 |
 |
FYI - the corner case from CMS is now more thoroughly understood. It was due to an unexpected change in functionality for the TTreeCache (or a bug, depending if your the ROOT or CMSSW devs). In fact, I'm not sure how useful this would be. Prior to the filling of the TTC, ROOT will ask Xrootd to clear the contents of the cache. So, if we're seeing two consecutive ReadV calls of the same clients, then ROOT will have manually cleared the XrdClient cache anyway. This illustrates one of the weaknesses of the system - the interface into ROOT basically forces design choices that making multiple TTree's difficult for frameworks. Brian On Jan 31, 2011, at 2:50 AM, Lukasz Janyst wrote: > On Fri, Jan 28, 2011 at 9:33 PM, Brian Bockelman <[log in to unmask]> wrote: >> I've asked a few times for Elvin's patches - do you know what it takes to get them? > > I will send them to you later today. > >>> BTW. your solution of yesterday should do the job just fine + I have >>> found another bug there, could you test the attached patch? >>> BTW2. have you managed to test the forking? >> >> I had to hold off on those tests until we upgraded versions of the XrdClient. I now have those, and am trying to hunt down a few issues like this one. > > OK, let me know. > > Cheers, > Lukasz