From: treumann@XXXXXXXXXX X-Lotus-FromDomain: IBMUS To: mpi-core@XXXXXXXXXXX cc: nissej@XXXXXXXXXX Date: Fri, 10 Dec 1999 08:58:33 -0500 Subject: Re: Fwd: Re: MPI_GET_PROCESSOR_NAME \0 or not? Sender: owner-mpi-core@XXXXXXXXXXX I agree that the IBM implementation has this bug involving MPI_GET_PROCESSOR_NAME in Fortran. We had already identified and fixed the bug in our development stream and now expect to fix it in existing releases via PTF as well. My interpretation of the correct response is that the Fortran character array passed in must be of length MPI_MAX_PROCESSOR_NAME so the character array output should be that length, left justified & blank filled. The resultlen parameter should report how many characters represent the real result, not counting the blanks. I propose this be the clarification but, because I am not an expert at Fortran's treatment of character strings I could be persuaded Bill is correct. I would welcome a prompt consensus. Nil's had contacted IBM service about this problem and felt he was engaged with IBM in a debate about the standard. I do not think the service representative he dealt with intended to debate the issue, only to make sure the issue was well defined and plausible before bringing the question to me. I thank Nils for raising the issue and making his point effectively. Dick Dick Treumann IBM RS/6000 SP Development Tele (914) 433-7846 Fax (914) 433-8363 William Gropp @mcs.anl.gov on 12/09/99 04:06:58 PM Sent by: owner-mpi-core@XXXXXXXXXXX To: mpi-core@XXXXXXXXXXX cc: Subject: Fwd: Re: MPI_GET_PROCESSOR_NAME \0 or not? The following question has come up. I agree with the questioner; the '\0' shouldn't be written in the Fortran version because it is not part of the name, and resultlen is not described as specifying the valid part of the return result. I propose the clarification to MPI_GET_PROCESSOR_NAME: Only the first resultlen characters of name are modified by MPI_GET_PROCESSOR_NAME. Bill