Hi Charles, It is difficult to come with a diagnostic. Could you try to link your application without the libNew.so library ? Why do you create TTree objects in the stack? Are you using 2.25/03 ? Rene Brun On Wed, 20 Sep 2000, Charles Leggett wrote: > > Has anyone seen this sort of a problem before? > > I'm integrating ROOT with Atlas and Gaudi software, and I have found > that if I link against the default ROOT libraries (I've tried 2.23, > and 2.25), which are compiled with -O2, I get very strange behaviour, > such as creating TTrees on the stack works fine, but creating them > on the heap causes their internals to be messed up. In desperation > I compiled my own root libraries, without the -O2 and with +g (BTW, > is there an easy way to do this? I had to hack the makefile) and the > problems went away. > > I could not reproduce the problem with simple, standalone programs. Only > when linking against the huge mess of code that is Atlas and Gaudi > (and these are not built optimized). Which use lots of dynamically loaded > libraries. > > I have no idea where the problem lies. Is it with the loading of the > shared libraries, a compiler bug, or something weird and esoteric. The > fact that removing the optimization caused the problem to vanish leads > me to believe that it might be compiler related. > > This is all under Linux (RH 6.1), with a 2.2.12-20smp kernel, and egcs-1.1.2. > > cheers... Charles. > > -- > +---------------------------------------------------------------------+ > | Charles Leggett | <CGLeggett@lbl.gov> | > | Lawrence Berkeley National Lab | HCG / NERSC : Atlas / D0 | > | 1 Cyclotron Road, MS 50B-3238 | | > | Berkeley, CA 94720 | Eagles may soar, but weasels don't | > | (510) 495-2930 room: 50B-3201 | get sucked into jet engines. | > +------------------- http://annwm.lbl.gov/~leggett -------------------+ >
This archive was generated by hypermail 2b29 : Tue Jan 02 2001 - 11:50:33 MET