gRandom access in SVN HEAD

Second try sending this ... hope it makes it through without
duplication]

With the new handling of MathCore something seems to have
gotten lost
in the way it handles gRandom.  Now I see:

    *******************************************
    *                                         *
    *        W E L C O M E  to  R O O T       *
    *                                         *
    *   Version   5.19/01   17 January 2008   *
    *                                         *
    *  You are welcome to visit our Web site  *
    *          http://root.cern.ch          
 *
    *                                         *
    *******************************************
ROOT 5.19/01 (trunk22548, Mar 10 2008, 00:33:04 on linux)
CINT/ROOT C/C++ Interpreter version 5.16.29, Jan 08, 2008

root [0] gRandom->Rndm()
Error: Symbol gRandom is not defined in current scope 
(tmpfile):1:
Error: Failed to evaluate gRandom->Rndm()
*** Interpreter error recovered ***
root [1] gSystem->Load("libMathCore");
root [2] gRandom->Rndm()
(Double_t)9.99741748906672001e-01

whereas in prior versions (e.g. v5.18/00a) the initial
statement worked.
Previously libCore.so had a reference to gRandom but it no
longer does
and so it fails to automatically load the libMathCore
library (because
it's not trying to resolve a TRandom class name and it knows
nothing about
"gRandom").  This is true for both Linux (SLF 3
& 4) and Mac OS X (Tiger).

Are users going to have to add
gSystem->Load("libMathCore") to their 
scripts or is there some means to get the previous behaviour
restored?
I don't fully understand the .rootmap mechanism, but would
it be possible
to use that to get the library automatically loaded for
these sort of
globals?

-robert

Robert W. Hatcher   | rhatcherfnal.gov   630-840-3102
FNAL CD/EXP (MINOS) | MS 220, PO Box 500, Batavia IL 60510

Robert,

The CINT autoload mechanism works only with class names, not
with 
globals or typedefs.
the following works
   TRandom r;
   r.Rndm();

but not gRandom->Rndm().

We are expecting to remove this limitation in a future
version of CINT.

Rene Brun

Robert Hatcher wrote:
> Second try sending this ... hope it makes it through
without duplication]
>
> With the new handling of MathCore something seems to
have gotten lost
> in the way it handles gRandom.  Now I see:
>
>    *******************************************
>    *                                         *
>    *        W E L C O M E  to  R O O T       *
>    *                                         *
>    *   Version   5.19/01   17 January 2008   *
>    *                                         *
>    *  You are welcome to visit our Web site  *
>    *          http://root.cern.ch          
 *
>    *                                         *
>    *******************************************
> ROOT 5.19/01 (trunk22548, Mar 10 2008,
00:33:04 on linux)
> CINT/ROOT C/C++ Interpreter version 5.16.29, Jan 08,
2008
>
> root [0] gRandom->Rndm()
> Error: Symbol gRandom is not defined in current scope 
(tmpfile):1:
> Error: Failed to evaluate gRandom->Rndm()
> *** Interpreter error recovered ***
> root [1] gSystem->Load("libMathCore");
> root [2] gRandom->Rndm()
> (Double_t)9.99741748906672001e-01
>
> whereas in prior versions (e.g. v5.18/00a) the initial
statement worked.
> Previously libCore.so had a reference to gRandom but it
no longer does
> and so it fails to automatically load the libMathCore
library (because
> it's not trying to resolve a TRandom class name and it
knows nothing 
> about
> "gRandom").  This is true for both Linux (SLF
3 & 4) and Mac OS X 
> (Tiger).
>
> Are users going to have to add
gSystem->Load("libMathCore") to their 
> scripts or is there some means to get the previous
behaviour restored?
> I don't fully understand the .rootmap mechanism, but
would it be possible
> to use that to get the library automatically loaded for
these sort of
> globals?
>
> -robert
>
> Robert W. Hatcher   | rhatcherfnal.gov   630-840-3102
> FNAL CD/EXP (MINOS) | MS 220, PO Box 500, Batavia IL
60510
>

On Mar 10, 2008, at 4:00 PM, Rene Brun wrote:

> The CINT autoload mechanism works only with class
names, not with  
> globals or typedefs.
> the following works
>   TRandom r;
>   r.Rndm();
>
> but not gRandom->Rndm().

Yes.  I just wondered if there was some way to coerce CINT
to
treat special entries in a .rootmap file such that it
handles
globals (e.g. gRandom) in the same way as actual classes.
I guess you're saying "it's not possible".

> We are expecting to remove this limitation in a future
version of  
> CINT.

If that isn't to happen all that soon, could we in the
interim just
have root.exe linked with "mathcore" by default.

-robert

> Rene Brun
>
> Robert Hatcher wrote:
>> Second try sending this ... hope it makes it
through without  
>> duplication]
>>
>> With the new handling of MathCore something seems
to have gotten lost
>> in the way it handles gRandom.  Now I see:
>>
>>    *******************************************
>>    *                                         *
>>    *        W E L C O M E  to  R O O T       *
>>    *                                         *
>>    *   Version   5.19/01   17 January 2008   *
>>    *                                         *
>>    *  You are welcome to visit our Web site  *
>>    *          http://root.cern.ch          
 *
>>    *                                         *
>>    *******************************************
>> ROOT 5.19/01 (trunk22548, Mar 10 2008,
00:33:04 on linux)
>> CINT/ROOT C/C++ Interpreter version 5.16.29, Jan
08, 2008
>>
>> root [0] gRandom->Rndm()
>> Error: Symbol gRandom is not defined in current
scope  (tmpfile):1:
>> Error: Failed to evaluate gRandom->Rndm()
>> *** Interpreter error recovered ***
>> root [1]
gSystem->Load("libMathCore");
>> root [2] gRandom->Rndm()
>> (Double_t)9.99741748906672001e-01
>>
>> whereas in prior versions (e.g. v5.18/00a) the
initial statement  
>> worked.
>> Previously libCore.so had a reference to gRandom
but it no longer  
>> does
>> and so it fails to automatically load the
libMathCore library  
>> (because
>> it's not trying to resolve a TRandom class name and
it knows  
>> nothing about
>> "gRandom").  This is true for both Linux
(SLF 3 & 4) and Mac OS X  
>> (Tiger).
>>
>> Are users going to have to add
gSystem->Load("libMathCore") to  
>> their scripts or is there some means to get the
previous behaviour  
>> restored?
>> I don't fully understand the .rootmap mechanism,
but would it be  
>> possible
>> to use that to get the library automatically loaded
for these sort of
>> globals?
>>
>> -robert
>>
>> Robert W. Hatcher   | rhatcherfnal.gov 
 630-840-3102
>> FNAL CD/EXP (MINOS) | MS 220, PO Box 500, Batavia
IL 60510
>>
>

Robert W. Hatcher   | rhatcherfnal.gov   630-840-3102
FNAL CD/REX (MINOS) | MS 220, PO Box 500, Batavia IL 60510