Re: [ROOT] TBranchClones and TClonesArray::RemoveAt

From: Manuel Sánchez García (M.Sanchez@gsi.de)
Date: Fri Jul 05 2002 - 19:14:58 MEST


Hi Rene,

Now I'm confused. You tell me I _should_ call RemoveAt/Compress for a
TClonesArray in a Tree; but this is what I do and according to the
reasoning below It cannot work.

Or was it a typo and your advise is I _shouldn't_ use RemoveAt/Compress
for a TClonesArray in a Tree?. I just want to make sure.

If I _should_ call RemoveAt/Compress. I would appreciate an explanation
about why my logic below is wrong.

Thanks for the quick answer, 
	Manuel


El vie, 05-07-2002 a las 17:25, Rene Brun escribió:
> Manuel,
> 
> You should call RemoveAt/Compress for a TClonesArray in a Tree.
> This is the assumption behind ExpandCreateFast.
> This should be made clear in the doc.
> 
> Rene Brun
> 
> 
> On 5 Jul 2002, Manuel [ISO-8859-1] Sánchez García wrote:
> 
> > Hi,
> > 
> > Could anybody tell me what's wrong with the following reasoning?
> > 
> > Let's say we have TClonesArray in a TTree. In my eventLoop I call
> > RemoveAt and then Compress(). Now, RemoveAt() doesn't free the memory in
> > fKeep->fCont[i] but calls the destructor; now the virtual table pointer
> > points to that of one TObject.  (This can be checked directly on the
> > interpreter)
> > 
> > Now, next event comes. In the TBranchClones::GetEntry a
> > ExpandCreateFast() is done; this expands the TClonesArray and prepares
> > it to dump the data from the branch. Now for each element 'i' in the
> > array 2 things may happen
> > 
> > 1.- fCont[i]==0  && fKeep->fCont[i]==0   => the object is created and
> > the default constructor called. This is fine
> > 
> > 2.- fCont[i]==0 && fKeep->fCont[i]!=0   -> The constructor is not
> > called; but if the object's destructor was called in advance the virtual
> > table is not correct.
> > 
> > This problem doesn't show up when only TClonesArray::Clear() is done
> > because that doesn't involve calling destructors.
> > 
> > Is this a bug or am I missing something?
> > 
> > Best regards,
> > 	Manuel
> > 
> 



This archive was generated by hypermail 2b29 : Sat Jan 04 2003 - 23:50:59 MET