Hi, Many thanks for your replies, > replacing the line: > X11.FindBestVisual: yes > by > X11.FindBestVisual: no ... didn't help ... (Hi, Jiri, I also had the idea of checking system variables, but nothing really interesting was there.) In the meantime ... I got more experience ... if I have a "big window" and I many times try ".class TVector" then sometimes it works properly and sometimes it breaks ... no idea what the problem is (I also tried to set X11.Sync to yes, but it didn't help). Well, maybe it is a bug in RH7.1 system libraries (just a precaution ... the system has all RH7.1 updates applied ...) ??? > For your second problem, replace the line: > #if !defined (__CINT__) || defined (__MAKECINT__) > by > #if !defined (__CINT__) I'm sorry that I did not tell you in my previous mail how to get around this problem. Of course, if rootcint does not see "Api.h" it does not break ! What is more interesting - one can completely remove the line "#include <Api.h>" and the code will still compile !!! While working on this problem, however, I noticed that there is an additional problem that I missed in 3.02/07 ... Let's look at the history ... the whole thing concerns TF77 ... pre 2.25/03 - I'm not sure when Masa implemented G__isinterpretedp2f, but it WAS WORKING like a charm 2.25/03 - problems with G__INTERPRETEDFUNC - a "brutal fix" implemented in TF77.cxx and [df]test.cxx, - "for (i = 0; i < 1; i++) sub(x, y, w);" automatically bytecompiles the "sub" 3.01/06 - "for (i = 0; i < 1; i++) sub(x, y, w);" NO LONGER automatically bytecompiles the "sub" when called inside of a macro 3.02/07 - "for (i = 0; i < 1; i++) sub(x, y, w);" AGAIN automatically bytecompiles the "sub", ... but ... - problems with G__BYTECODEFUNC discovered, no fix for now (I've just realized it today) 3.03/09 - "for (i = 0; i < 1; i++) sub(x, y, w);" AGAIN NO LONGER automatically bytecompiles the "sub" when called inside of a macro, ... plus - the same problems with G__BYTECODEFUNC ... plus - problems with Api.h Looking at the above history ... do you gently try to tell me something ??? For example that I should stay away from G__isinterpretedp2f & Co. ??? Best regards, Jacek. P.S. Of course, you might get curious what G__BYTECODEFUNC problems exist in 3.02/07+ .... well, get a 3.02/07 (mine is on RH6.2, egcs-1.1.2), then unpack the TF77.tar.gz, then : cd TF77 export CERNLIBDIR=/cernlib_home/2001/lib gmake ./root.exe root [0] .x dtest.cxx *** Break *** segmentation violation Root > Function dtest() busy flag cleared root [1] .q The "sub" got bytecompiled inside of the "dtest()", and ... *** Then edit dtest.cxx and comment out the line number 106, then : root [0] .x dtest.cxx This time "sub" is always interpreted (slow) and everything works ... Jacek.
This archive was generated by hypermail 2b29 : Sat Jan 04 2003 - 23:51:12 MET