Hi Rudiger, As you have found out, Root assumes a working environment with shared libs and a plug-in manager. It should be quite simple to run in pure statically linked mode with no references to external files and with a reduce functionality: no calls to the interpreter requiring access to include files, etc. We are willing to add a run option to support this mode of operation such that you do not get warnings/errors in case of missing files at initialisation time. Since you have successfully run in this mode, did you find all the places in the code that should be protected/modified ? Rene Brun Rüdiger Berlich wrote: > > Hi there, > > I'm using a program that has some compiled-in Root functions > (such as booking and filling histograms, interfacing to trees > -- or rather : chains, and integration of histograms). No graphics > output is done and I do not invoke the interpreter itself (i.e., I > never call TApplication or TRint). My assumption is now that > I'm dealing with only compiled functions, and that there is no > hidden usage of the interpreter. If so, I do not need > the header files and libraries associated with cint. Correct ? > The program is linked with "-static -lRoot -ldl" > > I cannot use a dynamically linked executable, as the target platform > (indeed a cluster, as this is an MPI application) doesn't have the > correct C-libraries installed. I work and compile on a SuSE 8.2 system > and run the programs on a Red Hat 7.2 cluster. I cannot compile > my programs on a cluster machine, as only g++ with a version > > 3.2 seems to be able to handle my application. So I need a static > executable. > > If I start my program, each instance of the MPI application emits the error : > > Error: cannot open /opt/root/share/root/cint/MAKEINFO > !!! There are examples of MAKEINFO files under > /opt/root/share/root/cint/platform/ !!! > Please refer to these examples and create for your platform > dlopen error: /tmp/iJtfkg_cintNM: invalid ELF header > Load Error: Failed to load Dynamic link library /tmp/iJtfkg_cintNM > > as if cint was started somewhere (which it isn't explicitly). What's > going on here ? And how can I stop cint from being used when > I only want to fill a histogram from a tree and integrate the histogram ? > > The program doesn't exit, though, and henceforth emits the > correct results coming from the integration of the histograms > (checked on the system I compiled the application on, > which has root/cint below /opt/root -- no error message on that > machine). > > What are the effects of this error ? > > On a related note : I recently asked a question regarding the > integration of root/cint as an extension language into another application. > This works like a charm. There is however a similar problem to those > mentioned above. When the interpreter starts it tries to load various dynamic > libraries. If my application was compiled on a machine with incompatible > symbols (again this problem arises when using SuSE 8.2 for compilation > and Red Hat 7.2 for execution) I get errors when the dynamic > libraries are loaded. I guess there is nothing I can do about that. > > In any case : I really love this tool. Root is great - thanks a lot !!! > > Kind Regards / Mit freundlichen Grüßen, > Rüdiger Berlich > > -- > Institut fuer Experimentalphysik I > Ruhr-Universitaet Bochum > Tel. +49 (0)7247 825678 > ruediger@berlich.de
This archive was generated by hypermail 2b29 : Thu Jan 01 2004 - 17:50:12 MET