RE: [ROOT] on "almost-uniqueness" of _R__UNIQUE__

From: Philippe Canal (pcanal@fnal.gov)
Date: Tue Feb 18 2003 - 14:08:59 MET 

Hi Alexei,

In root 3.03.07 and above, the macro ClassDefT2 is NOT needed (and actually
should not have any adverse effect).  You should have problem with
_R__UNIQUE_ only if you use the macro ClassImp in an header file (which is
not usual).  If you do need to have ClassImp in a 'header' file or
equivalent and you have a clash, you should use
ClassImpUnique(X,uniqueIdentifier) instead.

Cheers,
Philippe

-----Original Message-----
From: owner-roottalk@pcroot.cern.ch
[mailto:owner-roottalk@pcroot.cern.ch]On Behalf Of Alexei Dvoretskii
Sent: Tuesday, February 18, 2003 12:37 AM
To: roottalk@pcroot.cern.ch
Subject: [ROOT] on "almost-uniqueness" of _R__UNIQUE__


I am compiling a $(PACKAGE)Cint.cc class into a shared library. (__CINT__
macro undefined)

Suppose $(PACKAGE)Cint.h has

#include "A.h"
#include "B.h"

A has on line 50:

ClassDefT2(A,T)

B has on line 50:

ClassDefT2(B,T)

then there will be a conflict for a doubly defined variable
R__dummyholder50. I encountered this in 3.03-09 but it seems to be there
for the CVS HEAD as well.

_R_UNIQUE__ is defined in RConfig.h

/* produce an identifier that is almost unique inside a file */
#ifndef __CINT__
#   define _R__JOIN_(X,Y) _NAME2_(X,Y)
#   define _R__UNIQUE_(X) _R__JOIN_(X,__LINE__)
#else
    /* Currently CINT does not really mind to have duplicates and     */
    /* does not work correctly as far as merging tokens is concerned. */
#   define _R__UNIQUE_(X) X
#endif

This archive was generated by hypermail 2b29 : Thu Jan 01 2004 - 17:50:09 MET