Re: JacORB vs ORBit again

[Alejandro Imass <aimass aimass org>]

Paco Moya wrote:

> ORBit does not understand object references returned by method invocations
> to JacORB server (Big endian converting).
>

Hi. I just went through a similar experience and I found that JacORB was creating an UNEVEN number of bytes. Later I found that all IOR's should be of EVEN size.

It's ver easy to hack the ORBit code. In orb.c you will find the function CORBA_ORB_object_to_string(bla...)
here you will find a big switch and a for loop that decodes the string. just open a file here and start fprintf to see all the decision making in the switch, and you will come up with the answer quickly.

I included this right after all the variable declaratins in this function:
 // The Brain Police
 FILE * ofp;
 char* retval2;
 ofp = fopen("/home/aimass/libORBit-debug.txt","w");
 fprintf(ofp,"*** Start of Debug File *** \n", retval2);


In my case it was one of the first checks (the EVEN check):

 if((len % 2) || len <= 4) {
  CORBA_exception_set_system(ev, ex_CORBA_MARSHAL,
        CORBA_COMPLETED_NO);


 // The Brain Police
 fprintf(ofp,"Was NOT EVEN LEN \n", retval2);

  return(CORBA_OBJECT_NIL);
 }


Remember to close your file here:
 // The Brain Police
 fprintf(ofp,"Returning a supposedly valid retval \n", retval2);

 retval = ORBit_create_object_with_info(profiles, type_id, orb, ev);


 error_out:


 // The Brain Police
 fprintf(ofp,"Actual Return \n", retval2);



 // The Brain Police
 fclose(ofp);

 return retval;


It is probably a very unexpert way to debug this but it solved my problem...

What I did was to recompile ORBit in my home directory and copied my hacked liORBit.so.0.4.0 (I'm using ORBit 0.5.3) to /usr/lib. Remember to back up your original lib first. I think if this is a common thing w/ JacORB, perhaps we should try to see if we can fix JacORB or let ORBit be more tolerant to JacORB IORs.



Best Regards,
Alejandro Imass