Hi Volker, I know that TArray works. I repeat that abstract classes should be OK in this development version. There are many combinations where abstract classes can be found: - directly in derived classes - same with multiple inheritance - by composition - in templates - all case combinations above for classes in TClonesArray - same for split/no split mode In general, when ROOT finds an abstract class during I/O, its dictionary TStreamerInfo has already been computed and this is not a problem. I was just curious to know if you had a special case. The only way to know is to test it with the development version or to send me a small test case. Rene Brun On Sat, 1 Dec 2001, Volker Hejny wrote: > Hallo Rene, > > On Fri, Nov 30, 2001 at 02:25:47PM +0000, Rene Brun wrote: > > Hi Volker, > > > > I have recently made changes in teh development version in CVS for situations > > like you describe. If you could provide a small test case corresponding to your > > exact situation, I could tell you if this new version fixes this problem. > > In principle, you could test whether TArrayI is processed correctly, because > our case is basically the same, with the difference that the array will > not be recreated if the size is reduced. The behaviour in browsing is the > same. > > > There are sometimes class constructions non trivial to support when > > an abstract class is found during the split mode. ROOT needs to build > > the dictionary for these classes. To build teh dictionary, an instance > > of the class has to be created. If the dictionary has not already been built > > when the abstract class is found, there is no way to make an instance. > > The difficulty is in detecting in the parent (real classes) this situation. > > Well, abstract base classes are a nice instrument to define interfaces, > but for classes forseen to be written into a tree in split mode, it might > be easy enough to avoid them. So if itīs tricky to support these cases, > a warning in the TTree documentation should do the job. > > Best regards, > Volker > > -- > Dr. Volker Hejny Tel: 02461/616853 ** > Institut f. Kernphysik Fax: 02461/613930 ** > ---------------------------------------------------------------- ** ** --- > Forschungszentrum Juelich GmbH, D-52425 Juelich ** >
This archive was generated by hypermail 2b29 : Tue Jan 01 2002 - 17:51:10 MET