Re: [ROOT] Segfault while iterating over file

From: Rene Brun (Rene.Brun@cern.ch)
Date: Fri Sep 06 2002 - 10:53:29 MEST


To ietrate on all keys of a Root file, use a normal iterator, eg:

void iloop() {
   TFile f("hsimple.root");
   TIter nextkey(f.GetListOfKeys());
   TKey *key;
   while ((key=(TKey*)nextkey())) {
      TObject *obj = key->ReadObj();
      printf("at key:%s, object
class:%s\n",key->GetName(),obj->ClassName());
      delete obj;      
   }
}

Rene Brun

On Fri, 6 Sep 2002, Jeroen Hegeman wrote:

> Yes, I first tried 'delete currentObject', 'currentObject->Delete()' was 
> an attempt to solve the problem.
> Actually, the problem is in the GetObject() call, even after removing all 
> the delete lines, it still segfaults in that line (in the third run after 
> starting ROOT, in the second iteration, to be precise).
> 
> And it looked like such a simple task :(
> 
> Jeroen
> 
> On Thu, 5 Sep 2002, Valeri Fine wrote:
> 
> > Hello I wonder if your problem came from the statement
> > 
> > >     currentObject->Delete();
> > 
> > This implies the current object is in possesion of the "proper" method
> > "Delete()"
> > One needs very strong reason and solid knowledge to use this method.
> > 
> > Did you try
> > 
> >        delete currentObject;
> > 
> > instead  as it is suggested by
> > http://root.cern.ch/root/htmldoc/TFileIter.html#TFileIter:description ?
> > 
> > On other hand you may not test
> > 
> > >   if (currentObject) {
> > >     currentObject->Delete();
> > 
> > because TFileIter  checks this itself.
> > 
> > Just "delete currentObject;" should be sufficient.
> > 
> > But . . .
> > 
> > I'd like to mention one should not try to "delete" the histogram objects
> > from the "read-only" file.
> > This may cause you some extra troubles as well. Such kind of the object will
> > be deleted
> > automatically just TFile dtor is called
> > 
> > "deleting" the histogram object read from the "writable" TFile would delete
> > the object from the file.
> > So if your object is derived from TH1 class then  leave it alone, don't try
> > to delete it.
> > (see: http://root.cern.ch/root/htmldoc/TObject.html#TObject:InheritsFrom )
> > 
> > Hope this helps,
> >                  Cheers, Valeri
> > 
> > 
> > ----- Original Message -----
> > From: "Jeroen Hegeman" <jhegeman@nikhef.nl>
> > To: <roottalk@pcroot.cern.ch>
> > Sent: Thursday, September 05, 2002 10:38 AM
> > Subject: [ROOT] Segfault while iterating over file
> > 
> > 
> > > Hi everyone,
> > >
> > > While trying to loop over all objects in a .root file (in order to dump
> > > all
> > > histograms at once), I seem to have an unexspected problem. The below code
> > > results in a 'segmentation fault' every second or third run.
> > > The problem has nothing to do with the assignment to currentObject because
> > > the commented line in the code does the same (apart from leaking...).
> > > Replacing GetObject() with GetKeyName() gives no problems, the iterator is
> > > fine and when it does run, the values that come out are right.
> > > I have looked through roottalks and tried debugging but I must be missing
> > > something, anybody have a clue?
> > >
> > > Thanks in advance,
> > > Jeroen
> > >
> > > {
> > > // Load the library we need for TFileIter
> > > gSystem->Load("libStar");
> > >
> > > TObject* currentObject = NULL;
> > > TFile* inFile = new TFile("hsimple.root","READ");
> > > TFileIter* iter = new TFileIter(inFile);
> > >
> > > Int_t nrOfObjects = iter->TotalKeys();
> > >
> > > // Loop over all objects in inFile.
> > > for (iter->Reset(); (int)(*iter) < nrOfObjects; iter->SkipObjects()) {
> > >   cout << "\ngoing to extract object #" << (int)(*iter) << "\n";
> > >
> > >   //  cout << iter->GetObject() << "\n";
> > >   if (currentObject = (iter->GetObject())) {
> > >     cout << "current object extracted\n";
> > >   };
> > >
> > >   if (currentObject) {
> > >     currentObject->Delete();
> > >   };
> > > }; // (End of object-loop.)
> > >
> > > if (inFile->IsOpen()) {
> > >   inFile->Close();
> > > };
> > > delete inFile;
> > > };
> > >
> > >
> > 
> > 
> 



This archive was generated by hypermail 2b29 : Sat Jan 04 2003 - 23:51:08 MET