Hi Francois-Xavier,
I do not see any problems with your classes. I could not use your
Makefile,
I did the following:
rootcint -f cint.cpp -c TMuster.h TSecond.h LinkDef.h
g++ -fPIC -I$ROOTSYS/include -c cint.cpp TMuster.cpp TSecond.cpp
g++ -g -Wl,-soname,libMuster.so -shared cint.o TMuster.o TSecond.o -o
libMuster.so
then in a Root session
root > gSystem.Load("libMuster");
root > TSecond *s;
root > gM = new TMuster(5);
root > s = new TSecond();
fK = 5
Rene Brun
On Thu, 11 Apr 2002, GENTIT Francois-Xavier DAPNIA wrote:
> Dear rooters,
> I have a problem when using global pointers instantiated within CINT and
> then used in the compiled part of my code, when I create a .dll or .so. The
> problem nevers appear when I compile my code with a main program and create
> a .exe. Sometimes, not always, the pointer defined by a "new" inside the
> CINT macro appears inside the compiled code to be equal to 0.
>
> The pointer is correctly declared in the Linkdef.h file with:
>
> #pragma link C++ global gM;
>
> Is it a rule that when doing a .dll or .so, it is forbidden to use global
> pointers whose memory is allocated within CINT?
>
> I join to this mail a very small program to illustrate my point. This
> small program contains 2 (almost empty) classes TMuster and TSecond, plus a
> global pointer gM on TMuster.
> The following CINT macro
> {
> gSystem.Load("libMuster.dll");
> TSecond *s;
> gM = new TMuster(5);
> s = new TSecond();
> }
>
> does the following : the global pointer gM is used inside the constructor
> of TSecond. On Windows, everything works ok, on Linux gM is equal to 0
> inside the constructor of TSecond. I have also cases where the problem
> appears on Windows, unfortunatly, I was not able to condense the problem
> appearing on Windows into a small program.
>
> I am using ROOT 3.02/07, Windows NT or Linux.
>
> Thanks a lot if you can help me.
>
> F.X. Gentit
> DAPNIA/SPP CEN Saclay
> tel : 01 69 08 30 38 fax : 01 69 08 64 28
> web : http://home.cern.ch/~gentit/
>
> <<Problem.ZIP>>
>
This archive was generated by hypermail 2b29 : Sat Jan 04 2003 - 23:50:49 MET