Hi Christian, Your case was effectively not taken into account. Should be OK now in the CVS version. Let me know if you still have problems. Thanks for the detailed report. Rene Brun Christian Holm Christensen wrote: > > Hi ROOT'ers, > > I'm a bit puzzled about a when the TStreamerInfo object is written to > disk. The documentation says: > > Instead of generating the style of Streamer function shown above, a > shorter Streamer is generated. ... > > All the I/O is now managed automatically by TClass::WriteBuffer and > TClass::ReadBuffer. In turn these two functions invoke the services > of a new class TStreamerInfo that is the real I/O manager of the new > version of Root. > > So it seems like the TStreamerInfo object is written to disk when one > sends the message Write or indirectly when sending the Streamer > message. > > However, my problem is this: it seems that the TStreamerObject is > written to disk even if no object of that class is written to disk. > Here's a schematic of what I do: > > 1) Open a ROOT file (file1) and make a TTree in it. > 2) cd back to old TDirectory (dir1). > 3) Open another ROOT file (file2) and make histograms in it. > 4) Loop over events > a) Make objects of various non-ROOT - but derived from TObject, and > using the '+' option - classes. > b) Fill histograms in file2 > c) Fill TTree in file1 > 5) Write and Close file1 > 6) Write and Close file2 > 7) End of application. > > Now the idea of this, is that file1 will contain a TTree with all the > user defined classes, while file2 will contain diagnostics histograms > _and_nothing_else_. However, when I open file2 with the interactive > ROOT, ROOT complains with lines like: > > Warning in <TClass::TClass>: no dictionary for class <user class> is available > Error: class,struct,union or type <user class> not defined FILE:/tmp/filevBiSux_cint LINE:1 > > That is, somehow there's some information in that file pertaining to > the user defined class, though there's no objects of those classes in > the file. > > This is ofcourse not the intent. I found out, that by explicitly > changing directory to the file where the TTree is (file1) before doing > the Fill, everything works as it should. Schematically, I replaced > 4.c above with > > c) change directory to file1 > d) Fill TTree in file1 > e) restore old directory > > So it seems that some how, the TStreamerInfo object is written as a > consequnce of TTree::Fill, and it's written to current directory or > something. > > I have not been able to reproduce this in a simple way, so I guess you > have to take my word for it. My feeling is, that it has something to > do with the period flushing of buffers or something, but that's just a > thought. > > I think that the behaviour I outlined above is wrong (or I have > misundertood something), so I tought you sould know. > > Yours, > > Christian ----------------------------------------------------------- > Holm Christensen Phone: (+45) 35 35 96 91 > Sankt Hansgade 23, 1. th. Office: (+45) 353 25 305 > DK-2200 Copenhagen N Web: www.nbi.dk/~cholm > Denmark Email: cholm@nbi.dk
This archive was generated by hypermail 2b29 : Tue Jan 01 2002 - 17:50:47 MET