, 1. th.           Phone:  (+45) 35 35 96 91 
             DK-2200 Copenhagen N                Cell:   (+45) 28 82 16 23
             Denmark                             Office: (+45) 353  25 305 
    Email:   cholm@nbi.dk                        Web:    www.nbi.dk/~cholm
    



    This archive was generated by hypermail 2b30 : Wed Oct 31 2001 - 14:12:42 EST