Subject: Re: Handling errors in handle transfer functions To: William Gropp Cc: mpi-core@XXXXXXXXXXX,owner-mpi-core@XXXXXXXXXXXXX X-Mailer: Lotus Notes Release 5.0.5 September 22, 2000 From: "Richard Treumann" Date: Mon, 4 Jun 2001 18:02:25 -0400 X-MIMETrack: Serialize by Router on D01ML077/01/M/IBM(Release 5.0.7 SPR #MIAS4UTJ8H &S/390 SPR #JHEG4V8UT5|April 5, 2001) at 06/04/2001 06:03:26 PM MIME-Version: 1.0 Content-type: text/plain; charset=us-ascii The standard already says that the behavior is: an invalid handle in causes an invalid handle out. Take C to F conversion: To me that means that if you have a C handle that would trigger an specific error on being passed to a C routine, you should get back a Fortran handle that would yield an equivelent error when passed to an equivelent Fortran routine. Nothing more and probably nothing less. In PE/MPI the C & Fortran handles are no different and these functions should not need to be more than macros that wrap an assignment, I see no reason to go further in defining the behavior of these functions but if somebody feels a need to clarify then this is the clarification I would support. Dick Treumann RS/6000 SP Development IBM Poughkeepsie Unix Development Lab Dept 0lva / MS P963 -- 2455 South Road -- Poughkeepsie, NY 12601 Tele (845) 433-7846 Fax (845) 433-8363 William Gropp cc: Sent by: Subject: Handling errors in handle transfer functions owner-mpi-core@mpi -forum.org 06/04/2001 04:23 PM What should a high-quality implementation do if an invalid handle is given to one of the transfer functions (e.g., MPI_Comm_c2f)? I believe that the error handler on MPI_COMM_WORLD should be invoked (MPI_FILE_NULL for MPI_FILE handle transfers) because that is the specified error handler if no valid communicator/window/file is available. A user asked this recently; I'd like to propose this as a clarification. Bill